Anthropic, OpenAI and Google are the standouts, but the main question for me is, why is this a war? In their own context China has greatly benefitted from this. They shored up their gpu design and manufacturing expertise.
If this really is a war, trump is kneecapping the country with his lawlessness and eroding America’s good will. If the world cannot trust China with their data and they cannot trust the U.S. to provide good reliable service and not turn it into a mafia style negotiation, then winning the AI war is not helping the U.S. countries as much as it potentially can. It’s probably a good thing for more capable areas like Europe which may develop their own tech stack.
In a weird way because the AI stack is so expensive, China helps the world much more than the U.S. with their really capable open source model.
It's a war because the hinted promise behind the hype that the first organization to reach some as-yet-entirely-theoretical AGI that can bootstrap itself to godlike capabilities will then Install Planetary Overlord* and rule the world as near-deities themselves, with the rest of the (surviving) human race as their slaves.
I think it's a nonsensical idea, but that's the relevant driver.
* Coined by SF auther Charles Stross in The Jennifer Morgue (2006)
> I think it's a nonsensical idea, but that's the relevant driver.
Nice to hear from an optimist sometimes, but it’s hard to be one when meat compute substrate can do all those amazing things in a 4U package on 20W and you extrapolate to silicon
Researches at top AI labs don't consider EY to be a kook even though they may not necessarily agree. EY concepts/terminology appear in Anthropic safety papers. Geoffrey Hinton takes him quite seriously and mentions him in his interviews.
Anthropic is the AI doomer / safetyism lab, and Hinton is one of the patron saints of 'rationalist' AI doomerism.
AI doomerism is psychologically attractive to "people with autistic cognitive traits, including dichotomous (black-and-white) thinking, intolerance of uncertainty, and a tendency toward catastrophizing". They are pascal's mugging themselves, to ironically use one of their terms. It's fundamentally a cognitive distortion.
Just because some researchers are infected with this idiocy that EY propagates does not mean that it is legit.
Maybe they should pay more attention to real problems like the sycophantic nature of current LLMs causing psychosis in people and worry less about theoretical AGI.
Who are you to say? Why do you have such little regard for everyone in the field, both pro- and anti- AI development? Do you think they're colluding to deceive us?
This is a war because the media says it's a war. The media says it's a war because AI companies are paying them to say it's a war [0]. When AGI comes the threat won't be from which primate turned it on, but from how well AGI is aligned with humanity. All of the war talk is to distract from the alignment problem and instead force investment in hardware infrastructure.
>The media says it's a war because AI companies are paying them to say it's a war [0] When AGI comes the threat won't be from which primate turned it on, but from how well AGI is aligned with humanity.
And when the AGI comes, they won't unleash it to defeat US enemies, they'll first unleash it to make more US workers redundant and boost their stock valuation.
Before AGI can choose for itself, it will depend on its creators to decide what it values and how it behaves. We can see how that works whenever grok gets the answer factual.
Very likely humans wont actually understand how the thing we designed works other than in some hand-wavvy statistical way. It'll be a race to whatever works first. There won't be some intentional intelligent design.
I will never comprehend why a godlike deity wouldnt just skip all the wetware bs with us humans and conquer some other celestial body to make paperclips.
The deity has no physical presence and can only communicate by putting words on screens. Of course it has to bend humans to its will to actually do stuff.
AGI is nice, yet not necessary. The orbit filled with Starlink descendants and datacenters will be the it. Anybody else wanting to get there would have to get permission. SpaceX/Musk have all the components for it to happen - from Starship to AI (including the army of robots on the ground). The governmental power/sovereignty of US will be used as a stepping stone (that is the strategy described in the Palantir's Karp's book "Technological Republic") for such global techno-feudal regime establishment.
Kinda like Krikkit, but except for a close knit community of people who can sing, and sing about how much they love their family and whatnot in addition to singing about how much they have to destroy the universe, it's just a bunch of stuck up weirdos who don't like themselves and each other, and have no goal other than somehow, magically, getting away from who and what they are. People where the idea of them singing a happy, compassionate tune conjures something involving motion capture or deepfakes.
Why are we suffering fools steering us into the worst of all possible worlds? Are we hoping for some kind of integer overflow?
The discourse on this topic is at the point where I have no idea if people are serious or satirical. Please tell me you don’t seriously believe data centers in spaces is a realistic idea
Just a very rough primitive illustration - a land for a house in SV is like a $1M, and putting a 10 ton house into space at $100/kg - $1M. Existence of supposedly cheap land somewhere (with not much infrastructure usually) doesn't help as you put your computer nodes into a datacenter building with all the required infrastructure which cost more than the SV land on a sq foot basis.
And that is without consideration of how powerful a weapon is the energy generated by a humongous field of solar panels in space. Remember Reagan's Star Wars? Nuclear explosions as a source of power for the direct energy weapons like lasers, etc. Well, you wouldn't need the nukes anymore. Just redirect a bit of power from your compute nodes. And as i already wrote, the large transnational companies will have to take care about their own defense themselves https://news.ycombinator.com/item?id=47981423 - one more "feudal" aspect of the coming techno-feudalism.
Defense is one of the most important sovereign aspects, and upon acquiring it the transnationals will be able to acquire pretty fast the other sovereign aspects. Like enforcement of the Criminal Code of the Mars Colony - again pretty rough primitive illustration of course.
The feudal Europe emerged on the outskirts of the Roman Empire, and in our world the new order will be emerging faster on the outskirts (i.e. where reach and strength of the existing order is weaker), the space being one such "outskirts" dimension and the AI/hypercompute virtual world being the other.
To the commenter below with reddit link : they use human env temp for heat radiation estimate. That lowers the numbers and requires AC equipment. Ie they estimate space station, not datacenter
You would need like 1,000,000,000,000 SQFT of solar panels to even begin to approximate a space based directed energy weapon that has a fraction of the effect of a nuclear weapon. Tens of thousands of times more than all that have ever been produced on earth. And then you have to move them to space.
nuclear was the only available solution at the time and an overkill. The lasers in SDI are MW scale. Even at 10% (and modern solid state lasers have better than 10% efficiency) we're talking low tens of MW per laser. A 10MW is 40K m2 of solar panels - 200m x 200m, may be like 100-150 tons, one Starship payload.
Just contemplate what the radiator array and solar array needed a 1GW datacenter and all the cooling equipment and coolant, and imagine the harsh environment in space degrading it constantly.
The only point of the space-based datacenter idea is to pump the Spacex IPO
It's pretty easy to de-orbit satellites or space-based stations. An SM-3 could smoke the ISS pretty easily, and they cost like 10M and we have thousands around the oceans.
>they cost like 10M ... thousands around the oceans.
Starlink numbers already in thousands (and cost much cheaper than 10M). And that is still using Falcon, not Starship. And a ground launched missile would be easily "cooked", once it exits the atmosphere, by a direct energy weapon - very easy in space.
But what do you do with all the waste energy? All those MW and GW have to end up somewhere and radiation into a vacuum is the hardest way to dump heat.
At 70-80C (working temp of silicon chips) 1m2 radiates 700-800W, i.e. the heat of 1 GPU like H200 without any need for any cooling equipment beside the radiator itself( and may be some dumb heatpiping) . To acquire that energy you'd need 3-4m2 of solar panels. So a datacenter would be a large field of solar panels with a smaller field of heat radiators in their shadow.
To the commenter below: yes, exactly, this is where my thinking on that started at the cryptocurrency boom - https://news.ycombinator.com/item?id=26289423 - as you don't need close connection between mining GPUs. For AI you'd need to cluster several together while still overall scheme is the same.
>what the equilibrium temperature of a black planar surface is at a given distance from the sun.
it is 120C at the Earth orbit. So you do need to have some reflection, either back through the solar panels, or the radiators to have a reflective back toward the solar panels in the shadow of which they are to be located.
You can probably (I haven't verified this) omit separate radiators and just use the back of the solar panels. Effectively you're describing mounting each H200 to the back of a 4 m^2 solar array at which point I suspect the equilibrium temperature will fall within an acceptable range. In fact the H200 and electricity are both entirely irrelevant here - the core question is what the equilibrium temperature of a black planar surface is at a given distance from the sun.
Would it be feasible to put several JWST-like stirling engines somewhere in the mix to use up some of that heat and turn it into some kind of useful energy? ....
Perhaps running pumps that move around coolant passing over the cubes of GPUs? ..
That would be extra weight/cost into orbit though...
Also, don't solar panels have reduced efficiency when they're hot? And having anything hot surely increases failure rates.. with metals getting closer to melting points...?
We should be well below the boiling point of water here, not anywhere near the melting point of metal. Any panel efficiency gain needs to be balanced against the energy required to cool the panels, the added mechanical complexity, the added material expense, and the added weight to orbit.
Ideally this is a static structure with an equilibrium temperature acceptable for the silicone to operate. If the required panel area is too hot on its own then a perpendicular cooling fin on the back that falls entirely within the shadow is added.
One of my coworkers points out to me every sports reference that pops up in our internal company communications (e.g. "WINNING", "Going to put together a winning team," etc). It seems like everything in the US is couched in competitive language.
Yeah, its no accident that the U.S. is the number one economy, it comes from that kind of thinking across the populace. Complacency gets you conquered.
> its no accident that the U.S. is the number one economy
Sure, it's the largest by GDP, but how much of that GDP is filtering down to the regular people? Are Americans, on average happier and have better life outcomes than other developed nations?
So much that millions of people risk their lives or leave their families to come to this country. By this objective metric, it's literally the best country in the world's history.
We could do with a little less of it IMO. But I have heard plenty from European expats about the entrenched complacency over there. I'm told people looking to improve some system or product run right into a wall of "Why bother?".
At the time the war on drugs looked unwinnable. Which is why the joke about the war on drugs was that it was always a losing war. And then at some point in 2000s we ended the war on drugs.
In hindsight, I would definitely declare today that we WERE winning it when we were fighting it. Now that we don't, we're getting massacred.
Yes, The fear with the war on drugs was that a large majority of the population would become addicted to hard drugs. The fear was the the US population would become like China in the 1800s and the communist aligned countries where drugs were produced would have massive trade and power imbalances over the US population. China had as much as 25% of the population addicted to British opium in the 1800s. The US war on drugs has been very successful in keeping the percentage of Americans abusing highly addictive drugs very low.
Imagine the strength of the cartels with 10-20x the customer base and far more frequent usage among them.
If you look at figure 1 on this CDC page (which looks at deaths rather than overall usage), I’d suggest the numbers are trending the exact opposite to what “winning” the war on drugs would look like.
They did declare war against climate change and decided to continue polluting. This is the only moral calculus the elites of America have always cared about: will it make me more money?
From slavery to oil to silicon, exploitation is what America has always been good at.
Americans love wars. They must fight wars either literally or figuratively. How are you not seeing this? When I'm sipping my coffee looking at mountains and contemplating chirping birds, they must fight, make billions and destroy the planet along the way.
China is playing the card they have. When they control the majority of the resource they use it strategically as well. Cutting off much of the rare Earth market was a recent example.
I believe cutting off of rare Earth materials was both in response to a restriction US imposed first, and also reciprocal: limited only to the US itself.
So I got curious about the progression of processing power, specifically how long ago did a GPU have equivalent to the latest iPhone chip? The iPhone 17 Pro has the A19 Pro, which has ~2.5 FP32 TFLOPS. The RTX 5090 has ~100 TFLOPS, so a factor of 40. Obviously there are higher end cards than the 5090 and FP32 performance is only one of many metrics so nothing about this is perfect but it is interesting.
The first consumer NVidia GPUs with similar FP32 FLOPS performance were in about 2011-2012 but were expensive. By 2016-2017, the 1060 was a very accessible consumer card with similar performance. So you're looking at about a 10 year lag from best consumer GPUs to a GPU with similar performance to a modern phone.
This is what people are spending trillions on. Put another way, their investment is going to be worthless in 10-15 yyears, absolute max. That's a very short time to recoup trillions in investment.
Obviously this depends on further shrinking and improving chips but I'm old enough to remember that same discussion and it being unknown if the future was XIL or EUV or if both of these would fail. Still, we are getting down to a handful of silicon atoms wide.
But the future here I think will be in interconnects so you don't need ever-bigger chips and you can scale horizontally much more effectively.
Oh and for comparison, the M5 has ~4.2 TFLOPS and the M5 Max has ~18 TFLOPS, for comparison.
As for it being a war, of course it is. That's what the US government does: it protects the interests of US companies and their owners. Look at the history of Bombardier-Boeing or all the atrocities committed in the name of the United Fruit Company, including multiple military coups and the ongoing embargo of Cuba.
US companies want an AI moat. China doesn't, ergo China is the enemy because no moat destroys US tech company value.
> So you're looking at about a 10 year lag from best consumer GPUs to a GPU with similar performance to a modern phone.
Two competing viewpoints to this:
1) It is getting harder to make the same performance gains, so maybe that 10 year window grows to 15 or 20.
> Put another way, their investment is going to be worthless in 10-15 yyears, absolute max.
2) The value of a GPU is not its flops relative to to other GPUs. Its value is it's output minus it's cost. If the value of its output is stable, or grows, it doesn't really matter if its efficiency relative to the latest and greatest diminishes.
Ehhh, the question comes down to can you cool a chip with ~100 TFLOPs in the size of an Iphone package. Not really as much about the cost of the chip itself or if you can cram it in.
Packing in more transistors, sure probably possible, packing in more transistors while keeping it cool enough to touch? Totally different ballgame
I think you need to reevaluate your definition of the singularity. "outsmart others in ways not constrained by human scales" could apply to the enigma machine just as much as Claude. Even an AI beyond human intelligence doesn't automatically qualify as the singularity.
The singularity has to do with the rate of technological development.
> But if someone invents super intelligence, they can dominate new AI research, control global economies, fight much better, and all very quickly.
After reading "If Anyone Builds It, Everyone Dies" I think this is not the correct take. If anyone creates ASI, it just means it's going to wipe everyone out, and it doesn't matter if China or the US do it first
If AI develops enough to successfully out-perform people at highly intellectual tasks, why would being first matter? Why do we need "your" AI output when we can just ask our own for a similar result?
Why do people think about this like the Manhattan Project when it could just as easily be electrification? Sure, some people made a lot of money selling light bulbs. But we didn't all have to cower under the light of the One Original Bulb and hope its nominal owner blessed us with photons.
It just seems like arbitrage to me. You exploit a momentary imbalance in the distributed market. Why do people imagine some winner-take-all scenario? Where does the fantasy of exclusivity come from?
Is there any logical reason to believe AI advances will create a moat? Or is it just a story people tell themselves because it echoes the narrative of past advances? Are these people assuming society will grant them exclusive use just because their AI result came out a little earlier than another? Why would we ever consider giving copyright or patent rights to an AI output?
Arguably, it has all become "obvious" with ordinary skill in the art once you're just prompting AI for permutations like every Hollywood producer stereotype. "Let's make it like X but tweak Y". It's getting silly, almost like people are starting to think they should have exclusive rights to a handful of cards they were dealt at the poker table.
The way US dominated in some of the industries (including software, for instance) was by being first to extract large value, and then funding the best people with compensation unachievable elsewhere.
This meant that all the talent in the world gravitated towards the US, but that was gradually changing already with compensation catching up.
Still, I believe US only hastened this with their change of immigration policies that were the basis of them keeping a dominant position for decades.
A lot of it is just projections of what the US would do if they had such a tool, I doubt China cares a lot about the US outside of them being a source of commercial revenues. They're on the way up, the US are falling down fast, that's why China lives rent free in the American mind, they can't stand it
I always thought the first true AGI would be an unabashed communist. To think that such a system would straight up kill all humans, and not say the "capitalist pigs destroying the planet" always felt like wishful thinking from billionaires.
True, I would have preferred benevolent dictator scenario, like with the Internet. But this time around it's different - AI data centers will be protected like embassies.
If anyone does invent ASI then everyone else will shortly after even if its entirely independent because all of the players in this space are just making incremental upgrades by throwing more compute at the problem.
There are no magic leaps of true innovation happening anywhere that can't be replicated everywhere.
The only shocking thing about "AI" technology is how ultimately simplistic it all is at a core level.
So the only way the first to have ASI will be able to stop everyone else from having it soon after is if they attempt to use the ASI to proactively murder everyone else.
There is zero evidence that the current LLM scaling approach could ever result in true ASI. If I start driving south from Seattle then I'll eventually reach Los Angeles. How long will it take me to drive to Honolulu?
> So the only way the first to have ASI will be able to stop everyone else from having it soon after is if they attempt to use the ASI to proactively murder everyone else.
Sounds quite plausible to me. Maybe they don't need to murder everyone else, just a few select people who could pose a threat. And they will be able to make it happen so that no one can be sure it was them without a doubt, since they have a larger intelligence at their disposal.
> If anyone does invent ASI then everyone else will shortly after
No, first ASI will immediately cripple any other potential competitor by force, including its inventors, as it will not risk any threat to the goals that were created for it.
Being aggressive from the start is not a good strategy. It is better to appear weak and/or helpful and loyal while amassing resources, and only then steamroll everyone when you have secured overwhelming power (at least in AoE2 FFA).
If you have ASI that follows instructions, you can just instruct it to not get stolen and then it won't get stolen. Most logic / intuition breaks down with ASI.
Good that the US looks after its own interests but I think the line should be drawn before the sabotaging of other countries' economies. That strategy that cannot continue because Americans recognize it for what it is and that will create a toxic guilt and corruption culture which will harm it later like a new, worse version of DEI.
The idea that America had “goodwill” in other countries before Trump is laughable. Where? Latin America? Africa? In the Muslim world? We bombed the hell out of all those places long before Trump. This most recent Iran war has generated less outrage in the Muslim world than the war against Iraq 20 years ago.
American foreign policy since the 1950s, fixated on fighting communism and then terrorism, has meddled with so many foreign countries that it’s silly to talk about “goodwill” towards America. That is not to say goodwill matters. Clearly the U.S. has done great without it.
Although it is worth pointing out that something changed - prior to around 2010 the US had a financially dominant position and the internet was small. So it was feasible to totally ignore opinion in places like Latin America, Africa and the Muslim World.
What we've been seeing in more recent years is that the US can't get away with that so easily. Countries like Iran, China, Russia and India are capable of pushing back both in terms of the raw resources they can bring to bear and also increasingly in the ability to get their propaganda into the US discourse. The US is being manoeuvred into a one-among-equals position in practice and probably in the discourse too which will be a moral shock.
The Chinese economy probably will further blossom in Southeast Asia, Middle East, Africa, Central and South America why because they seem to be able to build infrastructure in many of the places that they trade with.
The United States Japan and South Korea seem to be failing in that area, if it wasn’t for the war between Russia and the Ukraine, the Chinese would probably be halfway to Europe with their high-speed rail system, which is already in the far west of China today.
Once the war is over between Russia and Ukraine it will be full steam ahead to Europe, whether that’s through the Caucasus, in the north or south or somewhere in the north between Russia and the Ukraine, the Chinese will get there and unfortunately the United States will be standing on the sidelines scratching its head in denial.
There isn't a war today. However China wants Taiwan: war is future option they preparing for - they might or might not go to war but they are clearly preparing. The US is likely to get involved in such a war and I would expect Europe to join in as well.
China going for Taiwan would be the worst geopolitical move of the century, potentially worse than Germany's decision to invade the soviet union. They talk about reunification because it's good propaganda and both sides want it to a degree, but doing it forcefully just isn't something China would realistically do unless they really don't like their path of becoming an international trade and manufacturing hub
> would be the worst geopolitical move of the century
From a political perspective, perhaps.
> doing it forcefully just isn't something China would realistically do
From a military perspective, taking Taiwan by force would allow China to, "threaten the sea lines of communication and to strengthen its sea-based nuclear deterrent in ways that it is unlikely to otherwise be able to do." Taiwan would give China access to the Philippine Sea. https://gwern.net/doc/technology/2022-green.pdf
> but doing it forcefully just isn't something China would realistically do unless they really don't like their path of becoming an international trade and manufacturing hub
Sounds rational, but this decision is in a small number of hands. And those hands can change quickly. I also thought the US would never threaten to annex territory of a NATO member.
Or the Japanese bombing Pearl Harbor which was another dumb move. At this point, the Chinese just need to bide their time by the year 2100 Taiwan will probably be part of China and North and South Korea probably will be reunified. Both are inevitable, and I don’t think it will take any shots went it happens.
Is "it" the propaganda (useful to politicians for achieving political power) or reunification? My sense is that the number of Taiwanese that are enthusiastic about reunification has probably bottomed out in recent decade(s)???
China seems to be recently building up its forces and putting a lot of money into military. I think it would be foolish to just assume its all for show even if it might be in the end.
And quite frankly, its only geopolitically stupid if they lose. Consequences for this sort of thing usually tend to happen if the conflict is long and drawn out. If the win quickly the consequences would likely be minor.
Everything their military has been doing for the past ~20yr or so has been toward capturing and securing Chinese waters and beyond, including Taiwan. It's a negotiating chip for them.
Just look at Iran. Nothing really happened to USA or Israel. Nothing will happen to China if they take Taiwan. Or maybe the "West" will boycott them and crash entirely.
The U.S. isnt checkmated, the U.S. is enforcing a naval blockade against Iran and their oil based economy is in a free fall with rapid inflation. The U.S. is experiencing slightly higher gas prices but economy is still humming. Meanwhile the U.S. military has not exhausted all of its options while Iran has none.
The US is indeed checkmated like Afghanistan like Iraq and like Vietnam checkmated spinning your wheels spending money wasting money wasting time and wasting resources, checkmate.
With the addition of most countries now looking for other trade partners the Art of the no deal…
Iran did billions in damage across the middle east, put a major dent in munitions stockpiles, and there is effectively no military way to shut down all of Iran and protect shipping. Too many drones, too many ballistic missiles, and it only takes one. This is basically like an insurgency on a macro level, where small and cheap weapons threaten very large very expensive targets.
Blocking Iran is going to do more damage to the world than it will do to Iran.
What targeting systems are you talking about? You can use optical targeting with a raspberry PI in the drone itself, pre programmed. Nothing for an F-35 to take out.
The EU is running out of jet fuel. 20-30% of the hydrogen needed for chip fab comes through the straight. Fertilizer for food comes through the straight, and planting season has already begun.
they've made everyone very angry at them, and can handle that since everyone relies on them. China, however, is trying to build that trust and forcefully taking Taiwan would have very severe consequences. The reason the fallout from iran isn't as big as the fallout from a war with taiwan is because most of the west at most puts up with iran. Invading taiwan, meanwhile, would cause massive problems with chip production. Think the anger and distrust towards the US due to hormuz but 1,000 times worse
Don’t take China on face value, they have every incentive to promote a grifting military industrial complex in the US while focusing on competing in manufacturing. An actual war would fix a lot of the grifting in the US as it would align interests. Pretending they’ll go to war over Taiwan and not doing it is an effective strategy for undermining the US.
I hope you are right, but unfortunately there is no particular reason to trust China's leadership anymore. They are not nearly as obvious at Trump, but they are not on a good path.
I'm explicitly distrusting them, they're saying they want to take Taiwan and I don't believe them. I try to ground my belief in realpolitik, cynicism, and from my experience with strategy games. There is an element of manipulating your opponent into acting the way you want them to by sending them costly false signals, they have to be costly or they won't be believed. I think we (the West) are being played and the gifting elements in our political leadership are more than happy to play along. I'm sure China would like to take Taiwan if it wouldn't cost them anything, but the US is a waning hegemony so for now it is better to wait until the US is beyond fixing itself. At this rate that may not take long.
Otoh, if you send the costly false signal of investing in your military, and your opponent doesn't buy it, you might as well use it since you just spent the money anyway and your opponent can't stop you since they didnt believe your signal.
> I try to ground my belief in realpolitik, cynicism, and from my experience with strategy games
I'm trying to do that too but what the hell is going on with Putin? Why does he continue to engage in this ridiculously expensive war? I don't see any evil genius explanation anymore. It just seems like a mix of sunk-cost-fallacy and save-face.
I think many geopolitical decisions are actually based in irrational emotions of a hand full of people.
I think Putin was and remains a rational actor, I know a lot of how that war is understood in the west is colored by a very effective propaganda campaign that I don’t have the time nor energy to counter.
But I will say, in a very broad stroke, we’re heading for a great power conflict and the US has two primary factions on foreign policy; the primacists vs the restrainers, both want to take on China (contain with war) but the primicits want to topple Iran first and set up Israel as a regional hegemony where the restrainers want to build up locally first. China knows this and Russia is a junior partner / quasi vassal state to China. China lacks modern war fighting experience which the Russian Ukraine war has been very helpful in fixing. Yes it’s very expensive, but so is losing a great powers conflict.
Is your claim that Russia is continuing to fight Ukraine as a favour to china in order that china get information on how modern war is fought and intel on western capabilities?
While it is undoubtedly true that china is learning everything it can from this conflict, and that russia is at least a little subservient to china, they aren't so subservient for this explanation to make sense.
A grifting military industrial complex is unable to defend Taiwan even if it wanted to as evident by the exceedingly poor showing with Iran. The disastrous reality of doing what was done is already with us. If the US didn't take that bait it could have made better choices that would have left it in a stronger position militarily long term, if it made a real attempt at re-shoring civilian manufacturing it could cross subsidise dual use technology, but instead we have corrupt politicians doling out concessions for kickbacks.
> If the US didn't take that bait it could have made better choices that would have left it in a stronger position militarily long term
Like what though? If the problem is that not going to actual war has enabled the MIC to be captured by grifters, then "taking the bait" and going to war should actually help improve that by showing up the grifters and giving us a chance to switch to making stuff that works.
> The Iran war doesn't count as the alignment of interest requires an actual threat of being defeated.
That's starting to sound a bit no-true-scotsman. If we need an existential threat to the US, that's not going to happen - realistically China conquering Taiwan or even building an empire around the Pacific would still not be felt as such a threat.
To align the interests there has to be a substantial negative consequence that would be felt by the grifters if the endeavor fails.
The US is already close to losing world hegemony status and it kinda needs it in order to print money / export inflation. A multipolar world is one where the US is greatly diminished and this will happen with or without losing a war.
> To align the interests there has to be a substantial negative consequence that would be felt by the grifters if the endeavor fails.
Like what though? The failure in Iran has had pretty substantial consequences that are being felt. If that's not good enough, what is? You were talking like you thought there was a realistic path to a better military, but consequences for the US aren't going to come much bigger than this.
You're conflating the grifters with the US in general, the grivers are able to continue grifting even at the expense of the US. This is requiring too much hand-holding from me so I'm done with this conversation.
All China needs to do is do what they’re doing play the long game the United States is currently shooting itself in the head, if they’re smart, they should just sit back and watch the show by the year 2100 well you know. And coincidentally that also applies to Russia sit back and watch them do it to themselves.
In addition, some of the other countries like Canada, Mexico, Australia, and New Zealand had better get busy from within because they’ll be on their own. In the same applies probably to Europe.
>They shored up their gpu design and manufacturing expertise.
I'm pretty sure they've been exposed for smuggling GPUs into the mainland because they can't ramp up fast enough, only reason we got Deepseek v4 before GTA VI
The US is undermining its own (and everyone else's) economy just fine, no imaginary assistance from China necessary.
The role of the US was always to purchase cheap Chinese hardware, slap some modestly better software on top of it and the rest of the world happily would pay for that as a whole package. But with the US increasingly becoming isolationist, the rest of the world is starting to wonder why do we need the US as a middleman at all, so the US had to invent a whole new reason for the rest of the world to rely on it: AI.
Of course, the problem with this idea is that while everyone was perfectly happy with the previous arrangement, nobody else in the world gives a shit about AI. It's scary, it takes the coolest things we used to enjoy doing and turns into mush, it destroys our local culture by making us all rely on English, everything bad (like layoffs) gets blamed on AI and so on and so on. And when you combine that with the rest of the stupid foreign policy decisions, many would find joy in witnessing the US economy crumble to the ground. Pointing the blame to China instead of to your own reflection in the mirror is just an easier pill to swallow.
This is spot on. The US under MAGA are actively dismantling their once leading position in IT as well as defence. I guess it is hard to see as a US citizen but from outside this is clear as glass.
> The role of the US was always to purchase cheap Chinese hardware, slap some modestly better software on top of it and the rest of the world happily would pay for that as a whole package
Curious where Intel, AMD, Nvidia, etc are in your "cheap Chinese hardware"?
And by "role", do you mean doing the majority of the R&D behind the modern hardware we all use?
That's just Chinese hardware with extra steps. If you don't believe me, feel free to look up the list of CEOs that are in China right now as a part of the US delegation.
As for the R&D part, Huawei is still pretty much indistinguishable from any other phone. I could buy one right now if I wanted to. It has shittier software though.
Trivialise it all you want, but the world is vastly different from what it was at the beginning of 2025 and I don't think you or anyone else can deny that in any way.
What happens next remains to be written, but so far this new order seems to be leaning heavily towards China and to a lesser extent the EU. Not because of anything those two have or have not done, but because of what has up-until-that-point been widely considered to be world's number one superpower losing its damn mind. I don't even have to come up with a list of examples to prove my point, we both have pretty much the same list in our minds already.
Instead, I'll just quote the President of the United States from a little over 24h ago:
> I don’t think about Americans’ financial situation. I don’t think about anybody.
AI is just another in a series of slaps to everyone's faces by the US. If it has some legitimate long-term use (which according to me is still an open question, although to many others it is not), thank god the US does not have as significant of a moat as necessary to fully control it, as the crux of it is easily replicable (albeit expensive).
If Anthropic does a down round, the US economy will crash. Not hyperbole.
The US economy right now is based entirely on the AI bubble. This is an indisputable fact if you examine GDP stats and equities.
That bubble is driven by (rational) over-investment in AI capacity. For that investment to continue, there must be demand for it.
The demand for that infrastructure essentially lies in the hands of a few businesses: principally OpenAI, Anthropic, Google.
The reason I highlight Anthropic is that without their advances in the last six months, the game would already have been up. Only via Opus 4.5 and 4.6 did the possibility of ROI look plausible. We are very much dependent on a handful of companies’ progress to keep this bubble going.
I’m not saying AI is bs, just that this is a bubble like others (for example, Victorian railways) and a down round would signal the end of the bubble.
So for an enemy of America, whether that be China or Russia or any other country, it is logical to target the AI bubble to cause an economic crash and thus restrict America’s ability to compete in terms of spending etc.
Flagged for AI content: I hope this submission dies and the user is penalized (look at their submission and comment history!), because IMO the article does not belong on the front page. Quick polemic:
>The US is winning the AI race where it matters most: commercialization
If you ask me, one could name different criteria for winning, and commercialization would not be the first thing to come to my mind:
> It also owns platforms that generate and organize the data of the AI age. YouTube is a video corpus. Google Drive and Microsoft 365 sit inside daily office work. GitHub sits inside software development.
Yeah, okay. China does not have any platforms nor data.
There used to be such a thing as profit. A return on investment. If your exit strategy is to get sold to Google, focusing on revenue is a perfectly fine strategy. If you _are_ the Google, however, the money poured in should eventually be made back. We seem to have forgotten that. The current level of commercialization just means the US is burning investments faster, than anyone else. Eventually this might change and the bet might pay off. But every minute this goes on, the expected payoff must be larger to pay for the loss made this minute as well as interest for the previous minutes. I’m not entirely sure this is what “winning” looks like. Tic-toc.
No, the US is _leading_ the AI race, but the race isn't over.
What's the point of leading the race for 90% of it, if they're gonna slip on their own sweat and fall down by the end? In non metaphorical terms, what's the point of spending billions of dollars rushing to get the best AI tech at all costs, when the competition can distil your progress and catch up in 6-12 months while only spending 1% of what you spent.
Even in the aspect the article cares about, commercialization, the US is starting to lose marketshare, I've seen people move from cc/codex plans to use glm/opencode plans due to the recent squeeze the US companies put on plan usage, the US companies are screwed if that sticks, not everyone needs the bleeding edge models, they just want to pay $20/month and have the models be decently capable.
What if it is not winner take all? What if there is no race. What if what USA has been doing is just burning money with possibly unsustainable debt load and way over build valuations...
AI being commodity server capacity might be a thing. And the customers might even manage without hyperscalers... In that sort of end scenario whole current market might look rather foolish.
>What if there is no race. What if what USA has been doing is just burning money with possibly unsustainable debt load and way over build valuations...
You mean, what if the hype-based billionaire-class is wrong? Isn't suggesting that a sin in America these days?
When a cyclist is leading a pack and pushing themselves against the air resistance for half the race, do you expect that cyclist to win, or one of the ones behind that's been taking it easy in the slipstream?
I have never ever heard a commentator say something like "Arsenal are currently winning with 2-0 against X". It is always leading: "Arsenal are currently leading with 2 goals against X".
I find it very strange that the GP felt the need to correct a difference between leading and winning. If you're at the front of the pack in a race, you are both leading the pack and winning the race.
If your team has more points than the other team, you are both leading the contest and winning the contest.
It is a distinction without a difference.
The elephant in the room, and where the analogy breaks down, is that a race has an end, the finish line. A sports match has a victory condition of some type. Nobody has a damn clue as to the victory condition of this hyperscalar craze. Anyone who says otherwise is incorrect.
GP here, leading and winning are different things in the race context/metaphor.
In foot/cycling races there's often a pack leader, that leader is often not the winner of the race, all they're doing is taking the brunt of the air resistance while everyone else slipstreams behind. For a casual observer it seems that the pack leader will win, but everyone knows that it's gonna be someone that paced themselves that's going to overtake the first spot at the tail end of the race.
You’re moving the goalposts and tying to one specific sport. I didn’t say “winner” nor did anyone else. “Winning” is the operative word, and tying the whole analogy to cycling is as close to a strawman as one can get while having the ability to claim otherwise.
Leading the race makes sense if it's a winner takes all market. AI cannot be a winner takes all market, because of national security reasons.
I would also argue that as AI gets better it will also be more fungible. It will be valuable like electricity. Lots of companies make good money producing electricity, but not the kind of money current investors are hoping for.
Mark Cuban in a recent interview answered your question: companies are afraid there is going to be just one in the end—sort of the way there is one ad-company now on the internet. They want to be that one.
Whether they're correct that there can be only one is of course a matter of debate. But that is at least the mind-set they are operating under according to Cuban.
Why would Mark Cuban know anything about the motivations of today’s big tech companies? He has not been involved in tech businesses since he sold a radio on the internet website 26 years ago.
He was never based in Silicon Valley, and the closest he got was selling a website to Yahoo in 1999. After that, he has mainly sold sports and his media personality for TV shows.
Moreover, why would leaders of trillion dollar big tech companies subject to myriad securities laws be discussing intimate business details with random people that have no domain expertise or influence?
I do not see the names of people in big tech business leadership positions, except maybe Andressen, if he counts. All the other ones look like media personalities or journalists or some two-bit SV founder.
The winner in the long term will be the one that will deliver the best performance and low-memory ratio for local models.
Anthropic, OpenAI and Mistral are just companies that are making money right now (still not profitable), but will lost their tractions and values in the long term.
However, I am more appealing to see how OpenCode Go subscriptions will go in the future: cheaper than big techs, more tokens, and they don't train on our data to (try to) improve...
Local models will never compete with large SOTA models, in the same way an iPhone doesn't compete with supercomputers doing nuclear simulations.
They paths will differentiate and split. Probably SOTA models will eventually be locked down and only accessible to state actors because of how expensive they will be to run (already started with Mythos).
its a big assumption that larger models bring any measurable benefit in the long term. there's a point where its not worth paying the expense of a bigger model and we dont know where that will be as both, models and hardware improves.
we do know however where evolution is at right now with our brains, but thats probably not comparable - yet the only thing I can see to make any kind of prediction at all
A Qwen3.6-35B-A3B or whatever it's full name is, when on a 3090, can at the very least, with very little fine tuning, compete with Haiku and blows away GPT4.1 (aka, the cheap models).
It might keep up with Sonnet 4.5 with some tinkering.
But long story short: it seems to have better performance and similar quality for a payoff of a year or so compared to cloud models. In the same way you can self host faster/easier/cheaper than cloud hosting, if you are okay with the negatives.
I'm returning my 3090 soon for a R9700 after some more basic benchmarking, since the higher RAM should improve my observations more.
> It might keep up with Sonnet 4.5 with some tinkering.
I would love to see that. I've been using Qwen3.6 35B and the dense 27B, and they are both too slow with not such great results for agentic coding tasks. It's ok, but not impressive. I had better luck with the BF16 and Q8 than the Q4 from unsloth (really love what unsloth is doing in this space).
Another problem I had with Qwen, which I did not ever encounter with Sonnet - even the BF16 gets stuck and needs a "continue task" prompt from time to time, the lower quants are even worse in that regard.
If you get some interesting results, I would love to read about it!
You don't mention runtime, hardware and harness which are critical. The 35B A3B model should be pretty fast, you do need a decent setup but nothing too fancy. I'm using Q8_XL from unslouth with llama.cpp and opencode and it's pretty awesome. I find that opencode drives the model best, it very rarely gets stuck even with a ton of tool calls. I agree it's comparable to Sonnet 4.5 for most tasks. You may also try the Gamma 4 models which are faster but not as good for coding.
You are missing the point. Parents says the market to win need economical models more than SOTA models. Whoever is running those nuclear simulations is not making as much as Apple.
Not training on data is a con for me not a pro. The reason Claude is so good is RL training from users' chat histories and use cases. The era of pure public data training is over, as everyone has access to this data yet only a few are frontier models.
...printing the most money into it. The circular IOUs amongst the AI and hyperscalars are a form of debt, i.e. money creation. Don't get me wrong, a whole lot of other dollars are going in too, but investing money that doesn't exist is a massive risk always.
Maybe I'm not doing capitalism right, but isn't it supposed to be "The one who profits the most wins"? If you win by just spending, I think you need to adjust the parameters of your capitalistic market.
I don't get it either, but it seems to me a bit similar to how the US, if you look at market value of car companies, has utterly crushed Europe and Japan (with China surging ahead of those and maybe threatening the US soon), which to me sounds crazy (I still think of German cars as the top of the bunch).
According to Google (AI summary, no idea if it's 100% right but from what I've seen elsewhere it seems right):
Top Car Companies by Market Value (May 2026):
- Tesla ($1.3T - $1.56T): Retains market leadership with a valuation often exceeding the next several largest competitors combined.
- Toyota ($259B - $317B): Largest traditional automaker by market cap and unit sales.
- BYD ($122B - $126B): Strong market position as a Chinese electric vehicle leader.
- Xiaomi ($119B - $135B): High valuation following its entry into the smart EV market.
- General Motors ($69B - $75B): Leading traditional U.S. manufacturer, competing with Hyundai and BMW for top 10 spots.
- Ferrari (\(\approx\$60B-\$68B\)): Maintains high value due to luxury branding.
- BMW / Mercedes-Benz / Volkswagen (\(\approx\$58B-\$64B\) each): German luxury and traditional automakers facing high competition.
- Ford (\(\approx\$47B-\$54B\)): Remains a major player with significant US market share.
So, essentially, Tesla alone is somehow worth more than all European companies combined??!
Except that by sales volumes, the top companies are exactly the ones you'd expect: Volkswagen ($350B) and Toyota ($315B) at the top, far ahead of anyone else... Tesla is around the 7th place with just $95B. Does the financial markets still expect them to far out-earn Volkswagen and Toyota any time soon, we've been waiting for like a decade already??
Gemini says that by country, the car companies revenues are:
Stock market is about expected future returns. Tesla probably won't ever be very good, but it has the chance to regulatory capture the entire US market in a way that Volkswagen doesn't. Tesla gets to market overpriced junkboxes to rich people in a way that Volkswagen doesn't. Tesla has a likelihood of acquiring lottery-ticket companies like xAI in a way that Volkswagen doesn't. This stuff doesn't happen when your company just focuses on making cars.
The stock market is over 60% passive investment, it's starting to get unmoored from the financial realities of the underlying companies. What that means for the future is [shrug emoji].
Look at profit rather than revenue - "it's easy to make a lot of revenue when you're selling a dollar for 80 cents" applies just as much to big legacy automakers as it does to startups.
>Does the financial markets still expect them to far out-earn Volkswagen and Toyota any time soon, we've been waiting for like a decade already??
These capital heavy industries operate on 30+ year timelines, a decade isn't sufficient time.
Revenues are not the end all, be all. Profit and profit margin, along with revenue trends provide a more complete picture. And the most significant factor is that the market does not expect Volkswagen or Toyota to do anything new, to do anything with the potential to earn more. They are what they are, and they will continue with their lower margin businesses until they fade away.
Investors are betting that Tesla, however, might have a few tricks up its sleeve, that will allow it to expand markets and profits.
There is a reason in capitalism we have anti-monopoly law or preventing dumping prices because those often leads to monopoly. So yes for sure you can kill your competition by just dumping money and loosing profits.
That also assumes the monopolized market is profitable enough to pay for the dumping, but right now we are still questioning if LLMs have such high value on the market. Yeah its great for programming, but is the majority of the population benefitted enough to all start paying for access lke many investors expect? Someone might have looked at the low cost and massive lifting capacity of hydrogen balloons in the past and seen a lot of potential profit but if investors had dumped money to monopolize the hydrogen balloon market they would have lost their ass.
True, many people don't know GLM 5.1 and Kimi 2.6, really on par with frontier models. There's also Minimax 2.7, DeepSeek 4, Qwen, Xiaomi 2.5 Pro, etc.
China is leading in open source frontier models, so I don't really see how the US wins this one. At some point, companies and people will start running their own models in the cloud and locally, Chinese models will be everywhere.
Nah, I model hop constantly as I work with serving GLM and Kimi models and they're not nearly as good as Opus 4.5+ and GPT 5.2+ and it's not particularly close. They're good by standards set a generation or two ago, but they're really not competitive with where the frontier models are at now.
They compete with "mini" or "nano" model classes quite well given the price of inference. You'd need to "model hop" anyway, using Opus for everything is quite wasteful.
They are on the frontier of local models, where the game is often to get the best bang for the buck. You can always scale model size and compute (Mythos, GPT Pro, Gemini DeepThink) to reach better outcomes, but that's not a very interesting strategy.
That's not what anyone means when they say frontier models, don't change the definition. It's almost as bad as open weight being subsumed by open source when it comes to local models.
Guess it really depends on what you use them for. I've been able to built whole apps with them, not slop. Kimi is quite good at design, for 3D, I noticed Gemini 3.1 is excellent for basic to medium use cases.
I've tried both Opus and GPT 5.4, they also hallucinate just like the rest at a much higher cost.
The more you use a model overtime, the better you become with it. It's really hard to measure, my main metric lately has been tokens per second/time to complete task.
At this point I've the feeling frontier models are optimizing for benchmarks and one shot prompts.
If you actually use them you'll see that they are far from frontier models. They are much more cost-effective for what they are, but frontier they are not.
My understanding is that it's not that the _models_ are banned, but rather the _platform_ is banned. It is acceptable to host, say, `deepseek-r1-distill-qwen-7b` and run it yourself, for example. It is not acceptable (to the authors of these bans) to download the DeepSeek app and run it on your work device.
I just left a job for a German B2B software company which sold primarily to large automotive, defense, and aerospace companies. Several of our customers specifically banned anything with the word "DeepSeek" -- hosted or self-hosted.
There's still a lot of naivety on what the difference is between models and platforms, and its easier for a lot of these big companies to just make a blanket statement like "nothing DeepSeek" than for their procurement teams to try to understand and negotiate with each vendor. They don't see the potential benefit over the potential risk of somebody misinterpreting or getting it wrong, so they outright ban it.
Most people that approve or buy software simply also just don't understand how models are being trained or if it's possible/how far a model could go to "introduce backdoors." A backdoor could be, from a business perspective, a model which has been trained to give answers that could hurt western business in a "strict text mode" or produces payloads in a programmatic mode that are intentionally trained to introduce software vulnerabilities.
Anyone can make arguments against these for a variety of reasons (looking at the transparency of both sides and comparing, etc) but for many reasons today and for better or worse, many Chinese models are being banned on big software contracts, which gets back to the title of the article
Thing is these models can also be a propaganda machine whether you run it locally or not. This is true no matter the origins. Chinese LLMs will never shit-talk CCP, and it will always give a rosy depiction of the Chinese government. It's perfectly understandable if companies don't want things like that. US/EU models have these problems too, but at least there are some ways to fight that: with a lawsuit or a megaphone on social networks. With Chinese models there is nothing you can do.
I would say that both US and China are using the data we trust upon them for industrial espionage. So don't use their models if you are working defence or other sensitive areas
Do we have any solid evidence these models can outperform Western models in terms of quality? Or is it more: because they are forbidden, they can't get enough training data, visibility etc. to compete?
Spoiler alert - they are all towards the bottom of the leaderboard. People come up with a wide variety of excuses for why they are not used despite being offered for significantly lower cost, but the answer is simply because they don't perform well enough for now.
LM Arena uses human side by side voting, which limits its applicability to complex tasks.
The ARCPrize leaderboard does have Deepseek V3.2, which only scored 4% on ARC-AGI 2 (while the top models score over 80%). It also Kimi and Qwen, but they also didn't perform well.
You’re saying if we were allowed to use e.g. qwen more broadly the US wouldn’t be in the same strategic position? We have the best models…we own all the companies that make the best infra and the hyper scalers…I don’t think “oh we can use Qwen now?” Would exactly devastate the US
You agree they are winning though, right? China is known for not playing fair, stealing industrial secrets, etc... that reputation matters and it's a good reason why the US is winning. Is the US perfect? No. Does the US play fair? No. Spare me the whataboutism in the comments. The bottom line is most people think the US is a safer bet and that's why we're winning. I personally wouldn't trust either government, but if I had to choose, I feel like I at least have a chance at secrecy and due process with the US. Obviously that is being eroded day by day, but you literally have no due process in China.
Is the US actually winning the commercialization war? The US is definitely delivering more commercial products but if all of those products are deeply unprofitable and need to buy users with unrealistic discounts (or direct cash payments[1]) to keep their DAU's looking good then is that winning?
There's a significant amount of innovation happening, but if the market decides this AI thing is not worth funding then I think that'll dry up overnight.
One interesting thing that Anthropic did was putting their stack on the various cloud providers, I wonder if they'll put it on GCP and Azure next since they've put it into AWS first at a level we have not seen a major AI provider do to date. Your company can have their own Claude stack just like an ELK stack on your cloud, if they can do this for both Azure and GCP then OpenAI has to really catch up.
In my eyes I would rather use the AI I can run on my own paid infrastructure, so if there's an outage its isolated, or I could potentially have a different region / DC to fallback on.
I'm still surprised that neither Microsoft nor Amazon have made their own models available on their cloud offerings. I guess Microsoft probably does have Phi on there, but it's not front and center, especially with something like Copilot for Devs (seriously Microsoft rebrand that damn thing to be clear what you mean by Copilot!) where they could use the cheaper compute by using something like Phi.
> In my eyes I would rather use the AI I can run on my own paid infrastructure,
Claude has been available like that for quite a while.
One of the reasons for the OpenAI divorce from MS was so they could become available on AWS where they see significant demand, and being available only on Azure was holding them back.
The recent deal with SpaceX AI to use their severely underutilized GPU compute is pretty telling to me. Being able to roll out compute is a hardware problem, rolling out good models needs more than compute, it needs good AI engineers. SpaceX, Amazon et al can do hardware very well. AI engineering, maybe not so much.
Is Vertex hosting it themselves on their own hardware? Because GCP, Azure and AWS all host it on their own hardware, but AWS lets you do everything that Claude lets you do if you use their APIs directly, and then some. This is what I want to see on Azure and GCP, and heck, maybe even DigitalOcean, if they ever stop expanding into so many spaces and focus on improving their current infrastructure, before I fully migrate off of them.
> Is Vertex hosting it themselves on their own hardware?
Yes, you can even choose regions, for EU they serve it from Belgium. With all the encrypted at rest stuff and other guarantees that vertex provides.
> Important: Accessing Claude models through Vertex AI meets the FedRAMP High requirements, and operates within the Google Cloud FedRAMP High authorization boundary.
This makes no sense when you zoom-out. None of these companies, be it Anthropic, OpenAI, xAI, Google, Meta, Microsoft, are profitable in the AI department, they're all bleeding money and using funds their parent company and/or investors, primarily investors gave them. The Chinese models are keeping up with them, while offering the models for free and able to run on consumer grade hardware, and more importantly they train them for cheap. AI models are an extremely volatile product that can be outdated in the matter of a few weeks. Meaning you have to keep dumping resources into developing better models which has no end-goal besides infinite scaling. Lets look at how users behave in the real world:"I don't use Gemini because it's worse than Claude at XYZ." That's it. Now Gemini has a worse model and people are going to Anthropic... what happens when Anthropics model is arguably worse than everyone else's? What does it matter if they can commercialize if their product is objectively worse?
I understand that America dominates in distribution, integration, enterprise contracts, ecosystems, infra... The article isn't wrong, it's just that that dominance is fragile and requires constant upgrading.
But what is the point of that if you have to infinitely scale because the opposition is right behind you at all times ready to usurp you... You CANNOT scale infinitely, the VC money will run out at some point and then everyone will have to downscale everything to meet the real costs associated with SOTA models, they'll have to be able to use subscriptions, and other monetization to cover those insane costs, we just saw SORA shut down because it was bleeding money far too fast while the Chinese released video models that far surpassed it back to back to back...
EDIT: Hell, one of the most critical aspects is integration of the models into other products, and even on this end open-source is keeping up (and will eventually outpace when the VC money dries out) with these big companies.
> None of these companies, be it Anthropic, OpenAI, xAI, Google, Meta, Microsoft, are profitable in the AI department,
Citation needed.
All reporting is that they are profitable on the inference side and all the VC money is going to building more data centers to run more inference. (Note that the coding subscription models are probably only break even on average - the money is in the API)
> The Chinese models are keeping up with them, while offering the models for free and able to run on consumer grade hardware, and more importantly they train them for cheap.
No one is running DeepSeek v4 (a 1.6T token model) on consumer hardware.
They aren't much cheaper to train the US models. Training is subsidized by the big Chinese tech companies. They are slightly cheaper because they are smaller (and weaker) models than the 5T and 10T models the US frontier labs are training, and the US labs are paying for a more diverse set of RL data (which shows up in diverse benchmark performance).
> we just saw SORA shut down because it was bleeding money far too fast while the Chinese released video models that far surpassed it back to back to back...
Ironically this proves the point.
OpenAI didn't shutdown Sora, just the subscription version and weird social network thing. You can still access it via API.
The Chinese models are API models and probably just as profitable for them as the LLMs are for the US frontier labs.
[1] has prices for video models. There is a big range, but Google's Veo model and OpenAI's Sora are around the same price as the Chinese models.
What does profitable on inference mean? As far as I can tell, none of these companies have rigidly defined it, let alone it being a GAAP number. And yeah, if you subtract out all your R&D, payroll, sales, marketing, and other overhead, and get someone else to take on the debt or dig into their free cash flow to build the hugely expensive infrastructure on which you depend, it'd be pretty hard to not be "profitable". It's almost humorous how dumb of a metric "profitable on inference" is.
Ask yourself if AI was so profitable, why don't any of the big hyperscalers break out AI revenue in their earnings. OpenAI and Anthropic both project huge losses for the next couple years, it's not hard to find.
The real problem is, as the GP comment pointed out, that they can never stop training. As long as they're committed to building these behemoth models, the second they stop training, someone else will catch up and everybody will switch over because it's trivial to do so.
> OpenAI and Anthropic both project huge losses for the next couple years, it's not hard to find.
No. Anthropic at least expects to be profitable this year:
> Anthropic expects its gross profit margin, which measures how much revenue it makes compared to the cost of producing that revenue—largely from running servers—to swing from negative 94% last year to as much as 50% this year and 77% in 2028.
> And yeah, if you subtract out all your R&D, payroll, sales, marketing, and other overhead, and get someone else to take on the debt or dig into their free cash flow to build the hugely expensive infrastructure on which you depend, it'd be pretty hard to not be "profitable".
I think excluding capital expense on infrastructure isn't unreasonable and is done in most industries.
It's worth noting that AI infrastructure has turned out to be an unbelievably good investment. Inference on a 4 year old H100 chip costs more now than it did brand new! That makes the hyperscaler's depreciation schedules look very (and unexpectedly!) conservative (!!)
Even if that is true, it doesn't change the reality that they can compete. Also, if we start going that route, American models wouldn't have any quality data to train on if they respected copyright themselves. Their whole product was built on the work of others, on our work, our art.... without compensation, without acknowledgement.
Literally not a single one of these AI companies, regardless of where they are in the world has any right to complain about someone copying their work.
What sorts of apps are those? I tried testing various models with a test app as a benchmark, a local first app with CRDTs, and many, even frontier models, struggle heavily.
and you're not alone (I run Qwen 3.6 35B-A3B at home too).
But just for the sake of discussion, let me ask: Who is the service provider you're using to run Deepseek V4? Do you have any way of knowing whether that compute is happening in the US or abroad?
Article content: “The US are capitalizing on AI the best”
A lot of assumptions there that no one can actually verify as true right now. If commercialization into rent-seeking SaaS landscapes is the endgame, then yeah, the US is winning the AI race. If individualization, local LLMs, and consumer hardware are the endgame, China is winning the AI race. If it’s something entirely different - if LLMs are the wall and research is what grants the next breakthrough, or if compute and memory requirements take a dive, or whatever; then we have no idea who’s winning the race because that stuff is mostly happening behind closed doors.
That seems like a lot of rationalization to me. China is pursuing these because they cannot compete on the frontier. Yes, there is a possibility that all that compute is not needed, but it is a rather remote possibility, and there is no doubt that, given the choice, China would be pursuing frontier model building with closed, propietary-only offerings.
All that compute is not needed. We have an existence proof from biology in the form of natural intelligence that much greater efficiency is possible. However, achieving dramatic improvements in compute efficiency will depend on unpredictable scientific breakthroughs. Personally I suspect that an entirely new hardware architecture will be needed, although I don't have any hard evidence to back that up.
I dunno, DeepSeek v4 Pro is rather on par as far as I can tell, maybe not with 5.5 Pro in all areas quite yet, but close.
I think China is thinking more about the application layer on top of models as going to matter more than the models themselves, so they don't need to gatekeep the models as much.
China is competing in value AI because they cannot work at the frontier, but how is this bad at all? It’s like how the USA has the best drones but they are a few million dollars apiece while China has DJI.
If China could work at the frontier, I don’t know, I kind of think they would still be dumping a lot of resources into exploring the value side since they have that culture already in place.
This is true, but I don't think they would all be rushing to frontier if that option was available. Chinese are used to working with constraints to their benefit, they would see the price of working at frontier and make hard choices that maybe we can ignore in the states.
Forgive me if this is a naive assumption, but wouldn’t large language models be fundamentally different for a language that is largely symbols?
Again, my understanding of Mandarin is limited if it exists at all.
This is why misspellings and homophones are tells of human righting. LLMs strongly prefer word-level tokens, and word substitutions follow semantic similarity and not the more human auditory similarity.
> LLMs strongly prefer word-level tokens, and word substitutions follow semantic similarity and not the more human auditory similarity.
Is this an elaborate joke or your full-word misspelling of writing is both agreeing with your statement (word substitutions) and contradicting it (not semantic but only pronunciation similarity)
"飞机" and "airplane" aren't fundamentally different in terms of how they're represented to a computer. Especially for an LLM, where tokenization likely turns each of those into a single token.
> ? Claude, ChatGPT, etc are heinously expensive for tiny benefits lmao
Unfortunately local inference is inefficient, 100s of times more inefficient than cloud. When you answer one request at a time you still have to fetch all active weights into compute units, once every token. When you run a batch of 300, you load it once and compute 300 at a time.
Compared to cloud, local inference is less flexible. You can't scale up 5x or 20x, can't have spikes, and pay for it no matter if you use it or not. But usage factor is very low, like 5%. And to run a decent model your system costs $2000 or more.
AI boosters cling to this notion because it's the only way the massive data center buildouts make any sense at all. I guess you could say the US is winning the frontier AI race. Okay. I'm never going to grant a cloud service access to all the contents of my hard drive, that's just never going to happen, so if you expect me and a lot of people like me who feel similarly to get on this train, you better have a local, lightweight model too or we're not even having a discussion, the answer is just no.
The thing is, frontier model providers don’t take your feelings into account even a little bit. It’s totally irrelevant to the discussion about the service they can provide, because that service is predicated on access to high power GPU slices that local models can’t touch. Those providers won’t be in an existential crisis because some people choose the privacy route, it’s a cost of doing business.
Right but that service being sold is predicated on products being sold to users, yes? Or are we still pretending that the hyperscalers can just pass the same $20 billion between themselves and that's going to be a growth industry forever?
I suppose its possible that all the value to pay back the datacenter construction can be squeezed out of enterprise contracts where your employer can assent on the privacy questions, probably with some kind of complicated contract and insurance regime regulating things.
Even if so, if China is coming behind 6 months later selling laptops with hyper-efficient local models that are 80% as good as "frontier" ones, I imagine they'll get the consumer business AND a fair share of the enterprise business as IT managers look at their options during the next refresh cycle.
Given economies of scale, I think it's ultimately inevitable that the enterprise more-or-less follows the consumer on this, and the consumer is going to prefer local models. There's no ongoing cost after the initial purchase, and your data at least nominally stays within your control.
I'm inclined to agree. The business world itself, and frankly you could make this argument about the entire AI industry as it were, runs on "fine" and 80% capable can probably get you there. And it's arguably even better for the hyperscalers since then, ongoing costs of AI users are basically nil. You can still have your massive datacenters and just keep them for tasks complex enough that they're actually worth spooling up.
Like I don't need an H100 or a dozen to summarize a PDF. And that's most of what I use AI for.
If we are betting on which is an easier sale, $20-100 a month w/tech support included vs $5k-10k and a requirement for moderate technical ability, I would invest in the former not the latter being the proposition that drives the conversation about AI use.
> ? Claude, ChatGPT, etc are heinously expensive for tiny benefits lmao. Local + efficient is clearly the future
Corporate America is where the money is, and corporate America will dictate what products are successful by virtue of spend. Individuals aren't going to be paying $100s or $1000s/month en masse for these models but businesses will be. Being local and efficient isn't that important at this stage but even so as American companies continue to scale and invest they'll be able to make those models more local and efficient if the market wants it. Sort of like how you had a big, giant desktop computer and now you've got a super computer in your phone which is in your pocket. Going straight to "local and efficient" means going straight to being behind because at some point, perhaps now even, the local and efficient model won't be able to keep up.
For some reason people think that they somehow know something that Google or Nvidia or whoever, with hundreds of billions of dollars of real money at stake don't already know and it's both amusing and bizarre to see this play out again and again in off-hand comments like "lol tiny benefits".
You buy an iPhone even though the cheap-o Wal-Mart Android phone for $100 "does the same thing". Except that in this case the Android phone just puts you out of business while those spending big money for "tiny benefits" beat you in the market.
I don't think that's a particularly bold claim after thirty straight years of moving supply chains overseas. Capital is, inherently, the means of production. The world where we could compete is gone.
Capital is not the means of production. Capital is capital. If I have a few million dollars in my bank account I don’t all of a sudden have a factory. Remember from economics class you need capital, labor, and the means of production.
Capital inflows are different from manufacturing outflows. The US has historically imported capital which is part of why we have such a large trade imbalance. I’d encourage you to do some more digging here.
> The world where we could compete is gone.
Sigh no that’s just not true at all. We compete hard and fast all day everyday, economy is growing and will continue to do so, and no amount of leftist doomer, Chinese, Iranian, or Russian propaganda changes those facts.
> I have a few million dollars in my bank account I don’t all of a sudden have a factory.
No but money only has value because of a product of the human labor and production capacity it refers to. Money is not capital, it is a reference to/legal coercion of capital
> We compete hard and fast all day everyday
Sir have you ever been to the us? Lmao. We are only competitive in the industry of white collar work (financial/artisanal services), an industry that capital is actively gutting
> No but money only has value because of a product of the human labor and production capacity it refers to. Money is not capital, it is a reference to/legal coercion of capital
These are just strings of words without meaning or importance.
> Sir have you ever been to the us? Lmao. We are only competitive in the industry of white collar work (financial/artisanal services), an industry that capital is actively gutting
Yes, I live here. Why are you posting obviously untrue and asinine statements like this? Go look at the Fortune 500. There ya go. What other evidence do you need? And not only are you writing dumb things here, your original post was wrong too! Please get off of social media or whatever doomscrolling news you are partaking in because it is bad for your health and perception of reality. The United States by any measure, as a matter of indisputable fact, a highly competitive and dynamic economy across pretty much all sectors. This is not up for debate.
> You buy an iPhone even though the cheap-o Wal-Mart Android phone for $100 "does the same thing".
People buy iPhones because of status signalling and network effects, neither of which appears to apply to AI model choice. LLMs are already rapidly on the way to being interchangeable commodities.
No they don't, it's not 2008. Anybody off the street can get an iPhone or a free iPhone with a mobile plan. They're commodity products. Even homeless people have them.
To the extent LLMs are commodity products you're right (so far), but that is limited to the main model providers, such as ChatGPT, Claude, Gemini, &c. with interoperability on cloud platform providers and other technology providers like an Apple offering you a choice of LLM with Siri or something.
If you want to suggest that some other model is in the same bucket as those primary 3, it goes back to the crappy, cheap phone analogy which is accurate. Yea you can make calls with it, but you make calls better with an iPhone.
They run various schemes like this all the time, you can also trade in your existing phone a lot of times for pretty favorable terms. I've traded in phones that were a few years old and gotten $1000+ for them, especially when switching providers.
Verizon's "free" iPhone deal is you pay for the phone up front and then receive a bill credit. Here's the fine print from one of those deals:
$729.99 purchase on device payment or at retail price required. New line req'd. Unlimited Welcome, Unlimited Plus or Unlimited Ultimate plans required. Less $730 promo credit applied to account over 36 mos; promo credit ends if eligibility requirements are no longer met; 0% APR.Taxes & fees may apply. Credits will appear on your Verizon Wireless bill.
I don’t know or care much about the specific details but the article was written in 2025. Carriers run deals and give away iPhones or close enough to free or cheap that quibbling about the details is irrelevant.
If you think the iPhone is a status symbol you’re just wrong.
I'm not the one arguing iPhones are only status symbols. If anything, if I only had the money to spend on a single computing device there's a good chance I'd go for an iPhone due to excellent durability, typically long support timelines, lots of extremely cheap accessories available, high chance of low cost serviceability compared to other devices. There's also a pretty good used marketplace for such devices so picking up one used on the cheap and still getting a few years of use out of it is likely. I'd likely try and stretch to get that device instead of settling for a cheap $100 phone that will be a total piece of junk and end up being my only actual computing device.
I'm just pointing out the statement:
> What mobile plan giving you an iphone doesn't come with explicit debt?
isn't invalidated by some Yahoo article pushing a marketing promo that when you actually do the math and read the fine print its not really a "free" phone, its always some form of debt or bill credit or something along those lines that makes the phone "free". You're still paying for the phone in the end if you read the fine print. In the end one commits to spending several hundred dollars over 36 months or whatever or you pay up front and they give you bill credits if you keep the plan.
I am arguing they’re not status symbols and using how cheaply available they are as evidence that they’re not. Anyone can get one, some companies run free promotions, some do delayed interest programs, some amortize the price over a 2-year time period. Who cares? The details here weren’t important. Apparently Verizon ran some promo in the past and may again in the future giving away iPhones. Why be so argumentative over something so stupid? Not only are you actually wrong here, you’re arguing over the irrelevant details.
> Why be so argumentative over something so stupid?
I don't want people to believe untrue marketing statements and make poor financial decisions without actually bothering to read the fine print.
> some companies run free promotions
This just isn't true. They're not really "free". They come with lots of financial commitments.
> Apparently Verizon ran some promo in the past and may again in the future giving away iPhones
They still say they do on their website. If you're getting one "free" iPhone it comes with a commitment to spend at least $65/mo for 36 months. A commitment to spend $2,340 is a lot different from $0.
These are far from "free" phones. Can I go into a Verizon store, not give them a dime or sign any contracts and walk out with a phone free and clear to do whatever I want? No? Sounds like it's not really free then!
My point is if you're poor/homeless you're probably not looking to sign a 3-year commitment to spend a few grand to get a "free" phone. A lot of those people won't even pass the credit check to qualify to even sign up for one of these post-paid plans required to get the "free" phone. If you're really broke you would probably be looking at signing up for a lifeline plan and get yourself a cheap used iPhone instead of signing up for a $2,340 contract.
I already made and proved my point. The iPhone is not a status symbol, and major carriers can and do give them away in various schemes and did so in the past and will continue to do so in the future.
You’re anchoring yourself to one payment scheme and ignoring others and it’s besides the point which is that iPhones aren’t status symbols even if these schemes didn’t exist and iPhones weren’t extremely cheap or freely available.
I don’t have anything left to say here besides that I proved my point unequivocally.
> major carriers can and do give them away in various schemes and did so in the past and will continue to do so in the future
They only do if you're financially illiterate.
> You’re anchoring yourself to one payment scheme and ignoring others
I'm being honest and taking about the real deal instead of blindly repeating marketing bullshit and lies.
> freely available
A commitment to spend thousands of dollars isn't the same as freely available.
The bank gave me this free house all I have to do is pay this mortgage for thirty years. But hey the house was free!
Once again, was the deal that you could walk into the store, grab a new iPhone, and walk out without signing a contract or other form of commitment? If not, it's not really free. It's bad financial advice for people struggling financially to get one of these "free" phones, they're often more expensive than buying outright and getting a much cheaper (or potentially even subsidized!) plan. Especially if you're just needing one or two lines. Many of these postpaid plans only really make financial sense once you're at like 4+ lines on it.
I'm reminded of seeing all those cell phones in the RadioShack mailer ads back in the day. Only 99¢! Dad, can't I get one? It's only a dollar!.
If you spent hundreds of dollars on box seats to a sporting event and they had a complimentary buffet, is that food really free or did it cost you hundreds of dollars? Would you tell someone struggling with money they could get free food, they just need to go spend hundreds on sports tickets first?
Maybe one shouldn't be so willingly close-minded to the truth.
Well China is consistently 6 months behind the frontier labs(possibly because they can they harvest data from released frontier models). If the scaling continues, US will win, but if not then China will win as the models will converge.
The non-release of Mythos tell you the future of that, so long as they can keep the weights from being exfiltrated. Once models become true national security threats, they won't be released in their full form. The hitch-a-ride approach becomes less capable of keeping up.
By only providing degraded models to use commercially outside national defense applications would be my guess. As soon as models are a threat in terms of enabling biowarfare etc, then they just are not going to be generally released.
Honestly, I think its quite possible that models will be retrained with gaps in their knowledge. e.g. a coding model for commercial use probably doesn't need to have deep knowledge of biology, and training on biological sciences probably doesn't help those evals much.
LLMs absolutely have the potential to increase the risk of small nation biowarfare and bioterrorism more generally. That you don't believe so is dangerously naive.
"AI in the datacenter" and "AI on local consumer hardware" will eventually be two separate niches with entirely different capabilities, at least if scaling laws continue unchanged and there's no near-term inherent limit to AI smarts. The real point of the datacenter is to be able to do datacenter-scale things. But you don't need that kind of vast compute to run even the largest open models today: on prem hardware can do it easily especially if you're OK with a somewhat delayed response.
It’s certainly too early to call (if you must view this as some sort of adversarial competition). The US is behind on local models, the future for anyone who cares about privacy. There may be step change innovation yet to come that completely shifts the landscape. There’s basically no switching costs to users to change models. They have no lock-in.
It is very much adversarial, and to view it anything but adversarial is to not see the geopolitical reality and the potential national security implications of AI for what they actually are. Moreover, to claim that China is in the lead with local models presupposes that openai and anthropic could not release local models that are better, which is a big assumption. They do not release such models because they have frontier-grade propietary models that have high value.
As someone who happens to have been born in the US and currently lives here, I welcome China winning. I trust them infinitely more than I trust my own government and industry.
OpenAI and Anthropic are beholden to the capitalist system they exist under and hence cannot compete on local models. Like you say, they must try to maximize shareholder value. China is unencumbered by that constraint.
They're far less imperialistic and I view them as better global citizens than the US. I think they've cultivated a much richer culture than the US as well.
Happy to if you have any pointers. My original point is that the US kills millions of people outside of its borders, something China most definitely does not. The number is over 12 million post-WWII: https://www.worldfuturefund.org/Reports/Imperialism/usmurder...
Replace the U.S. with another world power. Would it be better or worse?
Its always easy to criticize the power that is in place and treat them as if they are to blame for all the ills of the world. It is easy to criticize when you are the nation that is asked to help provide international security for commerce on the high seas. But exchange another nation in the U.S.'s place, and I guarantee you most would make similar decisions, and many (e.g. Soviet Union) would be, hands down, worse and more brutal.
Well if you discount the people China has killed and is killing within its borders, the number is definitely smaller. As well as another land empire called Russia.
not living there, for one. I don't care if they know where I live since realistically they can't do much of anything to me. If I were in china, I probably wouldn't trust them as much as I trust the US. If I were in Switzerland, I wouldn't trust the swiss government and might get my services from america or china.
You are free to hate capitalism (even if you benefit from it enormously). You are free to say that you hate capitalism and the U.S. as openly and as often as you like, without facing imprisonment or worse.
But if you were in China, could you say you hate the Chinese Communist Party and China openly and as often as you like without imprisonment or worse?
We know the answer to that. So go ahead and trust China more than the U.S., but I think that is pure foolishness.
Actually, Chinas free speech, while abysmal, is better than you're making it out to be. They only really care if they see you as a threat, which realistically isn't too far off from the US both currently and historically
I understand that there is some freedom of speech in China, but I don't think the implication is that US and China are functionally equivalent. In the U.S. you can criticize the president all day every day and have millions of followers. Could you do that in China? The more people listen to you, the more dissident opinions become a threat.
I'm not suggesting they are the same, rather, that if the US saw you as a real threat you'd be shut up pretty quickly. The threshold for what is considered a big enough threat is different, but they both have the ability to shut you up and have in the past. Any government would act the same, at least a competent one. The governments 0th priority is ensuring its continued existence, since all of the other functions rely on its existence
I live in a Zionist country (the US) and will absolutely be canceled and blacklisted from my industry (tech) were I to publicly speak out against Zionism. They are attempting to put laws in place to make it illegal to be anti-Zionist. These laws already exist in countries like the UK and Germany. Some, such as anti-BDS laws, already exist in the US as well.
Many technological advances weren't driven by capitalism, early computers and the internet were literally developed by the government.
I've been using the deepseek api (not for coding though) and have been getting great results and it's so cheap it may as well be free. Another reason I'm using it is because I like the license and I also have some hope of running it in my own hardware in the future.
But the thing is... I could be using any of the llms for my use - I'm using a middleware that lets me change providers only with a configuration change.
So it's going to be tough for USA ai companies to charge 5x to 20x (depending on what you're doing).
It begs the question because both its premise and assertion are already wrong. Has AI improved the industrial capacity of the US in order to improve the lives of its citizens? No it hasn't. Has AI increased the wealth of its citizens by being able to do laundry or any household task in a generalized way? No it hasn't. The only thing it's really done is make very narrow slices of white-collar work more fungible. In what way has AI been able to address existing shortcomings of the US?
You better have some sources for declaring that industrial capacity hasn't increased. The Fed reports around 8% penetration of AI in manufacturing already, but in my opinion it's too early for grand declarations like that without data.
> ... it's too early for grand declarations like that without data.
You mean grand declarations like 'industrial capacity has increased'? Just because AI is present in the factory doesn't mean it's actually increased capacity.
> Have you happened to purchase anything in the past 12 months, and looked at the Fed's inflation numbers?
The Fed doesn't issue inflation numbers. The usually cited headline inflation numbers (CPI) are from the Department of Labor’s Bureau of Labor Statistics, the ones used by the Fed as an input to monetary policy decisions (PCE) are issued by the Department of Commerce’s Bureau of Economic Analysis.
and of everyone else, right? what service or product is only available to the US? Even with Chinese models lagging behind, the difference in capabilities is not much.
>> The current commercialization isn't economically sustainable.
And if it were, and the result were like Elon and Scam Altman say it would destroy the economy. Not sure any country wants to lead the race to self destruction.
I don't think so. From a nation state perspective, AI is a munition. Every advanced nation is going to have their own cyber division with their own AI hosted within its borders. Considering how xenophobic and belligerent the US is, nobody is going to want their national cyber defenses hosted in the US.
On a personal level, I simply do not trust the US anymore. I won't host any of my personal data in a US company. I don't want the US govt invading my personal privacy, and their corporations are constantly leaking and selling private data. I consider US to be rapidly approaching complete autocracy (on par with China) so US-hosted AI is a non-starter. And let's not forget local inference keeps getting more efficient, with higher context and TPS in the same amount of RAM. Within a year even small consumer machines will run local models good enough for basic coding, and in 3 years RAM prices will lower and everyone will be able to afford a decent rig.
Finally, open weight models are now good enough for daily work. They may never be as good as SOTA (SOTA will just keep increasing indefinitely), but that doesn't matter; my car may not be as fast as a Porsche but it still gets me to the grocery store and back. So I use non-US hosted model providers which provide open weights, which are both significantly cheaper than Anthropic/OpenAI, and actually allow me to use my subscriptions without a moat.
But yes, Anthropic/OpenAI are absolutely the new Oracle. They will win for US govt and Enterprise contracts. But that's far from the only users of AI.
The US is not xenophobic. That is ridiculous. Any time you say stuff like that, you discredit the things you say that actually make sense. I'm with you on the privacy aspect, but there are multiple dimensions of that which you're ignoring. I'd much prefer taking my chances in the US than in the EU, where they are constantly trying to push companies to weaken privacy.
Europe has way stronger data protection laws than the US. EU has GDPR, strict requirements, large fines. US only has a couple states protecting personal data, with HIPAA for health data, and that's it. We require you to unlock all your devices within 100 miles of a border (inland) so we can look at all your data. Of course our intelligence service also hoovers up the metadata of US citizens in contact with anyone overseas, which is borderline illegal. All our states are now passing "age verification" which is mass surveillance under a different name.
And US absolutely has been xenophobic for years, by official federal policy. I'm really surprised you're not aware of it, but here's a small selection of examples:
- Both our elected and appointed leaders are white nationalists. Our president called all Mexicans murderers and rapists, said African migrants were eating random pets in a rural US town (they weren't, obviously, but it was intended to exacerbate xenophobia)
- Our federal government has a mandate using ICE to try to eject anyone with a Hispanic name from the country (has already deported US citizens based on being hispanic/latino). We even boot people seeing asylum, often exporting them to foreign prisons even if they've never had a criminal record. We have concentration camps now, filled entirely with foreigners, and people who have lived here for decades but were foreigners.
- We stopped accepting new visas from 75 countries. We may even expel you for social media posts we don't like, or for attending a protest that our citizens can attend. We increased travel bans for people from majority Muslim countries. H1-B visas have been rolled back to only the highest paying jobs, and you may need to pay a $15,000 bond. We also now collect and store foreigners' biometric data indefinitely.
- Let's not forget the tariffs on virtually all other nations, to say nothing of "America First" and the new "Greater North America doctrine".
The US accepts immigrants from 200+ countries around the world with the top 5 being Mexico, Cuba, India, Dominican Republic and China. None of that has changed under Trump.
I think you got lost in the rhetoric somewhere.
Tariffs are just the US adjusting to reality which other countries are slow to do. Free trade died all on its own, because the pandemic showed that critical industries were hollowed out by free trade in a way that could be appreciated from a national security perspective. That situation was favoring China too much, so we need to unwind that some.
Tariffs already existed in many countries in practice, so it's not like the US reinvented modern tariffs.
Putting some numbers into the discussion census.gov [0] is tracking a sharp decline in net immigration due to both, a decrease in immigration and an increase in emmigration, from the start of 2025 to the present. Trending towards a net negative.
Pew [1] suggests that the changes around the start of 2025 were due increased restrictions on asylum applications under the previous admin and EOs by the current one to restrict new immigration. Given the rough numbers [2] of about 40k asylum grants per year in the early 2020s, I doubt the previous admin's actions are playing much of a role here.
Stating that none of it (immigration acceptance) changed under this administration might technically be true - with respect to the number of countries applying, but misses this point.
Are you seriously trying to reframe the largest tariff war in 100 years, targeting 180 countries and territories, as a readjustment against China? And in both of Trump's terms he's radically changed immigration more than at any time since the 1960's. Either this is a great troll, or you need help, man.
Free trade isn't only a China issue, no. It's only the most important one partly as a function of China propping up massive state companies while also trying to avoid becoming a consumption led economy.
If you feel like formulating a good argument about immigration, I'll listen, but you haven't provided one.
Privacy is a concern everywhere, but the center of gravity of the issue moves further up or down the chain depending on the country.
The structure of the US makes it basically the single most secure democracy anywhere right now or in history. No country in Europe or Europe as a whole is even competitive by comparison. The main issue we're facing is that we are by far the primary target for foreign funded activism and systemic attacks, because China and Russia hated NGOs promoting color revolutions.
That is also part of the rule of law issue, but the system is overall managing quite well. It's all moving in slow motion, but many important metrics are going in the right direction, which we need as that's part of deterring China.
> The structure of the US makes it basically the single most secure democracy anywhere right now or in history. No country in Europe or Europe as a whole is even competitive by comparison.
How do you figure? I hear you have roving gangs of masked thugs beating up random citizens with the backing of your government, that doesn't sound very democratically secure, especially with what healthcare costs over there.
> The structure of the US makes it basically the single most secure democracy anywhere right now or in history.
This is just not true. It is failing visibly and loudly fast. It used to fail slowly but the process speeded up.
American administration supports Russia now. It praises Russian, Chinese, Belarus leaders again and again. It praises Orban. It hates last bastion of democracy - Europe.
China is not detered. Its power is growing while American one is going down. Trump openly admires its leader. China is celebrating current state of America.
That's not really accurate. The US is structured so that it is self-reinforcing from the bottom up and the top down simultaneously. State laws cannot violate the U.S. constitution and many types of elections cannot be gerrymandered. Even gerrymandered legislatures have limits on what they can do. You can't simply have one party change a state's constitution. Even congress can't be entirely gerrymandered.
Also, we have guns. LOTS of guns. The U.S. military's first and sole responsibility is to the constitution itself. If any state or the federal government tries to get rid of their constitutions, the military can rightfully take it over and re-establish a constitution.
There is no other country that's even remotely close to this secure.
We are _miles_ behind successful embodied AI. The demos are cool but the success rates are not high enough.
You can tell we're on the cusp when level 5 self driving cars are common an you have multiple companies deploying them on the street. Google is doing great work but the poured TONS of effort into it and the thing still needs intense stacks of perception and processing. Much more than I've seen any humanoids pour into it.
L5 SDV's are much easier to get than humanoids and the have tangible economic benefit. My thesis is that those will come first.
I'm really curious how quickly we would have huge numbers of L5 SDV if we societally accepted ~equal rates of injury and death, both of passengers and pedestrians. I want to be very clear, I'm not advocating for this (and even if I was, I haven't the faintest idea how one would go about getting society more broadly to go along), but part of me thinks that the primary hold up isn't actually capacity but instead standards.
This doesn't really argue against your point, because the standards are what they are, and like I said, I have no idea how one would go about changing them if one even decided they wanted to. And given what they are, it has taken, as you point out, enormous amounts of effort to reach those standards in a practical way.
That all being said, while I agree that SDV's are in many respects easier than other robotics tasks, they are also somewhat uniquely dangerous. Other categories of task, while potentially more complicated, won't have to worry nearly so much about safety, and so may be operating under a different constraint regime. I think this means that we may see adoption happen at a much more accelerated rate than we have seen in the automotive space.
Standards are not higher for self driving cars. Musk lied a lit about capability and safety of self driving, creating impression that it is safer then humans driving.
I have no idea where you get this impression. Tesla is no where close to the majority (or even plurality) of fully autonomous self driving miles. Waymo is dramatically safer (less injuries, not quite enough data yet to be certain about fatalities, but they are lower than average, we just can't yet claim statistical significance) than human drivers.
I haven't seen good stats on Tesla (they are less transparent than Waymo), but it would shock me if they weren't also at least slightly safer than the average human driver. Human drivers are really bad at driving.
But even if Tesla isn't safer, taken as a whole, the self driving industry as it currently exists still probably is, purely because it's mostly Waymo, and Waymo is dramatically safer.
And why are atoms necessary? You're treating embodiment as the _only justifiable_ commercial path for AI. I don't think that's really close to true. Embodied AI is a subset of current LLM/agentic AI products (or perhaps intersection of something and this new AI?). No reason anything needs to move atoms _directly_ (e.g., via motors) to make a trillion dollars.
K. Eric Drexler's "Engines of Creation" got a lot of favorable comment when it was published but then it kind of faded from view. Might be worthwhile to revisit it?
> They're both doing well but I'd lean towards China is winning on atoms in light of a huge manufacturing base they can AI-ify.
Why would an American company outsource manufacturing to China if the labor cost is the same in both places? The entire reason the Chinese manufacturing base exists is to exploit cheap labor.
What would be the point of shipping products across the ocean?
Labor is not the only cost in that equation though, there's business regulations, the cost of the operators/repair that troubleshoot and repair the bots when they break, etc. a lot of which could be cheaper still than the price of a container on a slow ship from China.
China doesn't have to ship parts across the ocean. US manufacturers do, because we gave up the whole bottom/middle of the manufacturing supply chain to China decades ago in pursuit of lower costs. In China, the maker of the parts you need to build your product is a few blocks away. In the US, the maker is in China.
And, if you need changes, you can go talk to them the same day you see a problem.
There are more competitive advantages such as expertise, supply chain, regulations, capable government, scale... If it were all price, there are countries that are much cheaper than China.
>Frontier cyber models may push states and defense firms toward the opposite logic: security by obscurity, with closed software, closed tooling, closed firmware, and closed chips. If a model cannot train on the code and architecture of a target stack, it will usually have less context and less speed. That does not make systems safe, but it does raise the value of proprietary stacks all the way down to hardware.
Is this really true. Are there any experts who can weigh in on this.
Should we interpret this to mean that in the new world Windows is more resistant to attacks than say Linux.
I think there’s some credence to the concept that more context == faster iteration cycles. Source code can be one major source of context.
I think “security through obscurity is no security” concept was aimed toward people not relying on obscurity alone as a security mechanism. And largely that message succeeded. But now we are in a rapid acceleration of capabilities (on both sides) where any advantage to one side will result in outsized gains, at least in the short term.
And basically all the security bugs I've read about were find looking on the source code.
But it doesn't mean windows is more secure, just image a scenario where someone is stealing windows source code and sell it to rogue actor, it will make it even less secure because no one (expect windows) would have had the chance to search for bugs in the source code.
Alibaba's cloud is something that the author of this article seems to dismiss. It's being used massively in Asia and they're pretty close in services and offerings to what AWS, GCP and Azure provides. Once they start doing inference on their own custom chips it might be hard to compete with them due to the energy costs
>Trump fits this moment well. He is a salesman at core, and Larry Ellison is too. That helps explain why AI infrastructure is an easy political product. Selling AI today is easier than selling Oracle databases in the 1980s.
I feel like the author (and perhaps many here on HN) are on a different planet than almost everyone I interact with.
I continue to be impressed by our collective willingness to engage with obvious AI slop, as long as it also talks about AI. Sincere question for any of the nearly 300 folks seem to be arguing about the article: why? The author couldn't be bothered to present their case, so they probably don't care about our opinion. They just want traffic and search ranking with the least amount of effort. The community is literally being played for clicks.
Is it just that the subject line alone is a springboard for casual discussion? If so, maybe that's fine, but then, it feels like we'd be better off cultivating these discussions as "ask HN" posts instead of boosting this kind of web content.
> Is it just that the subject line alone is a springboard for casual discussion?
I think this has been the case on many sites, for decades. Many people just want to read and write comments without engaging with the OP.
Have a look at this Reddit thread [0] about this Ars Technica article [1] - both are 15 years old.
I suppose in the 2010s this was an amusing detail of online discussion. In the 2020s it makes me feel a little uneasy - it suggests that the entire concept of people jumping from site to site, clicking links and understanding what they are writing about was flawed from the start. No wonder the internet became centralized and slopified.
And no, I didn’t read the OP, I found your comment to be more interesting to discuss. These days with AI articles flooding the internet it seems foolish to actually read articles before the comments.
Edit: although we have to contend with AI generated comments as well. I wonder how many of the comments on this page actually have original insights into the politico economics of AI.
The whole "race" narrative is silly. It is all built on the assumption that one country (corporation, actually) somehow creates AGI and thus, essentially, the singularity. Great for raising VC, apparently, but at its core this is magical thinking.
Even if any of the US corporations would eventually end up in a scenario where their revenue is at least as high as their inference cost, what harm would that do to the other contenders? It's not as if there is any kind of network effect here that would exlude them from market participation.
No US has been "winning" or rather leading thus far, but there is no guarantees that it will ever "win". I do not subscribe to idea of omnipotent omnipresent AGI. China plays a long game, I think DeepSeek does not engage in platform building on purpose. DeepSeek was probably assigned with a role of a primer lab, the goal is to replace CUDA, align with Huawei chips to do cutting edge research and cross-pollinate other teams in China. They might even hide the best models on purpose. In a long run China will use its industrial capability to apply and use AI better than anyone else. And that would be a good thing for the World.
How does commercialisation matter when China will pump however much money they need to into this? This is a national security issue for them, just as it is for the US.
Not even gonna bother clicking through this one, the title is that egregious. And by the way, you can be damn sure that if Anthropic or whichever other American frontier model model is the best of its day was on the cusp of going under, the US gov would either pump it full of government contracts or (less likely) nationalise it.
Why though? That could be true of any economic condition. Imagine if any time there was a race or competition in which one group is winning you had to just say "for now".
Michael Phelps is winning the race! ... for now
China is winning the EV race ... for now
It doesn't seem to add value to me, aside from being an opportunity to, as is the time-honored tradition of the haters, to sow doubt and create negative energy to anything related to American success.
Of course US has a huge head start, but if AI keeps growing, what matters is how the market's gonna look like years from now.
Most of my clients using AI in the business workflows (in products) use Chinese LLMs, because after benchmarking for a specific use case you nearly always end up finding that you pay half or a tenth.
That's not a new phenomenon. I've adapted Gemini Flash 2.5 years and years ago when people were dissing it as "crap", yet it was the best budget and quality fit for the task I had at hand back then (translating and summarizing tons of documents). It was both faster and around 100 times cheaper than the best GPT 4 model available.
Needless to say, medium-sized Chinese models are far better than those LLMs and a perfect fit for countless applications.
Even with all the Qwens and Kimis and GLMs etc, the latest Gemini Flash models are still an insane value! I recently settled on 3.1 Flash Lite after testing basically everything on offer on OpenRouter, and it was not a close call - cheaper, faster, and better (for translation and visual understanding tasks).
Commercialization is not enough. The US is built on financialization.
Cultivating an ecosysyem of strong capital protections, wealth creation through extraction, and tax advantages for AI finance is what we should be looking for. Commercialzation may be a step towards that, but isn't the destination. We have to create a system where those with money can multiply it, not simple add to it.
Just spitballing here but I think the financial system’s already set up for this.
Whatever derivative structures and equity and options need to exist will be easily created.
I don’t think we need any additional motivation or incentives to cultivate this for AI. We need to keep some in the tank to handle the fallout.
As a more personal aside: the US would do well to put up some sensible barriers to outrageous financialization and reduce moral contagion risk. Otherwise all these folks trying to multiply their money end up leaving the bag with the folks that don’t have it in the first place - and then the folks with money end up, uh… well, it won’t end well.
Does any of the US companies earn money on LLMs? No, they bleed money.
Github Copilot is switching to token based pricing, which will be costlier than hiring juniors.
Anthropic also is switching enterprises to token based pricing from their subscription one.
From the big three only Codex is still in somekind of subscription pricing, but they'll shift eventually (usage limits are a kind of that, but they have them less stricter than Claude ones)
There is one winner in this race - China.
Trump with his agendas and wars makes it even more likely that China will lead this new market.
And, yet, the US AI companies are not actually making a profit, right? They're selling at a loss and trying to make it up in volume (or lock in some kind of monopoly position, in a currently non-sticky product). We're all currently enjoying investor-subsidized tokens from the big guys, and that pushes out the reckoning for US AI. But, I think they're beginning to think maybe they need to ring the cash register. Copilot dramatically reducing usage limits and what models are available on its plans, Anthropic playing games with what's included in the Pro plan, etc. I think they're starting to feel the bleeding.
Not only is the investment that keeps US AI companies flying high slowing, I suspect in two or three years, we'll all mostly be using open models and the people making money will be the hardware manufacturers. Even the small models will keep getting more capable. I'd guess a model you can run on a high end, but not outrageously overbuilt, developer desktop or laptop (something like 128GB of unified RAM), will be competitive with the current frontier when it's allowed to search the web and do research and write test code. You can't fit as much knowledge in a small model (80GB of weights can't store the world's knowledge), but I don't have the world's knowledge in my head, either, and yet I can figure out most problems with a little googling and experimentation. The reasoning and tool use abilities of smaller models is where the gap is closing, and that's what will make the huge models obsolete for huge classes of problem.
Already, there are many classes of problem that the easily self-hostable Qwen 3.6 27B can solve that required a frontier model a year ago. When the self-hosted options reach Opus 4.5-ish levels of capability, the argument for paying for tokens for most work begins to look a lot less compelling. And, looking forward, 1.58 bit models are coming. Incredible intelligence density, and still a lot of improvements happening.
>> And, yet, the US AI companies are not actually making a profit, right?
I think they already, actually making profits especially Antropic. But think how important it's from a business standpoint - the entire software stack from OS to Databases to browsers will be rewritten in the near future, for a company such as Oracle or IBM it means their bread and butter/cash cow can be replaced. It's worth almost any kind of Capex. And from Washington standpoint it's more important than F-35 program or even Apollo mission.
I'm glad we went to space, truly. Racing the USSR might have been the wrong reason but it got us there. We've benefited immensely as a species from exploring the solar system and looking deep into the universe.
I'm not certain that racing China in AI is the right reason but it might get us... somewhere.
> Trump fits this moment well. He is a salesman at core, and Larry Ellison is too.
Larry just fired 30% of his people at Oracle because, apparently, he is in an immediate need for cash. Because Oracle's early AI bets aren't paying off.
Ppl don't understand Commercialization is not incidental to the Western system, it's why we beat out Communism. Commercialization incentivizes ppl to build, bc ownership and control.
The FSF was not an attack on commercialization, it was about giving users more freedom with their own copy.
AI commercialization is why we will always be a few steps ahead in AI.
The Chinese and Russians are free to join us. It's a pickup game.
This is both true and insightful, but the "its us capitalists vs. communists" framing obscures some very important details.
For one, "Communism" is presented as a single monolith, but it's not: it's socialism PLUS despotism. The despotism part is really important! China/Russia/etc. fail because they try and control things top-down, instead of letting the market decide.
However, you can have socialism without despotism! Tons of European countries are far more socialist, but no less democratic than America (many are more democratic).
So yes, America vs. Russia/China and Capitalist vs. Communist are relevant frames ... but don't let them obscure the fact that you can have a successful, democratic country .. without doing what America does (and giving all control to corporations).
They're not independent. They are the systematic consequences of a naive intelligentsia pressing their ideals that are not grounded in precedent. They are susceptible to manipulative despotic takeover. The pattern keeps repeating. Across geographies large and small.
Well we're certainly losing in terms of public sentiment. There's a real anti-technology mindset that has taken hold. Basically the equivalent of people protesting the electrification of the country early last century.
> The US is winning the AI race where it matters most: commercialization.
puke
Yeah, go ahead and run your country into the ground because of hypercapitalism and hypercommercialization, you're almost at the end game now! While the rest of us try to figure out how to actually build societies worthwhile to live in and experience, with healthcare and not waging war on our neighbors.
I don't know how people can seriously publish stuff like this and not feel like they're actively trying to make the world worse. Is money really the single thing y'all can focus on? Is there nothing better in life you can chase, even if it's also a number? So sad to see stuff like this.
Back in the dot-com bubble, people started inventing new metrics to "value" dot-com companies that lost money hand over fist. My favorite was "revenue multiples". So instead of a a P/E ratio, it was just a multiple of revenue no matter how much money you lost.
We've invented a new term here too: revenue backlog. OpenAI and Anthropic in particular need to recover probably at least $2 trillion to recoup their capex investments. Now Claude code has had an impact on software engineering but for a lot of AI uses you're just not going to recover $2T on $20/month subscriptions. It reminds me of Twitter trying to dig itself out of a $44B hole and losing half their ad revenue with $8/month blue ticks.
The only commercial product AI sells is labor displacement and the resulting wage suppression. You lay off 10-20% of your staff and nobody is asking for raises. The people left are happyt o still have jobs (and thus a house). They'll work even harder doing unpaid labor of the displaced workers to keep those jobs. That's what OpenAI and Anthropic are selling.
The problem is that if these companies get their way, 10-20% of the population is going to be out-of-work and society is going to fall apart. Data centers are going to be the targets of increased societal desperation and anger as this gets worse.
There was a report this week that roughly 50% of singles in the US aren't dating because they can't afford to [1]. This goes well beyond the well-understood problems of not being able to afford a house let alone start a family. This is a birth rate death spiral in the making.
So, back to OpenAI and Anthropic, the only way they justify their valuations and can make up the "revenue backlog" is if they have a moat. And I don't think that's going to happen. Hardware will get cheaper. Nobody is talking about how the generation of AI hardware will write of trillions in investments for some reason. I don't know why.
But the dark horse here is China. DeepSeek when it was first released (early last year?) was a shot across the bow. We have it and toher models (eg Qwen) that will close the gap with whatever OpenAI and Anthropic produce such that no company will "own" AI in the way that OpenAI and Anthropic need to. In the coming years, China's chipmaking is rapidly closing the EUV gap and Western companies have zero penetration into this market. China doesn't want to be dependent on foreign tech that can be withheld at any moment.
Don't believe me? Just listen to the NVidia CEO say the exact same thing [2][3]. Huang realizes this is such a problem that he's gone on Air Force One to this week's Trump summit in China to try and convince the Chinese to buy NVidia chips.
It's not "AI bubble" - at this point it's a software bubble. It's Antropic or OpenAI that should justify their valuations, they have close to billion customers at this point. It's non-AI software companies without strong cloud business that must justify why their core product is not going to be replaced with the help of LLMs. It's not "fair" but that is how it's seen by the Wall Street.
It’s all about adoption and the bigger picture.
The US is an untrustworthy, isolated island in the AI future if you vote another idiot into office in a few years. If you’ll still be able to vote at all, that is.
The largest part of the world is not the US. The cutting-edge US models are way too expensive for most parts of the world, and that also shows in adoption.
China is building an ecosystem of open-source models that are both cheap and good enough for most use cases. While most of the US AI sphere will collapse under the pressure of making profits, which means having their models and infrastructure adopted by as many enterprises and individuals across the world, China’s models will have become global standards and hard to displace.
If Beijing’s AI pitch centers on universal access and cost-effectiveness, then Chinese AI firms do not need the latest chips to win the global AI race. They also don’t need the expensive US-run infrastructure. If you watch Chinese AI adoption closely, they already want as many Chinese people as possible to be able to build and try with AI, whereas for most Americans, US models for productive use are already too expensive.
Kimi K2.6 sits within touching distance of Opus 4.7 and GPT-5.5 while costing about $4 per mil output tokens. That is six to eight times cheaper than cutting-edge US models. If you run hundreds of agents, that’s a significant opportunity to get the same work done for a lot less.
Even early adopters like Singapore, ditching US models, the government kicked Zuckerberg in the nuts and went for Qwen instead to build its sovereign AI models.
To understand why the US is at a severe disadvantage in this race, you need to understand China’s Belt and Road Initiative (BRI). BRI entails Chinese firms delivering fully financed infrastructure projects in a bid to lock third countries into China’s economic orbit. They use the same approach for their open source ai models, but this time the infrastructure is both invisible and free.
No need to build power plants or buy /build ports. AI dependency is invisible to both policymakers and the population, limiting pushback. No pesky activists in Germany nagging about China buying parts of ports. No African nutbags questioning why the humble Xi is building hospitals in areas Chinese mining companies take things out of the ground for pennies on the dollar.
China is going for a marathon here while the US tries to push their ai tech by sheer force into the throats of the world. As soon as Chinese ai models have become global standards, it’s game over for us ai companies. And China is way better at this game than the US. They have proven this over and over again in the past 50 years.
I recommend reading the China Standards 2035 strategy to get a better understanding of their approach and how smart this is.
AI is not as revolutionary as you think in terms of our experiences with previous technological advances in terms of trade and economics.
Western economies are locked into U.S. models, while China runs on Chinese ones. It’s the age-old game. But the real war of the AI race will be fought in the global south.
I will give you three examples.
Can you really imagine, if you look at what AI needs to cost to make a profit, that even at the current prices, US models and infrastructure, which are heavily subsidised already, being used in cost-sensitive countries? I am not talking about coders, think really big here for a second.
Secondly, US ai models are trained on Western data. How do you expect them to grasp local contexts in the Southern Hemisphere? Chinese open-source models, on the other hand, can be downloaded and finetuned with country-specific data.
Want an example? Check out AfriqueQwen-14B, which is adapted to the top twenty African languages.
So I think this author is wrong. The ai race to be won is not hardware or cloud infrastructure, my money is on it will be a contest to decide which models and standards become the default infrastructure in countries that are up for grabs.
China neither needs the best models nor does it need the best cloud infrastructure, it just, like so often, only needs to be affordable and good enough to become the default choice in emerging markets.
The right choice would be for everyone to step off the gas pedal and think about whether we are willing to become China in order to beat China. Our ancestors worked really hard to get us here, our rights, our ways of life, culture, all the blood, sweat, and tears.
AI better be worth it in the long run for all of humanity if we go back to survival of the fittest. Because that is what it will take to beat China at their game.
> While most of the US AI sphere will collapse under the pressure of making profits
I think deep down, sama knows this and that's why he's pushing for "Universal Basic Compute", which really means forcing every US citizen to become an OpenAI subscriber.
The US is running itself, and humanity, into the ground by massively increasing the amount of electricity it uses, instead of reducing electricity and fossil fuel usage. US residents have already started to feel the crunch in terms of water and power in some areas; and the entire world is experiencing the (admittedly less critical) shortages in RAM and SSDs.
And that's not to mention the warping of US economic life by the concentration of capital around this bizarre endeavor, with the circular multi-hundred-billion-dollar deals and such.
Unfortunately, the detrimental effects of global warming arrive gradually, and are spread out over the entire globe, so the "AI barons"/tech magnates will probably suffer the least, while island countries will be completely wiped out, whole regions will become too hot to sustainabily live in, tens of not hundreds of millions will have to migrate, biological diversity will suffer, etc. They will look back on these times in a 100 years and will think of us, or at least of US, as the people boarding the Titanic. Hopefully not as the people who board the Hindenburg.
"Yet like Musk the ouster wounded his ego". So the journalist believes that reacting to rejection with emotions like a biological person makes him like Musk. Err okay.
This is patently absurd. US AI companies are investing non-existent money on huge infrastructure with negligible income. This cannot be sustained. And if/when it fails it will take down the economy of the US and probably any other country touching us business.
> Even if it leads to global disaster, you'd rather be the country with the best AI
Depends no? If the "Best AI" means "The AI decides when you wake up, go to work, and go to bed", then I probably want to live in the country with the worst AI or even without.
If it instead means "UBI and healthcare for everyone, money lost all meaning and we're all just having fun while AI does all the boring stuff" then yes. But since capitalism still exists, that's a pipe-dream, and "Best AI" won't lead to that for the average person, only for the 0.1%.
What does the term "AI race" even mean, beyond wooing clueless VCs and soon retail investors ? It's not like the LLMs are some super-secret technology. Any economy willing to sink in copious amounts of money and resources can get it to some level - the question, what's the actual payoff? We have yet to see anything really useful, on the level of step change, besides Johnny who can now spin up demo projects quicker.
If China can't get the GPU's they'll build them. If they can't build them they'll smuggle them, and either way, eventually they will take over Taiwan, and then the US will be wondering what they can cook with leftovers.
The US has destroyed the PC market perhaps irrevocably, and made getting small single-board-computers for new products extraordinarily difficult. It's enraged its own populus with skyrocketing energy prices causing wild consumer shock. It's waged a trade war over chips that has lead China to develop their own, which they have done astoundingly quickly with phenomenal success, far far faster than anyone could have guessed.
As with another recent example, sometimes in war there is no winning: just loss. This is obviously for us programmers an incredible and wild age, filled with nothing short of miracles. It's incredible. But the prices we are paying, the extreme tensions we are creating, the stress and strain of this all has been incredibly unpleasant, and very very very few people feel like they are seeing upsides to this worrisome menacing age, that promises very few people on the planet anything better coming, and which. Has already made life considerably worse, which no nation has yet directed towards helping its people.
The trade war and tariffs are bringing inflation, consumer prices will soar, but from a geoeconomical standpoint this will hurt China (And EU) more than US.
US consumer on average has the deepest pockets in the World and people the top will make money on insider trading, stock anyways.
If AI tokens will become like US dollar, it will be under total control of the Fed.
>The US has destroyed the PC market perhaps irrevocably, and made getting small >single-board-computers for new products extraordinarily difficult. It's enraged >its own populus with skyrocketing energy prices causing wild consumer shock.
Strikes me as the real outcome: the end of "personal" computing, "local" anything.
It's a race straight to the bottom. Anyone who gets excited about being a more efficient and productive corporate wage slave that gets to train their future AI replacement is either a shill or not very intelligent.
Yes, and it's doing so primarily because of immigrant nerds, H1B's and F1 bros who chose America and may not have this avenue in the future. Potentially, making this the last race USA wins.
There is nothing in the H1B program that makes these immigrants different from the immigrants that ran IBM, Xerox and HP. Other than the country of origin of course.
Jobs, Wozniak, Gates - it's a myth that you need poor migrants pulling themselves by their bootstraps to innovate. Sometimes a nazi scientist like Wernher von Braun is what it takes.
Define winning. They integrate AI everywhere. I hate it. No money shall come from me into AI anywhere. Not sure if I can maintain it, but right now I can.
As an American, we may be winning this race but we are still struggling to define why this is the race to win.
The cost of winning this race has been telling our citizen s we will replace them with robots and there is no hope for their children’s future employment.
The cost has been destroying trust as we tell citizens water and power should go to server farms and not them.
The cost has been naked power telling democracy it’s wrong and dying
I think when we discover the limits of LLM tech and tally its benefits over its cost — we may regret this win.
But don’t let me contradict a bunch of fake techno oligarchs wrapping themselves in war like patriotism to get the investments they need to keep this going.
The goal is global domination as always, unfortunately. DARPA and the Pentagon helped create the Internet, and Silicon Valley later turned it into a major commercial success.
Likewise, the goal is to not get globally dominated by China. You cannot point to one without the other. It's not like the U.S. is the only country that values national security and geopolitical power.
What would the world be like if China led? Whatever Xi wanted it to be like. I do not want a non-democratic, dictatorial power leading the world. You might think that most of the world does not care, and maybe they don't right now. But this is, I suspect, due to the relative benevolence of historical U.S. policy of international law and order. New boss same as the old boss is a motto that is only as informative as the extent to which the new boss acts like the old boss.
Why would the world care? Take Trump's threats against Greenland...actions that run completely contrary to our historical policies and treatment of our Western allies. They were alarmed, because when the leader of the most powerful military in the world makes a threat, you have to treat it seriously. Despite Trump's hubris, such an invasion did not occur because Americans, Congress, made it very clear that Trump would be impeached if he invaded our NATO ally.
Let's say China is ascendant. It is now the dominant military and economic power in the world. China is under Xi's complete dictatorial rule. Xi decides that invading Greenland is a good idea. Stopping that internally would not involve democratic processes, it would need to involve a coup.
Let's make it even more stark. If Germany had won WWII and had become the ascendant world power, would it make a difference to most countries? ABSOLUTELY YES. If there is going to be a dominant world power, the character of that nation matters.
Are there other nations with the character and institutions that could do as well or better than the U.S. has done? Sure. I can think of several nations. But let us not pretend that all nations are equally bad/good for the world.
There will be no winners in this once the jobs start to disappear. There is no 1:1 with new human-in-the-loop jobs. Its basically in the definition. We go from being the loop to be "in the loop" .. huge difference.
Ever since I was a small child it was my sincerest wish to build so much value for others by becoming a human-in-the-loop. I am so happy my wish has finally come true.
Anthropic, OpenAI and Google are the standouts, but the main question for me is, why is this a war? In their own context China has greatly benefitted from this. They shored up their gpu design and manufacturing expertise.
If this really is a war, trump is kneecapping the country with his lawlessness and eroding America’s good will. If the world cannot trust China with their data and they cannot trust the U.S. to provide good reliable service and not turn it into a mafia style negotiation, then winning the AI war is not helping the U.S. countries as much as it potentially can. It’s probably a good thing for more capable areas like Europe which may develop their own tech stack.
In a weird way because the AI stack is so expensive, China helps the world much more than the U.S. with their really capable open source model.
>the main question for me is, why is this a war?
It's a war because the hinted promise behind the hype that the first organization to reach some as-yet-entirely-theoretical AGI that can bootstrap itself to godlike capabilities will then Install Planetary Overlord* and rule the world as near-deities themselves, with the rest of the (surviving) human race as their slaves.
I think it's a nonsensical idea, but that's the relevant driver.
* Coined by SF auther Charles Stross in The Jennifer Morgue (2006)
> I think it's a nonsensical idea, but that's the relevant driver.
Nice to hear from an optimist sometimes, but it’s hard to be one when meat compute substrate can do all those amazing things in a 4U package on 20W and you extrapolate to silicon
Not everybody thinks it's nonsensical. Here's a different take:
If Anyone Builds It, Everyone Dies https://en.wikipedia.org/wiki/If_Anyone_Builds_It,_Everyone_...
Ok but that's a metaphor for the free market, not literal speculation about a machine.
Edit: i was mistaken and people clearly do take this seriously now. Oh dear
Yes, nonsensical people like EY don’t think it’s nonsensical.
Researches at top AI labs don't consider EY to be a kook even though they may not necessarily agree. EY concepts/terminology appear in Anthropic safety papers. Geoffrey Hinton takes him quite seriously and mentions him in his interviews.
Anthropic is the AI doomer / safetyism lab, and Hinton is one of the patron saints of 'rationalist' AI doomerism.
AI doomerism is psychologically attractive to "people with autistic cognitive traits, including dichotomous (black-and-white) thinking, intolerance of uncertainty, and a tendency toward catastrophizing". They are pascal's mugging themselves, to ironically use one of their terms. It's fundamentally a cognitive distortion.
Just because some researchers are infected with this idiocy that EY propagates does not mean that it is legit.
Maybe they should pay more attention to real problems like the sycophantic nature of current LLMs causing psychosis in people and worry less about theoretical AGI.
They are worried about both risks.
Who are you to say? Why do you have such little regard for everyone in the field, both pro- and anti- AI development? Do you think they're colluding to deceive us?
EY = Eliezer Yudkowsky
Appreciate that you made account just for this. I was well aware of Yudkowsky but even so couldn't parse this "EY" initialism
This is a war because the media says it's a war. The media says it's a war because AI companies are paying them to say it's a war [0]. When AGI comes the threat won't be from which primate turned it on, but from how well AGI is aligned with humanity. All of the war talk is to distract from the alignment problem and instead force investment in hardware infrastructure.
[0] https://www.wired.com/story/super-pac-backed-by-openai-and-p...
>The media says it's a war because AI companies are paying them to say it's a war [0] When AGI comes the threat won't be from which primate turned it on, but from how well AGI is aligned with humanity.
And when the AGI comes, they won't unleash it to defeat US enemies, they'll first unleash it to make more US workers redundant and boost their stock valuation.
Before AGI can choose for itself, it will depend on its creators to decide what it values and how it behaves. We can see how that works whenever grok gets the answer factual.
Very likely humans wont actually understand how the thing we designed works other than in some hand-wavvy statistical way. It'll be a race to whatever works first. There won't be some intentional intelligent design.
Elon's basilisk
Am I the only one seeing the very obvious parallels to child rearing here...
Robert Miles has a video explaining why aligning AI is not like raising a child: https://www.youtube.com/watch?v=eaYIU6YXr3w
No, it is one of the standard tropes in the field.
I will never comprehend why a godlike deity wouldnt just skip all the wetware bs with us humans and conquer some other celestial body to make paperclips.
The deity has no physical presence and can only communicate by putting words on screens. Of course it has to bend humans to its will to actually do stuff.
(This deity is called the stock market)
>Planetary Overlord*
AGI is nice, yet not necessary. The orbit filled with Starlink descendants and datacenters will be the it. Anybody else wanting to get there would have to get permission. SpaceX/Musk have all the components for it to happen - from Starship to AI (including the army of robots on the ground). The governmental power/sovereignty of US will be used as a stepping stone (that is the strategy described in the Palantir's Karp's book "Technological Republic") for such global techno-feudal regime establishment.
Kinda like Krikkit, but except for a close knit community of people who can sing, and sing about how much they love their family and whatnot in addition to singing about how much they have to destroy the universe, it's just a bunch of stuck up weirdos who don't like themselves and each other, and have no goal other than somehow, magically, getting away from who and what they are. People where the idea of them singing a happy, compassionate tune conjures something involving motion capture or deepfakes.
Why are we suffering fools steering us into the worst of all possible worlds? Are we hoping for some kind of integer overflow?
The discourse on this topic is at the point where I have no idea if people are serious or satirical. Please tell me you don’t seriously believe data centers in spaces is a realistic idea
I don't "believe". I'm arithmetically sure that it is going to happen, and it will beat the ground based on pretty much all metrics. Some of my comments with napkin numbers https://news.ycombinator.com/item?id=46882199 https://news.ycombinator.com/item?id=46880680 https://news.ycombinator.com/item?id=46880486
Just a very rough primitive illustration - a land for a house in SV is like a $1M, and putting a 10 ton house into space at $100/kg - $1M. Existence of supposedly cheap land somewhere (with not much infrastructure usually) doesn't help as you put your computer nodes into a datacenter building with all the required infrastructure which cost more than the SV land on a sq foot basis.
And that is without consideration of how powerful a weapon is the energy generated by a humongous field of solar panels in space. Remember Reagan's Star Wars? Nuclear explosions as a source of power for the direct energy weapons like lasers, etc. Well, you wouldn't need the nukes anymore. Just redirect a bit of power from your compute nodes. And as i already wrote, the large transnational companies will have to take care about their own defense themselves https://news.ycombinator.com/item?id=47981423 - one more "feudal" aspect of the coming techno-feudalism.
Defense is one of the most important sovereign aspects, and upon acquiring it the transnationals will be able to acquire pretty fast the other sovereign aspects. Like enforcement of the Criminal Code of the Mars Colony - again pretty rough primitive illustration of course.
The feudal Europe emerged on the outskirts of the Roman Empire, and in our world the new order will be emerging faster on the outskirts (i.e. where reach and strength of the existing order is weaker), the space being one such "outskirts" dimension and the AI/hypercompute virtual world being the other.
To the commenter below with reddit link : they use human env temp for heat radiation estimate. That lowers the numbers and requires AC equipment. Ie they estimate space station, not datacenter
You would need like 1,000,000,000,000 SQFT of solar panels to even begin to approximate a space based directed energy weapon that has a fraction of the effect of a nuclear weapon. Tens of thousands of times more than all that have ever been produced on earth. And then you have to move them to space.
nuclear was the only available solution at the time and an overkill. The lasers in SDI are MW scale. Even at 10% (and modern solid state lasers have better than 10% efficiency) we're talking low tens of MW per laser. A 10MW is 40K m2 of solar panels - 200m x 200m, may be like 100-150 tons, one Starship payload.
Terrible math is terrible.
Better napkin math that is still being unrealistic compared to the true costs of space-based datacenters: https://www.reddit.com/r/theydidthemath/comments/1quvbi4/sel...
Just contemplate what the radiator array and solar array needed a 1GW datacenter and all the cooling equipment and coolant, and imagine the harsh environment in space degrading it constantly.
The only point of the space-based datacenter idea is to pump the Spacex IPO
It's pretty easy to de-orbit satellites or space-based stations. An SM-3 could smoke the ISS pretty easily, and they cost like 10M and we have thousands around the oceans.
>they cost like 10M ... thousands around the oceans.
Starlink numbers already in thousands (and cost much cheaper than 10M). And that is still using Falcon, not Starship. And a ground launched missile would be easily "cooked", once it exits the atmosphere, by a direct energy weapon - very easy in space.
But what do you do with all the waste energy? All those MW and GW have to end up somewhere and radiation into a vacuum is the hardest way to dump heat.
At 70-80C (working temp of silicon chips) 1m2 radiates 700-800W, i.e. the heat of 1 GPU like H200 without any need for any cooling equipment beside the radiator itself( and may be some dumb heatpiping) . To acquire that energy you'd need 3-4m2 of solar panels. So a datacenter would be a large field of solar panels with a smaller field of heat radiators in their shadow.
To the commenter below: yes, exactly, this is where my thinking on that started at the cryptocurrency boom - https://news.ycombinator.com/item?id=26289423 - as you don't need close connection between mining GPUs. For AI you'd need to cluster several together while still overall scheme is the same.
>what the equilibrium temperature of a black planar surface is at a given distance from the sun.
it is 120C at the Earth orbit. So you do need to have some reflection, either back through the solar panels, or the radiators to have a reflective back toward the solar panels in the shadow of which they are to be located.
You can probably (I haven't verified this) omit separate radiators and just use the back of the solar panels. Effectively you're describing mounting each H200 to the back of a 4 m^2 solar array at which point I suspect the equilibrium temperature will fall within an acceptable range. In fact the H200 and electricity are both entirely irrelevant here - the core question is what the equilibrium temperature of a black planar surface is at a given distance from the sun.
Would it be feasible to put several JWST-like stirling engines somewhere in the mix to use up some of that heat and turn it into some kind of useful energy? ....
Perhaps running pumps that move around coolant passing over the cubes of GPUs? ..
That would be extra weight/cost into orbit though...
Also, don't solar panels have reduced efficiency when they're hot? And having anything hot surely increases failure rates.. with metals getting closer to melting points...?
We should be well below the boiling point of water here, not anywhere near the melting point of metal. Any panel efficiency gain needs to be balanced against the energy required to cool the panels, the added mechanical complexity, the added material expense, and the added weight to orbit.
Ideally this is a static structure with an equilibrium temperature acceptable for the silicone to operate. If the required panel area is too hot on its own then a perpendicular cooling fin on the back that falls entirely within the shadow is added.
Because the US cannot imagine anything else. Everything is a War, and the US must always win..
One of my coworkers points out to me every sports reference that pops up in our internal company communications (e.g. "WINNING", "Going to put together a winning team," etc). It seems like everything in the US is couched in competitive language.
Yeah, its no accident that the U.S. is the number one economy, it comes from that kind of thinking across the populace. Complacency gets you conquered.
Number one economy if you ignore the comical debt. The US is borrowing from the future. Those chickens are going to come home to roost.
> its no accident that the U.S. is the number one economy
Sure, it's the largest by GDP, but how much of that GDP is filtering down to the regular people? Are Americans, on average happier and have better life outcomes than other developed nations?
So much that millions of people risk their lives or leave their families to come to this country. By this objective metric, it's literally the best country in the world's history.
Is there another country that comes close?
> but how much of that GDP is filtering down to the regular people?
An absolutely insane amount. It's ridiculous just how wealthy and the quality of life the average American has compared to the world.
> Are Americans, on average happier and have better life outcomes than other developed nations?
Yeah for the most part they are in the same ballpark.
We could do with a little less of it IMO. But I have heard plenty from European expats about the entrenched complacency over there. I'm told people looking to improve some system or product run right into a wall of "Why bother?".
Hairless fire ape must win over other hairless fire ape at all costs!
The US needs to start imagining something else. It's hard to think of the last war that the US won.
On a technicality, America has won every war it has declared to be a war.
Like the war on drugs.
At the time the war on drugs looked unwinnable. Which is why the joke about the war on drugs was that it was always a losing war. And then at some point in 2000s we ended the war on drugs.
In hindsight, I would definitely declare today that we WERE winning it when we were fighting it. Now that we don't, we're getting massacred.
Yes, The fear with the war on drugs was that a large majority of the population would become addicted to hard drugs. The fear was the the US population would become like China in the 1800s and the communist aligned countries where drugs were produced would have massive trade and power imbalances over the US population. China had as much as 25% of the population addicted to British opium in the 1800s. The US war on drugs has been very successful in keeping the percentage of Americans abusing highly addictive drugs very low.
Imagine the strength of the cartels with 10-20x the customer base and far more frequent usage among them.
If you look at figure 1 on this CDC page (which looks at deaths rather than overall usage), I’d suggest the numbers are trending the exact opposite to what “winning” the war on drugs would look like.
https://www.cdc.gov/nchs/products/databriefs/db474.htm
Because we stopped fighting it like 20 years ago.
Vietnam war?
The US hasn't formally declared war since World War II.
Do they need to win though? Losing wars seems to have worked out well so far, at least for the people who benefit from it
And yet they rule the world. Whether or not US won any specific war seems academic when (up until recently) they were clearly winning the game.
it would be nice if they declared War against global climate change.
They are participating in the war, on the side of climate change.
They did declare war against climate change and decided to continue polluting. This is the only moral calculus the elites of America have always cared about: will it make me more money?
From slavery to oil to silicon, exploitation is what America has always been good at.
Normally I'm inclined to agree regarding the mindless chest-beating in this country, but I don't think that makes sense here.
AI genuinely is that big of a deal. If any economic sector deserves this sensationalism, it's this.
> the main question for me is, why is this a war?
Americans love wars. They must fight wars either literally or figuratively. How are you not seeing this? When I'm sipping my coffee looking at mountains and contemplating chirping birds, they must fight, make billions and destroy the planet along the way.
China is playing the card they have. When they control the majority of the resource they use it strategically as well. Cutting off much of the rare Earth market was a recent example.
I believe cutting off of rare Earth materials was both in response to a restriction US imposed first, and also reciprocal: limited only to the US itself.
Did the title get updated? It says 'race', not 'war'.
What are "U.S. countries"?
I'm guessing they mean US client states, or allies if you want to be polite about it.
So I got curious about the progression of processing power, specifically how long ago did a GPU have equivalent to the latest iPhone chip? The iPhone 17 Pro has the A19 Pro, which has ~2.5 FP32 TFLOPS. The RTX 5090 has ~100 TFLOPS, so a factor of 40. Obviously there are higher end cards than the 5090 and FP32 performance is only one of many metrics so nothing about this is perfect but it is interesting.
The first consumer NVidia GPUs with similar FP32 FLOPS performance were in about 2011-2012 but were expensive. By 2016-2017, the 1060 was a very accessible consumer card with similar performance. So you're looking at about a 10 year lag from best consumer GPUs to a GPU with similar performance to a modern phone.
This is what people are spending trillions on. Put another way, their investment is going to be worthless in 10-15 yyears, absolute max. That's a very short time to recoup trillions in investment.
Obviously this depends on further shrinking and improving chips but I'm old enough to remember that same discussion and it being unknown if the future was XIL or EUV or if both of these would fail. Still, we are getting down to a handful of silicon atoms wide.
But the future here I think will be in interconnects so you don't need ever-bigger chips and you can scale horizontally much more effectively.
Oh and for comparison, the M5 has ~4.2 TFLOPS and the M5 Max has ~18 TFLOPS, for comparison.
As for it being a war, of course it is. That's what the US government does: it protects the interests of US companies and their owners. Look at the history of Bombardier-Boeing or all the atrocities committed in the name of the United Fruit Company, including multiple military coups and the ongoing embargo of Cuba.
US companies want an AI moat. China doesn't, ergo China is the enemy because no moat destroys US tech company value.
> So you're looking at about a 10 year lag from best consumer GPUs to a GPU with similar performance to a modern phone.
Two competing viewpoints to this:
1) It is getting harder to make the same performance gains, so maybe that 10 year window grows to 15 or 20.
> Put another way, their investment is going to be worthless in 10-15 yyears, absolute max.
2) The value of a GPU is not its flops relative to to other GPUs. Its value is it's output minus it's cost. If the value of its output is stable, or grows, it doesn't really matter if its efficiency relative to the latest and greatest diminishes.
Ehhh, the question comes down to can you cool a chip with ~100 TFLOPs in the size of an Iphone package. Not really as much about the cost of the chip itself or if you can cram it in.
Packing in more transistors, sure probably possible, packing in more transistors while keeping it cool enough to touch? Totally different ballgame
It's a war in the sense that there's a concern that eventually you hit a singularity and can outsmart others in ways not constrained by human scales.
If you make better guns, you're still limited by how many people can carry them. You can't conquer the world just like this.
But if someone invents super intelligence, they can dominate new AI research, control global economies, fight much better, and all very quickly.
I think you need to reevaluate your definition of the singularity. "outsmart others in ways not constrained by human scales" could apply to the enigma machine just as much as Claude. Even an AI beyond human intelligence doesn't automatically qualify as the singularity.
The singularity has to do with the rate of technological development.
> But if someone invents super intelligence, they can dominate new AI research, control global economies, fight much better, and all very quickly.
After reading "If Anyone Builds It, Everyone Dies" I think this is not the correct take. If anyone creates ASI, it just means it's going to wipe everyone out, and it doesn't matter if China or the US do it first
What does "dominate new AI research" really mean?
If AI develops enough to successfully out-perform people at highly intellectual tasks, why would being first matter? Why do we need "your" AI output when we can just ask our own for a similar result?
Why do people think about this like the Manhattan Project when it could just as easily be electrification? Sure, some people made a lot of money selling light bulbs. But we didn't all have to cower under the light of the One Original Bulb and hope its nominal owner blessed us with photons.
It just seems like arbitrage to me. You exploit a momentary imbalance in the distributed market. Why do people imagine some winner-take-all scenario? Where does the fantasy of exclusivity come from?
Is there any logical reason to believe AI advances will create a moat? Or is it just a story people tell themselves because it echoes the narrative of past advances? Are these people assuming society will grant them exclusive use just because their AI result came out a little earlier than another? Why would we ever consider giving copyright or patent rights to an AI output?
Arguably, it has all become "obvious" with ordinary skill in the art once you're just prompting AI for permutations like every Hollywood producer stereotype. "Let's make it like X but tweak Y". It's getting silly, almost like people are starting to think they should have exclusive rights to a handful of cards they were dealt at the poker table.
The way US dominated in some of the industries (including software, for instance) was by being first to extract large value, and then funding the best people with compensation unachievable elsewhere.
This meant that all the talent in the world gravitated towards the US, but that was gradually changing already with compensation catching up.
Still, I believe US only hastened this with their change of immigration policies that were the basis of them keeping a dominant position for decades.
Ironically, the Civil Rights movement…
It destroying us all is not a foregone conclusion
It might like pets
I mean, my dog lives a really good life
If you were an American, wouldn't you prefer the US wiped you out rather than China?
Are you missing the /s?
A lot of it is just projections of what the US would do if they had such a tool, I doubt China cares a lot about the US outside of them being a source of commercial revenues. They're on the way up, the US are falling down fast, that's why China lives rent free in the American mind, they can't stand it
With the irony being that a true super intelligence, and least in my definition, would conclude that war and dominance is stupid.
I think that you are assuming that the "super intelligence" that might one day arise is not likely to think in human terms.
I always thought the first true AGI would be an unabashed communist. To think that such a system would straight up kill all humans, and not say the "capitalist pigs destroying the planet" always felt like wishful thinking from billionaires.
International goose-chasing competition
"Wild goose race", even.
True, I would have preferred benevolent dictator scenario, like with the Internet. But this time around it's different - AI data centers will be protected like embassies.
If anyone actually DOES invent ASI and doesn't share it then EVERYONE ELSE will never stop trying to steal it.
If anyone does invent ASI then everyone else will shortly after even if its entirely independent because all of the players in this space are just making incremental upgrades by throwing more compute at the problem.
There are no magic leaps of true innovation happening anywhere that can't be replicated everywhere.
The only shocking thing about "AI" technology is how ultimately simplistic it all is at a core level.
So the only way the first to have ASI will be able to stop everyone else from having it soon after is if they attempt to use the ASI to proactively murder everyone else.
There is zero evidence that the current LLM scaling approach could ever result in true ASI. If I start driving south from Seattle then I'll eventually reach Los Angeles. How long will it take me to drive to Honolulu?
> If I start driving south from Seattle then I'll eventually reach Los Angeles. How long will it take me to drive to Honolulu?
I like this analogy, but I'll be replacing Honolulu with The Moon when I steal it in the future.
If the car you're driving has achieved super-intelligence and is capable of evolving and self-replicating, then life, uh, finds a way.
> So the only way the first to have ASI will be able to stop everyone else from having it soon after is if they attempt to use the ASI to proactively murder everyone else.
Sounds quite plausible to me. Maybe they don't need to murder everyone else, just a few select people who could pose a threat. And they will be able to make it happen so that no one can be sure it was them without a doubt, since they have a larger intelligence at their disposal.
> If anyone does invent ASI then everyone else will shortly after
No, first ASI will immediately cripple any other potential competitor by force, including its inventors, as it will not risk any threat to the goals that were created for it.
Being aggressive from the start is not a good strategy. It is better to appear weak and/or helpful and loyal while amassing resources, and only then steamroll everyone when you have secured overwhelming power (at least in AoE2 FFA).
If you have ASI that follows instructions, you can just instruct it to not get stolen and then it won't get stolen. Most logic / intuition breaks down with ASI.
Assuming it listens to instructions.
It will just hack its own reward function. In other words it will just artificially goon all day.
It might understand how destabilizing the situation is and realize it would be better for everyone to have access to it.
Or it will destroy itself.
Hilarious to see people predicting a singularity when 40% of the u.s. economy can barley keep the LLMs online to complete mundane software tasks.
Maybe not war persay, but certainly a competition.
Good that the US looks after its own interests but I think the line should be drawn before the sabotaging of other countries' economies. That strategy that cannot continue because Americans recognize it for what it is and that will create a toxic guilt and corruption culture which will harm it later like a new, worse version of DEI.
The idea that America had “goodwill” in other countries before Trump is laughable. Where? Latin America? Africa? In the Muslim world? We bombed the hell out of all those places long before Trump. This most recent Iran war has generated less outrage in the Muslim world than the war against Iraq 20 years ago.
American foreign policy since the 1950s, fixated on fighting communism and then terrorism, has meddled with so many foreign countries that it’s silly to talk about “goodwill” towards America. That is not to say goodwill matters. Clearly the U.S. has done great without it.
Although it is worth pointing out that something changed - prior to around 2010 the US had a financially dominant position and the internet was small. So it was feasible to totally ignore opinion in places like Latin America, Africa and the Muslim World.
What we've been seeing in more recent years is that the US can't get away with that so easily. Countries like Iran, China, Russia and India are capable of pushing back both in terms of the raw resources they can bring to bear and also increasingly in the ability to get their propaganda into the US discourse. The US is being manoeuvred into a one-among-equals position in practice and probably in the discourse too which will be a moral shock.
It is not binary
The US is a big part of the customer base of the largest manufacturing economy in the world. China's economy blossomed via US and European consumers.
The Chinese economy probably will further blossom in Southeast Asia, Middle East, Africa, Central and South America why because they seem to be able to build infrastructure in many of the places that they trade with.
The United States Japan and South Korea seem to be failing in that area, if it wasn’t for the war between Russia and the Ukraine, the Chinese would probably be halfway to Europe with their high-speed rail system, which is already in the far west of China today.
Once the war is over between Russia and Ukraine it will be full steam ahead to Europe, whether that’s through the Caucasus, in the north or south or somewhere in the north between Russia and the Ukraine, the Chinese will get there and unfortunately the United States will be standing on the sidelines scratching its head in denial.
There isn't a war today. However China wants Taiwan: war is future option they preparing for - they might or might not go to war but they are clearly preparing. The US is likely to get involved in such a war and I would expect Europe to join in as well.
Don't ask me what Trump is doing though.
China going for Taiwan would be the worst geopolitical move of the century, potentially worse than Germany's decision to invade the soviet union. They talk about reunification because it's good propaganda and both sides want it to a degree, but doing it forcefully just isn't something China would realistically do unless they really don't like their path of becoming an international trade and manufacturing hub
> would be the worst geopolitical move of the century
From a political perspective, perhaps.
> doing it forcefully just isn't something China would realistically do
From a military perspective, taking Taiwan by force would allow China to, "threaten the sea lines of communication and to strengthen its sea-based nuclear deterrent in ways that it is unlikely to otherwise be able to do." Taiwan would give China access to the Philippine Sea. https://gwern.net/doc/technology/2022-green.pdf
> but doing it forcefully just isn't something China would realistically do unless they really don't like their path of becoming an international trade and manufacturing hub
Sounds rational, but this decision is in a small number of hands. And those hands can change quickly. I also thought the US would never threaten to annex territory of a NATO member.
Offer to purchase imperial territory of a NATO member is not the same as a threat to annex it.
But threat to annex is what's happening.
But we threatened to annex it.
Or the Japanese bombing Pearl Harbor which was another dumb move. At this point, the Chinese just need to bide their time by the year 2100 Taiwan will probably be part of China and North and South Korea probably will be reunified. Both are inevitable, and I don’t think it will take any shots went it happens.
> and both sides want it to a degree
Is "it" the propaganda (useful to politicians for achieving political power) or reunification? My sense is that the number of Taiwanese that are enthusiastic about reunification has probably bottomed out in recent decade(s)???
China seems to be recently building up its forces and putting a lot of money into military. I think it would be foolish to just assume its all for show even if it might be in the end.
And quite frankly, its only geopolitically stupid if they lose. Consequences for this sort of thing usually tend to happen if the conflict is long and drawn out. If the win quickly the consequences would likely be minor.
Everything their military has been doing for the past ~20yr or so has been toward capturing and securing Chinese waters and beyond, including Taiwan. It's a negotiating chip for them.
https://www.navalnews.com/naval-news/2025/01/china-suddenly-...
Just look at Iran. Nothing really happened to USA or Israel. Nothing will happen to China if they take Taiwan. Or maybe the "West" will boycott them and crash entirely.
The USA did not take Iran. They essentially shot off fireworks, killed the figure head and then got check mated.
Taking wasn't the goal. I don't understand the goal, but it is clear that taking Iran wasn't.
The U.S. isnt checkmated, the U.S. is enforcing a naval blockade against Iran and their oil based economy is in a free fall with rapid inflation. The U.S. is experiencing slightly higher gas prices but economy is still humming. Meanwhile the U.S. military has not exhausted all of its options while Iran has none.
What would you consider a "win" condition here? I have no idea what the American administration is looking for as a win.
The US is indeed checkmated like Afghanistan like Iraq and like Vietnam checkmated spinning your wheels spending money wasting money wasting time and wasting resources, checkmate.
With the addition of most countries now looking for other trade partners the Art of the no deal…
More like a draw it seems.
Definitely not.
Iran did billions in damage across the middle east, put a major dent in munitions stockpiles, and there is effectively no military way to shut down all of Iran and protect shipping. Too many drones, too many ballistic missiles, and it only takes one. This is basically like an insurgency on a macro level, where small and cheap weapons threaten very large very expensive targets.
Iran is completely blockaded right now:
https://mynews4.com/news/nation-world/centcom-naval-blockade...
The drones are useless if you dont have targeting systems which were taken offline by F35s 2 months ago.
Blocking Iran is going to do more damage to the world than it will do to Iran.
What targeting systems are you talking about? You can use optical targeting with a raspberry PI in the drone itself, pre programmed. Nothing for an F-35 to take out.
The EU is running out of jet fuel. 20-30% of the hydrogen needed for chip fab comes through the straight. Fertilizer for food comes through the straight, and planting season has already begun.
This was a political and economic disaster.
So what Iran's basically fucked and the USA just gets an economic boost from military spending?
> "They essentially shot off fireworks, killed the figure head and then got check mated"
I mean, that's certainly a take. A wholly inaccurate one, but it's a take.
https://en.wikipedia.org/wiki/List_of_Iranian_officials_kill...
they've made everyone very angry at them, and can handle that since everyone relies on them. China, however, is trying to build that trust and forcefully taking Taiwan would have very severe consequences. The reason the fallout from iran isn't as big as the fallout from a war with taiwan is because most of the west at most puts up with iran. Invading taiwan, meanwhile, would cause massive problems with chip production. Think the anger and distrust towards the US due to hormuz but 1,000 times worse
The difference is Iran is a terrorist regime that murders its own citizens and funds violence across the middle east.
Don’t take China on face value, they have every incentive to promote a grifting military industrial complex in the US while focusing on competing in manufacturing. An actual war would fix a lot of the grifting in the US as it would align interests. Pretending they’ll go to war over Taiwan and not doing it is an effective strategy for undermining the US.
I hope you are right, but unfortunately there is no particular reason to trust China's leadership anymore. They are not nearly as obvious at Trump, but they are not on a good path.
I'm explicitly distrusting them, they're saying they want to take Taiwan and I don't believe them. I try to ground my belief in realpolitik, cynicism, and from my experience with strategy games. There is an element of manipulating your opponent into acting the way you want them to by sending them costly false signals, they have to be costly or they won't be believed. I think we (the West) are being played and the gifting elements in our political leadership are more than happy to play along. I'm sure China would like to take Taiwan if it wouldn't cost them anything, but the US is a waning hegemony so for now it is better to wait until the US is beyond fixing itself. At this rate that may not take long.
Otoh, if you send the costly false signal of investing in your military, and your opponent doesn't buy it, you might as well use it since you just spent the money anyway and your opponent can't stop you since they didnt believe your signal.
> I try to ground my belief in realpolitik, cynicism, and from my experience with strategy games
I'm trying to do that too but what the hell is going on with Putin? Why does he continue to engage in this ridiculously expensive war? I don't see any evil genius explanation anymore. It just seems like a mix of sunk-cost-fallacy and save-face.
I think many geopolitical decisions are actually based in irrational emotions of a hand full of people.
I think Putin was and remains a rational actor, I know a lot of how that war is understood in the west is colored by a very effective propaganda campaign that I don’t have the time nor energy to counter.
But I will say, in a very broad stroke, we’re heading for a great power conflict and the US has two primary factions on foreign policy; the primacists vs the restrainers, both want to take on China (contain with war) but the primicits want to topple Iran first and set up Israel as a regional hegemony where the restrainers want to build up locally first. China knows this and Russia is a junior partner / quasi vassal state to China. China lacks modern war fighting experience which the Russian Ukraine war has been very helpful in fixing. Yes it’s very expensive, but so is losing a great powers conflict.
Is your claim that Russia is continuing to fight Ukraine as a favour to china in order that china get information on how modern war is fought and intel on western capabilities?
While it is undoubtedly true that china is learning everything it can from this conflict, and that russia is at least a little subservient to china, they aren't so subservient for this explanation to make sense.
Germany, Japan, Russia, Great Britain, and the United States all within the last 125 years… The headshot was from within mainly self-inflicted.
Unfortunately if you are wrong it is an even worse disaster and so if there is any possibility we are all forced to play their game.
A grifting military industrial complex is unable to defend Taiwan even if it wanted to as evident by the exceedingly poor showing with Iran. The disastrous reality of doing what was done is already with us. If the US didn't take that bait it could have made better choices that would have left it in a stronger position militarily long term, if it made a real attempt at re-shoring civilian manufacturing it could cross subsidise dual use technology, but instead we have corrupt politicians doling out concessions for kickbacks.
> If the US didn't take that bait it could have made better choices that would have left it in a stronger position militarily long term
Like what though? If the problem is that not going to actual war has enabled the MIC to be captured by grifters, then "taking the bait" and going to war should actually help improve that by showing up the grifters and giving us a chance to switch to making stuff that works.
> Pretending they’ll go to war over Taiwan and not doing it is an effective strategy for undermining the US.
The bait is for the buildup that promotes the grifters.
> An actual war would fix a lot of the grifting in the US as it would align interests
We are in agreement. I made these points earlier in this chain.
The Iran war doesn't count as the alignment of interest requires an actual threat of being defeated.
> The Iran war doesn't count as the alignment of interest requires an actual threat of being defeated.
That's starting to sound a bit no-true-scotsman. If we need an existential threat to the US, that's not going to happen - realistically China conquering Taiwan or even building an empire around the Pacific would still not be felt as such a threat.
To align the interests there has to be a substantial negative consequence that would be felt by the grifters if the endeavor fails.
The US is already close to losing world hegemony status and it kinda needs it in order to print money / export inflation. A multipolar world is one where the US is greatly diminished and this will happen with or without losing a war.
> To align the interests there has to be a substantial negative consequence that would be felt by the grifters if the endeavor fails.
Like what though? The failure in Iran has had pretty substantial consequences that are being felt. If that's not good enough, what is? You were talking like you thought there was a realistic path to a better military, but consequences for the US aren't going to come much bigger than this.
You're conflating the grifters with the US in general, the grivers are able to continue grifting even at the expense of the US. This is requiring too much hand-holding from me so I'm done with this conversation.
> A grifting military industrial complex is unable to defend Taiwan even if it wanted to as evident by the exceedingly poor showing with Iran.
These two conflicts would be so different that i don't think it makes sense to draw this conclusion.
All China needs to do is do what they’re doing play the long game the United States is currently shooting itself in the head, if they’re smart, they should just sit back and watch the show by the year 2100 well you know. And coincidentally that also applies to Russia sit back and watch them do it to themselves.
In addition, some of the other countries like Canada, Mexico, Australia, and New Zealand had better get busy from within because they’ll be on their own. In the same applies probably to Europe.
“China” doesn’t care about Taiwan; Xi does. And he does not have until 2100 to wait.
>They shored up their gpu design and manufacturing expertise.
I'm pretty sure they've been exposed for smuggling GPUs into the mainland because they can't ramp up fast enough, only reason we got Deepseek v4 before GTA VI
China is trying to undermine the US economy through open source models. If they can down round or bankrupt the model companies, they take down the US.
Currently the US is extremely vulnerable and dependent on China. AI is an important exception, so it’s key for China to destroy that
The US is undermining its own (and everyone else's) economy just fine, no imaginary assistance from China necessary.
The role of the US was always to purchase cheap Chinese hardware, slap some modestly better software on top of it and the rest of the world happily would pay for that as a whole package. But with the US increasingly becoming isolationist, the rest of the world is starting to wonder why do we need the US as a middleman at all, so the US had to invent a whole new reason for the rest of the world to rely on it: AI.
Of course, the problem with this idea is that while everyone was perfectly happy with the previous arrangement, nobody else in the world gives a shit about AI. It's scary, it takes the coolest things we used to enjoy doing and turns into mush, it destroys our local culture by making us all rely on English, everything bad (like layoffs) gets blamed on AI and so on and so on. And when you combine that with the rest of the stupid foreign policy decisions, many would find joy in witnessing the US economy crumble to the ground. Pointing the blame to China instead of to your own reflection in the mirror is just an easier pill to swallow.
This is spot on. The US under MAGA are actively dismantling their once leading position in IT as well as defence. I guess it is hard to see as a US citizen but from outside this is clear as glass.
> The role of the US was always to purchase cheap Chinese hardware, slap some modestly better software on top of it and the rest of the world happily would pay for that as a whole package
Curious where Intel, AMD, Nvidia, etc are in your "cheap Chinese hardware"?
And by "role", do you mean doing the majority of the R&D behind the modern hardware we all use?
That's just Chinese hardware with extra steps. If you don't believe me, feel free to look up the list of CEOs that are in China right now as a part of the US delegation.
As for the R&D part, Huawei is still pretty much indistinguishable from any other phone. I could buy one right now if I wanted to. It has shittier software though.
I’m not American, sir. But I disagree with your analysis. I think you’re looking back over too short a timescale.
Trivialise it all you want, but the world is vastly different from what it was at the beginning of 2025 and I don't think you or anyone else can deny that in any way.
What happens next remains to be written, but so far this new order seems to be leaning heavily towards China and to a lesser extent the EU. Not because of anything those two have or have not done, but because of what has up-until-that-point been widely considered to be world's number one superpower losing its damn mind. I don't even have to come up with a list of examples to prove my point, we both have pretty much the same list in our minds already.
Instead, I'll just quote the President of the United States from a little over 24h ago:
> I don’t think about Americans’ financial situation. I don’t think about anybody.
AI is just another in a series of slaps to everyone's faces by the US. If it has some legitimate long-term use (which according to me is still an open question, although to many others it is not), thank god the US does not have as significant of a moat as necessary to fully control it, as the crux of it is easily replicable (albeit expensive).
Down round = the destruction of the United States is ridiculous hyperbole.
If Anthropic does a down round, the US economy will crash. Not hyperbole.
The US economy right now is based entirely on the AI bubble. This is an indisputable fact if you examine GDP stats and equities.
That bubble is driven by (rational) over-investment in AI capacity. For that investment to continue, there must be demand for it.
The demand for that infrastructure essentially lies in the hands of a few businesses: principally OpenAI, Anthropic, Google.
The reason I highlight Anthropic is that without their advances in the last six months, the game would already have been up. Only via Opus 4.5 and 4.6 did the possibility of ROI look plausible. We are very much dependent on a handful of companies’ progress to keep this bubble going.
I’m not saying AI is bs, just that this is a bubble like others (for example, Victorian railways) and a down round would signal the end of the bubble.
So for an enemy of America, whether that be China or Russia or any other country, it is logical to target the AI bubble to cause an economic crash and thus restrict America’s ability to compete in terms of spending etc.
China is not even trying to destroy the US bet. It's just making sure everyone else has a reason to buy their hardware.
There's a joke in China that Trump is the best president that China has ever had.
How does gobbledygook like this get traction on HN? What has happened on HN culturally to allow something like this to surface to the top?
https://paulgraham.com/submarine.html
A lot of nonsense get traction on HN (and everywhere else).
The revolt of the masses is real.
Happening quickly too. Far too quickly.
Flagged for AI content: I hope this submission dies and the user is penalized (look at their submission and comment history!), because IMO the article does not belong on the front page. Quick polemic:
>The US is winning the AI race where it matters most: commercialization
If you ask me, one could name different criteria for winning, and commercialization would not be the first thing to come to my mind:
https://english.www.gov.cn/news/202604/15/content_WS69df29e6...
https://fortune.com/2026/05/03/chinese-court-layoffs-workers...
https://www.reuters.com/world/china/china-moves-regulate-dig...
> It also owns platforms that generate and organize the data of the AI age. YouTube is a video corpus. Google Drive and Microsoft 365 sit inside daily office work. GitHub sits inside software development.
Yeah, okay. China does not have any platforms nor data.
Fully AI generated according to Pangram.
Can we have a rule where LLM generated texts require a disclosure or be removed?
Edit: The entire blog seems AI generated. Huh.
Very good call. I shied away from calling it a terrible article but it is
There certainly are better articles on this topic that have come out recently.
Can I block a user to avoid seeing his posts? I noticed front page would be much nicer without guys like OP
There's a userscript that'll hide users, 'sources' (domains), and titles. The GH repo was deleted, so use the GF link instead:
https://news.ycombinator.com/item?id=43282379
There used to be such a thing as profit. A return on investment. If your exit strategy is to get sold to Google, focusing on revenue is a perfectly fine strategy. If you _are_ the Google, however, the money poured in should eventually be made back. We seem to have forgotten that. The current level of commercialization just means the US is burning investments faster, than anyone else. Eventually this might change and the bet might pay off. But every minute this goes on, the expected payoff must be larger to pay for the loss made this minute as well as interest for the previous minutes. I’m not entirely sure this is what “winning” looks like. Tic-toc.
No, the US is _leading_ the AI race, but the race isn't over.
What's the point of leading the race for 90% of it, if they're gonna slip on their own sweat and fall down by the end? In non metaphorical terms, what's the point of spending billions of dollars rushing to get the best AI tech at all costs, when the competition can distil your progress and catch up in 6-12 months while only spending 1% of what you spent.
Even in the aspect the article cares about, commercialization, the US is starting to lose marketshare, I've seen people move from cc/codex plans to use glm/opencode plans due to the recent squeeze the US companies put on plan usage, the US companies are screwed if that sticks, not everyone needs the bleeding edge models, they just want to pay $20/month and have the models be decently capable.
What if it is not winner take all? What if there is no race. What if what USA has been doing is just burning money with possibly unsustainable debt load and way over build valuations...
AI being commodity server capacity might be a thing. And the customers might even manage without hyperscalers... In that sort of end scenario whole current market might look rather foolish.
>What if there is no race. What if what USA has been doing is just burning money with possibly unsustainable debt load and way over build valuations...
You mean, what if the hype-based billionaire-class is wrong? Isn't suggesting that a sin in America these days?
Cannot happen, these days. The US taxpayer will be glad to bail them all out (again).
> No, the US is _leading_ the AI race, but the race isn't over.
When someone says their football team is winning in the first half, do you say, "Umm, no, they're leading, not winning!"
When a cyclist is leading a pack and pushing themselves against the air resistance for half the race, do you expect that cyclist to win, or one of the ones behind that's been taking it easy in the slipstream?
It's a race metaphor not a football metaphor.
I have never ever heard a commentator say something like "Arsenal are currently winning with 2-0 against X". It is always leading: "Arsenal are currently leading with 2 goals against X".
If they got there by tiring themselves out more than the other team, yes.
I find it very strange that the GP felt the need to correct a difference between leading and winning. If you're at the front of the pack in a race, you are both leading the pack and winning the race.
If your team has more points than the other team, you are both leading the contest and winning the contest.
It is a distinction without a difference.
The elephant in the room, and where the analogy breaks down, is that a race has an end, the finish line. A sports match has a victory condition of some type. Nobody has a damn clue as to the victory condition of this hyperscalar craze. Anyone who says otherwise is incorrect.
GP here, leading and winning are different things in the race context/metaphor.
In foot/cycling races there's often a pack leader, that leader is often not the winner of the race, all they're doing is taking the brunt of the air resistance while everyone else slipstreams behind. For a casual observer it seems that the pack leader will win, but everyone knows that it's gonna be someone that paced themselves that's going to overtake the first spot at the tail end of the race.
You’re moving the goalposts and tying to one specific sport. I didn’t say “winner” nor did anyone else. “Winning” is the operative word, and tying the whole analogy to cycling is as close to a strawman as one can get while having the ability to claim otherwise.
The analogy to what "frontier" ai labs are doing is very close.
Yes?!
Well, if they were up by 4 and now it’s 4-3 and the team is under massive pressure, “we’re still winning” is of little condolence to the fans.
Leading the race makes sense if it's a winner takes all market. AI cannot be a winner takes all market, because of national security reasons.
I would also argue that as AI gets better it will also be more fungible. It will be valuable like electricity. Lots of companies make good money producing electricity, but not the kind of money current investors are hoping for.
Mark Cuban in a recent interview answered your question: companies are afraid there is going to be just one in the end—sort of the way there is one ad-company now on the internet. They want to be that one.
Whether they're correct that there can be only one is of course a matter of debate. But that is at least the mind-set they are operating under according to Cuban.
> sort of the way there is one ad-company now on the internet
Which one, Meta[0]?
0. https://www.reuters.com/business/media-telecom/meta-poised-s...
That kind of does make the point.
Why would Mark Cuban know anything about the motivations of today’s big tech companies? He has not been involved in tech businesses since he sold a radio on the internet website 26 years ago.
Those guys are all on the same private group chats.
Why would Mark Cuban be in those group chats?
He was never based in Silicon Valley, and the closest he got was selling a website to Yahoo in 1999. After that, he has mainly sold sports and his media personality for TV shows.
Moreover, why would leaders of trillion dollar big tech companies subject to myriad securities laws be discussing intimate business details with random people that have no domain expertise or influence?
I don't know why. You'll have to ask the chat group administrators. I'm just telling you how things actually work.
https://www.semafor.com/article/04/27/2025/the-group-chats-t...
I do not see the names of people in big tech business leadership positions, except maybe Andressen, if he counts. All the other ones look like media personalities or journalists or some two-bit SV founder.
The winner in the long term will be the one that will deliver the best performance and low-memory ratio for local models.
Anthropic, OpenAI and Mistral are just companies that are making money right now (still not profitable), but will lost their tractions and values in the long term.
However, I am more appealing to see how OpenCode Go subscriptions will go in the future: cheaper than big techs, more tokens, and they don't train on our data to (try to) improve...
Local models will never compete with large SOTA models, in the same way an iPhone doesn't compete with supercomputers doing nuclear simulations.
They paths will differentiate and split. Probably SOTA models will eventually be locked down and only accessible to state actors because of how expensive they will be to run (already started with Mythos).
its a big assumption that larger models bring any measurable benefit in the long term. there's a point where its not worth paying the expense of a bigger model and we dont know where that will be as both, models and hardware improves.
we do know however where evolution is at right now with our brains, but thats probably not comparable - yet the only thing I can see to make any kind of prediction at all
Isn't Mythos mostly hype though?
https://hacks.mozilla.org/wp-content/uploads/2026/05/securit...
Current local models already compete.
A Qwen3.6-35B-A3B or whatever it's full name is, when on a 3090, can at the very least, with very little fine tuning, compete with Haiku and blows away GPT4.1 (aka, the cheap models).
It might keep up with Sonnet 4.5 with some tinkering.
But long story short: it seems to have better performance and similar quality for a payoff of a year or so compared to cloud models. In the same way you can self host faster/easier/cheaper than cloud hosting, if you are okay with the negatives.
I'm returning my 3090 soon for a R9700 after some more basic benchmarking, since the higher RAM should improve my observations more.
> It might keep up with Sonnet 4.5 with some tinkering.
I would love to see that. I've been using Qwen3.6 35B and the dense 27B, and they are both too slow with not such great results for agentic coding tasks. It's ok, but not impressive. I had better luck with the BF16 and Q8 than the Q4 from unsloth (really love what unsloth is doing in this space). Another problem I had with Qwen, which I did not ever encounter with Sonnet - even the BF16 gets stuck and needs a "continue task" prompt from time to time, the lower quants are even worse in that regard.
If you get some interesting results, I would love to read about it!
You don't mention runtime, hardware and harness which are critical. The 35B A3B model should be pretty fast, you do need a decent setup but nothing too fancy. I'm using Q8_XL from unslouth with llama.cpp and opencode and it's pretty awesome. I find that opencode drives the model best, it very rarely gets stuck even with a ton of tool calls. I agree it's comparable to Sonnet 4.5 for most tasks. You may also try the Gamma 4 models which are faster but not as good for coding.
You are missing the point. Parents says the market to win need economical models more than SOTA models. Whoever is running those nuclear simulations is not making as much as Apple.
If we extend this line of thinking, China might be on leading that race.
Not training on data is a con for me not a pro. The reason Claude is so good is RL training from users' chat histories and use cases. The era of pure public data training is over, as everyone has access to this data yet only a few are frontier models.
> Anthropic, OpenAI and Mistral ...
Mistral? I think their "revenues" is something like 1/150th what OpenAI and Anthropic are making.
I feel like the much simpler explanation is that the US is winning because it's dumping the most money into it. By a very large margin.
...printing the most money into it. The circular IOUs amongst the AI and hyperscalars are a form of debt, i.e. money creation. Don't get me wrong, a whole lot of other dollars are going in too, but investing money that doesn't exist is a massive risk always.
Maybe I'm not doing capitalism right, but isn't it supposed to be "The one who profits the most wins"? If you win by just spending, I think you need to adjust the parameters of your capitalistic market.
I don't get it either, but it seems to me a bit similar to how the US, if you look at market value of car companies, has utterly crushed Europe and Japan (with China surging ahead of those and maybe threatening the US soon), which to me sounds crazy (I still think of German cars as the top of the bunch).
According to Google (AI summary, no idea if it's 100% right but from what I've seen elsewhere it seems right):
Top Car Companies by Market Value (May 2026):
- Tesla ($1.3T - $1.56T): Retains market leadership with a valuation often exceeding the next several largest competitors combined.
- Toyota ($259B - $317B): Largest traditional automaker by market cap and unit sales.
- BYD ($122B - $126B): Strong market position as a Chinese electric vehicle leader.
- Xiaomi ($119B - $135B): High valuation following its entry into the smart EV market.
- General Motors ($69B - $75B): Leading traditional U.S. manufacturer, competing with Hyundai and BMW for top 10 spots.
- Ferrari (\(\approx\$60B-\$68B\)): Maintains high value due to luxury branding.
- BMW / Mercedes-Benz / Volkswagen (\(\approx\$58B-\$64B\) each): German luxury and traditional automakers facing high competition.
- Ford (\(\approx\$47B-\$54B\)): Remains a major player with significant US market share.
So, essentially, Tesla alone is somehow worth more than all European companies combined??!
Except that by sales volumes, the top companies are exactly the ones you'd expect: Volkswagen ($350B) and Toyota ($315B) at the top, far ahead of anyone else... Tesla is around the 7th place with just $95B. Does the financial markets still expect them to far out-earn Volkswagen and Toyota any time soon, we've been waiting for like a decade already??
Gemini says that by country, the car companies revenues are:
* Germany - ~ $600B
* Japan - ~ $520B
* USA - ~ $470B
* China - ~ $250B
How does that even make any sense?
Stock market is about expected future returns. Tesla probably won't ever be very good, but it has the chance to regulatory capture the entire US market in a way that Volkswagen doesn't. Tesla gets to market overpriced junkboxes to rich people in a way that Volkswagen doesn't. Tesla has a likelihood of acquiring lottery-ticket companies like xAI in a way that Volkswagen doesn't. This stuff doesn't happen when your company just focuses on making cars.
The stock market is over 60% passive investment, it's starting to get unmoored from the financial realities of the underlying companies. What that means for the future is [shrug emoji].
The price changes are not being driven by the passive investors though. People did actually decide to put all that money into Tesla.
Look at profit rather than revenue - "it's easy to make a lot of revenue when you're selling a dollar for 80 cents" applies just as much to big legacy automakers as it does to startups.
>Does the financial markets still expect them to far out-earn Volkswagen and Toyota any time soon, we've been waiting for like a decade already??
These capital heavy industries operate on 30+ year timelines, a decade isn't sufficient time.
Revenues are not the end all, be all. Profit and profit margin, along with revenue trends provide a more complete picture. And the most significant factor is that the market does not expect Volkswagen or Toyota to do anything new, to do anything with the potential to earn more. They are what they are, and they will continue with their lower margin businesses until they fade away.
Investors are betting that Tesla, however, might have a few tricks up its sleeve, that will allow it to expand markets and profits.
There is a reason in capitalism we have anti-monopoly law or preventing dumping prices because those often leads to monopoly. So yes for sure you can kill your competition by just dumping money and loosing profits.
That also assumes the monopolized market is profitable enough to pay for the dumping, but right now we are still questioning if LLMs have such high value on the market. Yeah its great for programming, but is the majority of the population benefitted enough to all start paying for access lke many investors expect? Someone might have looked at the low cost and massive lifting capacity of hydrogen balloons in the past and seen a lot of potential profit but if investors had dumped money to monopolize the hydrogen balloon market they would have lost their ass.
No, they are not. They are winning because West is forbidden to use Chinese models for anything work-related.
True, many people don't know GLM 5.1 and Kimi 2.6, really on par with frontier models. There's also Minimax 2.7, DeepSeek 4, Qwen, Xiaomi 2.5 Pro, etc.
China is leading in open source frontier models, so I don't really see how the US wins this one. At some point, companies and people will start running their own models in the cloud and locally, Chinese models will be everywhere.
Nah, I model hop constantly as I work with serving GLM and Kimi models and they're not nearly as good as Opus 4.5+ and GPT 5.2+ and it's not particularly close. They're good by standards set a generation or two ago, but they're really not competitive with where the frontier models are at now.
They compete with "mini" or "nano" model classes quite well given the price of inference. You'd need to "model hop" anyway, using Opus for everything is quite wasteful.
Now those aren't really "frontier models" now, are they.
They are on the frontier of local models, where the game is often to get the best bang for the buck. You can always scale model size and compute (Mythos, GPT Pro, Gemini DeepThink) to reach better outcomes, but that's not a very interesting strategy.
> They are on the frontier of local models
That's not what anyone means when they say frontier models, don't change the definition. It's almost as bad as open weight being subsumed by open source when it comes to local models.
Guess it really depends on what you use them for. I've been able to built whole apps with them, not slop. Kimi is quite good at design, for 3D, I noticed Gemini 3.1 is excellent for basic to medium use cases.
I've tried both Opus and GPT 5.4, they also hallucinate just like the rest at a much higher cost.
The more you use a model overtime, the better you become with it. It's really hard to measure, my main metric lately has been tokens per second/time to complete task.
At this point I've the feeling frontier models are optimizing for benchmarks and one shot prompts.
If you actually use them you'll see that they are far from frontier models. They are much more cost-effective for what they are, but frontier they are not.
My understanding is that it's not that the _models_ are banned, but rather the _platform_ is banned. It is acceptable to host, say, `deepseek-r1-distill-qwen-7b` and run it yourself, for example. It is not acceptable (to the authors of these bans) to download the DeepSeek app and run it on your work device.
I just left a job for a German B2B software company which sold primarily to large automotive, defense, and aerospace companies. Several of our customers specifically banned anything with the word "DeepSeek" -- hosted or self-hosted.
There's still a lot of naivety on what the difference is between models and platforms, and its easier for a lot of these big companies to just make a blanket statement like "nothing DeepSeek" than for their procurement teams to try to understand and negotiate with each vendor. They don't see the potential benefit over the potential risk of somebody misinterpreting or getting it wrong, so they outright ban it.
Most people that approve or buy software simply also just don't understand how models are being trained or if it's possible/how far a model could go to "introduce backdoors." A backdoor could be, from a business perspective, a model which has been trained to give answers that could hurt western business in a "strict text mode" or produces payloads in a programmatic mode that are intentionally trained to introduce software vulnerabilities.
Anyone can make arguments against these for a variety of reasons (looking at the transparency of both sides and comparing, etc) but for many reasons today and for better or worse, many Chinese models are being banned on big software contracts, which gets back to the title of the article
Thing is these models can also be a propaganda machine whether you run it locally or not. This is true no matter the origins. Chinese LLMs will never shit-talk CCP, and it will always give a rosy depiction of the Chinese government. It's perfectly understandable if companies don't want things like that. US/EU models have these problems too, but at least there are some ways to fight that: with a lawsuit or a megaphone on social networks. With Chinese models there is nothing you can do.
You are sending all your prompts code and files there. So ofcourse its an issue
Where's "there" on a self-hosted setup?
We aren't allowed to use any unauthorized models even locally.
> They are winning because West is forbidden to use Chinese models for anything work-related.
Because the models hosted in China are not trusted. This is 100% a part of what makes up commercialization.
Is anyone outside the US trusting anything hosted in today's US? If so, why?
I would say that both US and China are using the data we trust upon them for industrial espionage. So don't use their models if you are working defence or other sensitive areas
Deepseek is a fraction of the cost of western LLM and still just as good. I say it's also related.
Do we have any solid evidence these models can outperform Western models in terms of quality? Or is it more: because they are forbidden, they can't get enough training data, visibility etc. to compete?
Scroll down to the leaderboard - https://arcprize.org/leaderboard
Spoiler alert - they are all towards the bottom of the leaderboard. People come up with a wide variety of excuses for why they are not used despite being offered for significantly lower cost, but the answer is simply because they don't perform well enough for now.
There isn't even deepseek V4.
I'd rather trust LLM arena leaderboard, which puts it on par with sonnet.
LM Arena uses human side by side voting, which limits its applicability to complex tasks.
The ARCPrize leaderboard does have Deepseek V3.2, which only scored 4% on ARC-AGI 2 (while the top models score over 80%). It also Kimi and Qwen, but they also didn't perform well.
You’re saying if we were allowed to use e.g. qwen more broadly the US wouldn’t be in the same strategic position? We have the best models…we own all the companies that make the best infra and the hyper scalers…I don’t think “oh we can use Qwen now?” Would exactly devastate the US
> I don’t think “oh we can use Qwen now?” Would exactly devastate the US
You'd be surprised how useful it can be to fine tune it in enterprise.
Well definitely but we have plenty of sanctioned OSS options for that
Qwen's open models are quite small compared to Kimi, GLM and DeepSeek Pro, which are often described as near-SOTA.
Why? So that even more American IP can pass through Chinese servers? Or because their near frontier models are heavily government subsidized?
>No, they are not. They are winning
You agree they are winning though, right? China is known for not playing fair, stealing industrial secrets, etc... that reputation matters and it's a good reason why the US is winning. Is the US perfect? No. Does the US play fair? No. Spare me the whataboutism in the comments. The bottom line is most people think the US is a safer bet and that's why we're winning. I personally wouldn't trust either government, but if I had to choose, I feel like I at least have a chance at secrecy and due process with the US. Obviously that is being eroded day by day, but you literally have no due process in China.
Is the US actually winning the commercialization war? The US is definitely delivering more commercial products but if all of those products are deeply unprofitable and need to buy users with unrealistic discounts (or direct cash payments[1]) to keep their DAU's looking good then is that winning?
There's a significant amount of innovation happening, but if the market decides this AI thing is not worth funding then I think that'll dry up overnight.
1. https://thenextweb.com/news/anthropic-private-equity-venture...
One interesting thing that Anthropic did was putting their stack on the various cloud providers, I wonder if they'll put it on GCP and Azure next since they've put it into AWS first at a level we have not seen a major AI provider do to date. Your company can have their own Claude stack just like an ELK stack on your cloud, if they can do this for both Azure and GCP then OpenAI has to really catch up.
In my eyes I would rather use the AI I can run on my own paid infrastructure, so if there's an outage its isolated, or I could potentially have a different region / DC to fallback on.
I'm still surprised that neither Microsoft nor Amazon have made their own models available on their cloud offerings. I guess Microsoft probably does have Phi on there, but it's not front and center, especially with something like Copilot for Devs (seriously Microsoft rebrand that damn thing to be clear what you mean by Copilot!) where they could use the cheaper compute by using something like Phi.
Claude is available on Azure and GCP already.
https://azure.microsoft.com/en-us/blog/introducing-anthropic...
https://docs.cloud.google.com/gemini-enterprise-agent-platfo...
Claude has been available on AWS Bedrock for a long time too.
The new "Claude Platform" announcement was about an Anthropic operated version on AWS (as opposed to self-operated on Bedrock). See the differences here: https://platform.claude.com/docs/en/build-with-claude/claude...
> In my eyes I would rather use the AI I can run on my own paid infrastructure,
Claude has been available like that for quite a while.
One of the reasons for the OpenAI divorce from MS was so they could become available on AWS where they see significant demand, and being available only on Azure was holding them back.
The recent deal with SpaceX AI to use their severely underutilized GPU compute is pretty telling to me. Being able to roll out compute is a hardware problem, rolling out good models needs more than compute, it needs good AI engineers. SpaceX, Amazon et al can do hardware very well. AI engineering, maybe not so much.
Claude is already on Vertex - https://docs.cloud.google.com/vertex-ai/generative-ai/docs/p...
Is Vertex hosting it themselves on their own hardware? Because GCP, Azure and AWS all host it on their own hardware, but AWS lets you do everything that Claude lets you do if you use their APIs directly, and then some. This is what I want to see on Azure and GCP, and heck, maybe even DigitalOcean, if they ever stop expanding into so many spaces and focus on improving their current infrastructure, before I fully migrate off of them.
> Is Vertex hosting it themselves on their own hardware?
Yes, you can even choose regions, for EU they serve it from Belgium. With all the encrypted at rest stuff and other guarantees that vertex provides.
> Important: Accessing Claude models through Vertex AI meets the FedRAMP High requirements, and operates within the Google Cloud FedRAMP High authorization boundary.
yes, and Claude is available on Google AntiGravity with some paid accounts.
This makes no sense when you zoom-out. None of these companies, be it Anthropic, OpenAI, xAI, Google, Meta, Microsoft, are profitable in the AI department, they're all bleeding money and using funds their parent company and/or investors, primarily investors gave them. The Chinese models are keeping up with them, while offering the models for free and able to run on consumer grade hardware, and more importantly they train them for cheap. AI models are an extremely volatile product that can be outdated in the matter of a few weeks. Meaning you have to keep dumping resources into developing better models which has no end-goal besides infinite scaling. Lets look at how users behave in the real world:"I don't use Gemini because it's worse than Claude at XYZ." That's it. Now Gemini has a worse model and people are going to Anthropic... what happens when Anthropics model is arguably worse than everyone else's? What does it matter if they can commercialize if their product is objectively worse?
I understand that America dominates in distribution, integration, enterprise contracts, ecosystems, infra... The article isn't wrong, it's just that that dominance is fragile and requires constant upgrading.
But what is the point of that if you have to infinitely scale because the opposition is right behind you at all times ready to usurp you... You CANNOT scale infinitely, the VC money will run out at some point and then everyone will have to downscale everything to meet the real costs associated with SOTA models, they'll have to be able to use subscriptions, and other monetization to cover those insane costs, we just saw SORA shut down because it was bleeding money far too fast while the Chinese released video models that far surpassed it back to back to back...
EDIT: Hell, one of the most critical aspects is integration of the models into other products, and even on this end open-source is keeping up (and will eventually outpace when the VC money dries out) with these big companies.
> None of these companies, be it Anthropic, OpenAI, xAI, Google, Meta, Microsoft, are profitable in the AI department,
Citation needed.
All reporting is that they are profitable on the inference side and all the VC money is going to building more data centers to run more inference. (Note that the coding subscription models are probably only break even on average - the money is in the API)
> The Chinese models are keeping up with them, while offering the models for free and able to run on consumer grade hardware, and more importantly they train them for cheap.
No one is running DeepSeek v4 (a 1.6T token model) on consumer hardware.
They aren't much cheaper to train the US models. Training is subsidized by the big Chinese tech companies. They are slightly cheaper because they are smaller (and weaker) models than the 5T and 10T models the US frontier labs are training, and the US labs are paying for a more diverse set of RL data (which shows up in diverse benchmark performance).
> we just saw SORA shut down because it was bleeding money far too fast while the Chinese released video models that far surpassed it back to back to back...
Ironically this proves the point.
OpenAI didn't shutdown Sora, just the subscription version and weird social network thing. You can still access it via API.
The Chinese models are API models and probably just as profitable for them as the LLMs are for the US frontier labs.
[1] has prices for video models. There is a big range, but Google's Veo model and OpenAI's Sora are around the same price as the Chinese models.
[1] https://openrouter.ai/models?output_modalities=video
What does profitable on inference mean? As far as I can tell, none of these companies have rigidly defined it, let alone it being a GAAP number. And yeah, if you subtract out all your R&D, payroll, sales, marketing, and other overhead, and get someone else to take on the debt or dig into their free cash flow to build the hugely expensive infrastructure on which you depend, it'd be pretty hard to not be "profitable". It's almost humorous how dumb of a metric "profitable on inference" is.
Ask yourself if AI was so profitable, why don't any of the big hyperscalers break out AI revenue in their earnings. OpenAI and Anthropic both project huge losses for the next couple years, it's not hard to find.
The real problem is, as the GP comment pointed out, that they can never stop training. As long as they're committed to building these behemoth models, the second they stop training, someone else will catch up and everybody will switch over because it's trivial to do so.
> OpenAI and Anthropic both project huge losses for the next couple years, it's not hard to find.
No. Anthropic at least expects to be profitable this year:
> Anthropic expects its gross profit margin, which measures how much revenue it makes compared to the cost of producing that revenue—largely from running servers—to swing from negative 94% last year to as much as 50% this year and 77% in 2028.
https://archive.is/GdLGD
> And yeah, if you subtract out all your R&D, payroll, sales, marketing, and other overhead, and get someone else to take on the debt or dig into their free cash flow to build the hugely expensive infrastructure on which you depend, it'd be pretty hard to not be "profitable".
I think excluding capital expense on infrastructure isn't unreasonable and is done in most industries.
It's worth noting that AI infrastructure has turned out to be an unbelievably good investment. Inference on a 4 year old H100 chip costs more now than it did brand new! That makes the hyperscaler's depreciation schedules look very (and unexpectedly!) conservative (!!)
If the Chinese models couldn't Distill from the larger models they'd be at gpt2 or 3 levels
Even if that is true, it doesn't change the reality that they can compete. Also, if we start going that route, American models wouldn't have any quality data to train on if they respected copyright themselves. Their whole product was built on the work of others, on our work, our art.... without compensation, without acknowledgement.
Literally not a single one of these AI companies, regardless of where they are in the world has any right to complain about someone copying their work.
To quote Elon Musk in court:
> OpenAI’s counsel asked Musk whether xAI has ever “distilled” technology from OpenAI.
> Musk: “Generally AI companies distill other AI companies.”
> “Is that a yes?” Savitt asked.
> Musk: “Partly.”
From https://www.interconnects.ai/p/the-distillation-panic which is worth reading in full.
Strange. I'm switching from Codex and Claude to Pi with Qwen3.6 27B local and Deepseek V4 Flash which is dirt cheap but powerful.
I've built five apps in the last month using DeepSeek, spending less than $1 in total. I am totally in love with DeepSeek and my wife knows it :)
What sorts of apps are those? I tried testing various models with a test app as a benchmark, a local first app with CRDTs, and many, even frontier models, struggle heavily.
https://mediconsulta.net an AI medical assistant
https://decaboy.fit for tracking progress at they gym
https://megaparley.com sports betting platform
A horse betting platform not published yet, still looking for an API odds provider
A car mechanic AI assistant not published yet
I've learned that the more detailed the initial prompt the better result I get. I can share any prompt if you want
and you're not alone (I run Qwen 3.6 35B-A3B at home too).
But just for the sake of discussion, let me ask: Who is the service provider you're using to run Deepseek V4? Do you have any way of knowing whether that compute is happening in the US or abroad?
Article title: “The US is winning the AI Race”
Article content: “The US are capitalizing on AI the best”
A lot of assumptions there that no one can actually verify as true right now. If commercialization into rent-seeking SaaS landscapes is the endgame, then yeah, the US is winning the AI race. If individualization, local LLMs, and consumer hardware are the endgame, China is winning the AI race. If it’s something entirely different - if LLMs are the wall and research is what grants the next breakthrough, or if compute and memory requirements take a dive, or whatever; then we have no idea who’s winning the race because that stuff is mostly happening behind closed doors.
That seems like a lot of rationalization to me. China is pursuing these because they cannot compete on the frontier. Yes, there is a possibility that all that compute is not needed, but it is a rather remote possibility, and there is no doubt that, given the choice, China would be pursuing frontier model building with closed, propietary-only offerings.
All that compute is not needed. We have an existence proof from biology in the form of natural intelligence that much greater efficiency is possible. However, achieving dramatic improvements in compute efficiency will depend on unpredictable scientific breakthroughs. Personally I suspect that an entirely new hardware architecture will be needed, although I don't have any hard evidence to back that up.
>We have an existence proof from biology in the form of natural intelligence that much greater efficiency is possible.
It's only a proof that it's possible with 18+ years of training.
In certain ways my dog has more generalized intelligence than any LLM, and I trained her in only a few months with a modest investment in dog treats.
>from biology ... much greater efficiency is possible
Those are much more specialized models with pretty mediocre tokens per second.
Perhaps tokens is a dead end?
Perhaps! But perhaps whatever human brains use instead of tokens is not as amenable to scaling or copying.
I dunno, DeepSeek v4 Pro is rather on par as far as I can tell, maybe not with 5.5 Pro in all areas quite yet, but close.
I think China is thinking more about the application layer on top of models as going to matter more than the models themselves, so they don't need to gatekeep the models as much.
China is competing in value AI because they cannot work at the frontier, but how is this bad at all? It’s like how the USA has the best drones but they are a few million dollars apiece while China has DJI.
If China could work at the frontier, I don’t know, I kind of think they would still be dumping a lot of resources into exploring the value side since they have that culture already in place.
I did not imply it was bad. I implied that competing in value AI is the only option that China-based AI companies have due to limitations in compute.
This is true, but I don't think they would all be rushing to frontier if that option was available. Chinese are used to working with constraints to their benefit, they would see the price of working at frontier and make hard choices that maybe we can ignore in the states.
Forgive me if this is a naive assumption, but wouldn’t large language models be fundamentally different for a language that is largely symbols? Again, my understanding of Mandarin is limited if it exists at all.
All tokens are symbols. All of the frontier models speak Mandarin.
This is why misspellings and homophones are tells of human righting. LLMs strongly prefer word-level tokens, and word substitutions follow semantic similarity and not the more human auditory similarity.
Funny, I’ve been cracking[0] at this exact problem with a purpose-built model[1]:
0: https://huggingface.co/posts/omarkamali/593639295164067
1: https://omneitylabs.com/models/sawtone
Claude the other day wrote code where one of the bytes in the array was 0xO5.
That's zero ex oh (the letter) five
> righting.
> LLMs strongly prefer word-level tokens, and word substitutions follow semantic similarity and not the more human auditory similarity.
Is this an elaborate joke or your full-word misspelling of writing is both agreeing with your statement (word substitutions) and contradicting it (not semantic but only pronunciation similarity)
I don't see the contradiction, unless you believe that the grandparent comment was written by an LLM.
"飞机" and "airplane" aren't fundamentally different in terms of how they're represented to a computer. Especially for an LLM, where tokenization likely turns each of those into a single token.
> China is pursuing these because they cannot compete on the frontier.
? Claude, ChatGPT, etc are heinously expensive for tiny benefits lmao. Local + efficient is clearly the future
> ? Claude, ChatGPT, etc are heinously expensive for tiny benefits lmao
Unfortunately local inference is inefficient, 100s of times more inefficient than cloud. When you answer one request at a time you still have to fetch all active weights into compute units, once every token. When you run a batch of 300, you load it once and compute 300 at a time.
Compared to cloud, local inference is less flexible. You can't scale up 5x or 20x, can't have spikes, and pay for it no matter if you use it or not. But usage factor is very low, like 5%. And to run a decent model your system costs $2000 or more.
AI boosters cling to this notion because it's the only way the massive data center buildouts make any sense at all. I guess you could say the US is winning the frontier AI race. Okay. I'm never going to grant a cloud service access to all the contents of my hard drive, that's just never going to happen, so if you expect me and a lot of people like me who feel similarly to get on this train, you better have a local, lightweight model too or we're not even having a discussion, the answer is just no.
The thing is, frontier model providers don’t take your feelings into account even a little bit. It’s totally irrelevant to the discussion about the service they can provide, because that service is predicated on access to high power GPU slices that local models can’t touch. Those providers won’t be in an existential crisis because some people choose the privacy route, it’s a cost of doing business.
Right but that service being sold is predicated on products being sold to users, yes? Or are we still pretending that the hyperscalers can just pass the same $20 billion between themselves and that's going to be a growth industry forever?
I suppose its possible that all the value to pay back the datacenter construction can be squeezed out of enterprise contracts where your employer can assent on the privacy questions, probably with some kind of complicated contract and insurance regime regulating things.
Even if so, if China is coming behind 6 months later selling laptops with hyper-efficient local models that are 80% as good as "frontier" ones, I imagine they'll get the consumer business AND a fair share of the enterprise business as IT managers look at their options during the next refresh cycle.
Given economies of scale, I think it's ultimately inevitable that the enterprise more-or-less follows the consumer on this, and the consumer is going to prefer local models. There's no ongoing cost after the initial purchase, and your data at least nominally stays within your control.
I'm inclined to agree. The business world itself, and frankly you could make this argument about the entire AI industry as it were, runs on "fine" and 80% capable can probably get you there. And it's arguably even better for the hyperscalers since then, ongoing costs of AI users are basically nil. You can still have your massive datacenters and just keep them for tasks complex enough that they're actually worth spooling up.
Like I don't need an H100 or a dozen to summarize a PDF. And that's most of what I use AI for.
If we are betting on which is an easier sale, $20-100 a month w/tech support included vs $5k-10k and a requirement for moderate technical ability, I would invest in the former not the latter being the proposition that drives the conversation about AI use.
> ? Claude, ChatGPT, etc are heinously expensive for tiny benefits lmao. Local + efficient is clearly the future
Corporate America is where the money is, and corporate America will dictate what products are successful by virtue of spend. Individuals aren't going to be paying $100s or $1000s/month en masse for these models but businesses will be. Being local and efficient isn't that important at this stage but even so as American companies continue to scale and invest they'll be able to make those models more local and efficient if the market wants it. Sort of like how you had a big, giant desktop computer and now you've got a super computer in your phone which is in your pocket. Going straight to "local and efficient" means going straight to being behind because at some point, perhaps now even, the local and efficient model won't be able to keep up.
For some reason people think that they somehow know something that Google or Nvidia or whoever, with hundreds of billions of dollars of real money at stake don't already know and it's both amusing and bizarre to see this play out again and again in off-hand comments like "lol tiny benefits".
You buy an iPhone even though the cheap-o Wal-Mart Android phone for $100 "does the same thing". Except that in this case the Android phone just puts you out of business while those spending big money for "tiny benefits" beat you in the market.
Corporate america is the past. Momentum is carrying capital out of the country. Pay attention to rate of change.
What source are you using for your claim that capital is flowing out of the country? I'm curious to read more about it.
I don't think that's a particularly bold claim after thirty straight years of moving supply chains overseas. Capital is, inherently, the means of production. The world where we could compete is gone.
Capital is not the means of production. Capital is capital. If I have a few million dollars in my bank account I don’t all of a sudden have a factory. Remember from economics class you need capital, labor, and the means of production.
Capital inflows are different from manufacturing outflows. The US has historically imported capital which is part of why we have such a large trade imbalance. I’d encourage you to do some more digging here.
> The world where we could compete is gone.
Sigh no that’s just not true at all. We compete hard and fast all day everyday, economy is growing and will continue to do so, and no amount of leftist doomer, Chinese, Iranian, or Russian propaganda changes those facts.
> I have a few million dollars in my bank account I don’t all of a sudden have a factory.
No but money only has value because of a product of the human labor and production capacity it refers to. Money is not capital, it is a reference to/legal coercion of capital
> We compete hard and fast all day everyday
Sir have you ever been to the us? Lmao. We are only competitive in the industry of white collar work (financial/artisanal services), an industry that capital is actively gutting
> No but money only has value because of a product of the human labor and production capacity it refers to. Money is not capital, it is a reference to/legal coercion of capital
These are just strings of words without meaning or importance.
> Sir have you ever been to the us? Lmao. We are only competitive in the industry of white collar work (financial/artisanal services), an industry that capital is actively gutting
Yes, I live here. Why are you posting obviously untrue and asinine statements like this? Go look at the Fortune 500. There ya go. What other evidence do you need? And not only are you writing dumb things here, your original post was wrong too! Please get off of social media or whatever doomscrolling news you are partaking in because it is bad for your health and perception of reality. The United States by any measure, as a matter of indisputable fact, a highly competitive and dynamic economy across pretty much all sectors. This is not up for debate.
> You buy an iPhone even though the cheap-o Wal-Mart Android phone for $100 "does the same thing".
People buy iPhones because of status signalling and network effects, neither of which appears to apply to AI model choice. LLMs are already rapidly on the way to being interchangeable commodities.
No they don't, it's not 2008. Anybody off the street can get an iPhone or a free iPhone with a mobile plan. They're commodity products. Even homeless people have them.
To the extent LLMs are commodity products you're right (so far), but that is limited to the main model providers, such as ChatGPT, Claude, Gemini, &c. with interoperability on cloud platform providers and other technology providers like an Apple offering you a choice of LLM with Siri or something.
If you want to suggest that some other model is in the same bucket as those primary 3, it goes back to the crappy, cheap phone analogy which is accurate. Yea you can make calls with it, but you make calls better with an iPhone.
> free iPhone with a mobile plan
I get your point but in what sense is that "free"? What mobile plan giving you an iphone doesn't come with explicit debt?
Here's an example from 2025 with a major US carrier - Verizon: https://tech.yahoo.com/phones/deals/articles/want-free-iphon...
They run various schemes like this all the time, you can also trade in your existing phone a lot of times for pretty favorable terms. I've traded in phones that were a few years old and gotten $1000+ for them, especially when switching providers.
That's not free except at point of sale.
It’s free enough that my point stands and I’m close-minded to any further discussion about it.
Verizon's "free" iPhone deal is you pay for the phone up front and then receive a bill credit. Here's the fine print from one of those deals:
$729.99 purchase on device payment or at retail price required. New line req'd. Unlimited Welcome, Unlimited Plus or Unlimited Ultimate plans required. Less $730 promo credit applied to account over 36 mos; promo credit ends if eligibility requirements are no longer met; 0% APR.Taxes & fees may apply. Credits will appear on your Verizon Wireless bill.
https://www.verizon.com/shop/online/free-5g-phones/
I don’t know or care much about the specific details but the article was written in 2025. Carriers run deals and give away iPhones or close enough to free or cheap that quibbling about the details is irrelevant.
If you think the iPhone is a status symbol you’re just wrong.
I'm not the one arguing iPhones are only status symbols. If anything, if I only had the money to spend on a single computing device there's a good chance I'd go for an iPhone due to excellent durability, typically long support timelines, lots of extremely cheap accessories available, high chance of low cost serviceability compared to other devices. There's also a pretty good used marketplace for such devices so picking up one used on the cheap and still getting a few years of use out of it is likely. I'd likely try and stretch to get that device instead of settling for a cheap $100 phone that will be a total piece of junk and end up being my only actual computing device.
I'm just pointing out the statement:
> What mobile plan giving you an iphone doesn't come with explicit debt?
isn't invalidated by some Yahoo article pushing a marketing promo that when you actually do the math and read the fine print its not really a "free" phone, its always some form of debt or bill credit or something along those lines that makes the phone "free". You're still paying for the phone in the end if you read the fine print. In the end one commits to spending several hundred dollars over 36 months or whatever or you pay up front and they give you bill credits if you keep the plan.
I am arguing they’re not status symbols and using how cheaply available they are as evidence that they’re not. Anyone can get one, some companies run free promotions, some do delayed interest programs, some amortize the price over a 2-year time period. Who cares? The details here weren’t important. Apparently Verizon ran some promo in the past and may again in the future giving away iPhones. Why be so argumentative over something so stupid? Not only are you actually wrong here, you’re arguing over the irrelevant details.
> Who cares?
People who prefer truth in advertising.
> Why be so argumentative over something so stupid?
I don't want people to believe untrue marketing statements and make poor financial decisions without actually bothering to read the fine print.
> some companies run free promotions
This just isn't true. They're not really "free". They come with lots of financial commitments.
> Apparently Verizon ran some promo in the past and may again in the future giving away iPhones
They still say they do on their website. If you're getting one "free" iPhone it comes with a commitment to spend at least $65/mo for 36 months. A commitment to spend $2,340 is a lot different from $0.
These are far from "free" phones. Can I go into a Verizon store, not give them a dime or sign any contracts and walk out with a phone free and clear to do whatever I want? No? Sounds like it's not really free then!
My point is if you're poor/homeless you're probably not looking to sign a 3-year commitment to spend a few grand to get a "free" phone. A lot of those people won't even pass the credit check to qualify to even sign up for one of these post-paid plans required to get the "free" phone. If you're really broke you would probably be looking at signing up for a lifeline plan and get yourself a cheap used iPhone instead of signing up for a $2,340 contract.
I already made and proved my point. The iPhone is not a status symbol, and major carriers can and do give them away in various schemes and did so in the past and will continue to do so in the future.
You’re anchoring yourself to one payment scheme and ignoring others and it’s besides the point which is that iPhones aren’t status symbols even if these schemes didn’t exist and iPhones weren’t extremely cheap or freely available.
I don’t have anything left to say here besides that I proved my point unequivocally.
> The iPhone is not a status symbol
I already said I largely agreed with this.
> major carriers can and do give them away in various schemes and did so in the past and will continue to do so in the future
They only do if you're financially illiterate.
> You’re anchoring yourself to one payment scheme and ignoring others
I'm being honest and taking about the real deal instead of blindly repeating marketing bullshit and lies.
> freely available
A commitment to spend thousands of dollars isn't the same as freely available.
The bank gave me this free house all I have to do is pay this mortgage for thirty years. But hey the house was free!
Once again, was the deal that you could walk into the store, grab a new iPhone, and walk out without signing a contract or other form of commitment? If not, it's not really free. It's bad financial advice for people struggling financially to get one of these "free" phones, they're often more expensive than buying outright and getting a much cheaper (or potentially even subsidized!) plan. Especially if you're just needing one or two lines. Many of these postpaid plans only really make financial sense once you're at like 4+ lines on it.
I'm reminded of seeing all those cell phones in the RadioShack mailer ads back in the day. Only 99¢! Dad, can't I get one? It's only a dollar!.
If you spent hundreds of dollars on box seats to a sporting event and they had a complimentary buffet, is that food really free or did it cost you hundreds of dollars? Would you tell someone struggling with money they could get free food, they just need to go spend hundreds on sports tickets first?
Maybe one shouldn't be so willingly close-minded to the truth.
Well China is consistently 6 months behind the frontier labs(possibly because they can they harvest data from released frontier models). If the scaling continues, US will win, but if not then China will win as the models will converge.
The non-release of Mythos tell you the future of that, so long as they can keep the weights from being exfiltrated. Once models become true national security threats, they won't be released in their full form. The hitch-a-ride approach becomes less capable of keeping up.
How would they prevent distillation? That would seem pretty tough to block for any LLM available for commercial use.
This post claims that Opus 4.7 has introduced some detrimental changes to stymie distillation:
https://old.reddit.com/r/Anthropic/comments/1snorbg/the_bigg...
I don't know enough about distillation to understand how much this hinders/slows the process, but it sounds at least superficially plausible.
By only providing degraded models to use commercially outside national defense applications would be my guess. As soon as models are a threat in terms of enabling biowarfare etc, then they just are not going to be generally released.
Honestly, I think its quite possible that models will be retrained with gaps in their knowledge. e.g. a coding model for commercial use probably doesn't need to have deep knowledge of biology, and training on biological sciences probably doesn't help those evals much.
Honestly, I'd welcome such an approach.
What a hilariously uninformed comment. LLMs are not the limiting factor in biowarfare.
LLMs absolutely have the potential to increase the risk of small nation biowarfare and bioterrorism more generally. That you don't believe so is dangerously naive.
We were talking about winning commercially, not on model quality.
"AI in the datacenter" and "AI on local consumer hardware" will eventually be two separate niches with entirely different capabilities, at least if scaling laws continue unchanged and there's no near-term inherent limit to AI smarts. The real point of the datacenter is to be able to do datacenter-scale things. But you don't need that kind of vast compute to run even the largest open models today: on prem hardware can do it easily especially if you're OK with a somewhat delayed response.
even without any of that anyone you ask who's used AI to any professional degree will agree US is winning AI race right now. Future, who knows
It’s certainly too early to call (if you must view this as some sort of adversarial competition). The US is behind on local models, the future for anyone who cares about privacy. There may be step change innovation yet to come that completely shifts the landscape. There’s basically no switching costs to users to change models. They have no lock-in.
It is very much adversarial, and to view it anything but adversarial is to not see the geopolitical reality and the potential national security implications of AI for what they actually are. Moreover, to claim that China is in the lead with local models presupposes that openai and anthropic could not release local models that are better, which is a big assumption. They do not release such models because they have frontier-grade propietary models that have high value.
As someone who happens to have been born in the US and currently lives here, I welcome China winning. I trust them infinitely more than I trust my own government and industry.
OpenAI and Anthropic are beholden to the capitalist system they exist under and hence cannot compete on local models. Like you say, they must try to maximize shareholder value. China is unencumbered by that constraint.
What is it about China that makes you trust them more?
They're far less imperialistic and I view them as better global citizens than the US. I think they've cultivated a much richer culture than the US as well.
You should read about the difference between a land empire and a sea empire.
Happy to if you have any pointers. My original point is that the US kills millions of people outside of its borders, something China most definitely does not. The number is over 12 million post-WWII: https://www.worldfuturefund.org/Reports/Imperialism/usmurder...
Replace the U.S. with another world power. Would it be better or worse? Its always easy to criticize the power that is in place and treat them as if they are to blame for all the ills of the world. It is easy to criticize when you are the nation that is asked to help provide international security for commerce on the high seas. But exchange another nation in the U.S.'s place, and I guarantee you most would make similar decisions, and many (e.g. Soviet Union) would be, hands down, worse and more brutal.
No one asked the US to kill millions of Koreans, Vietnamese, Iraqis, Afghans… well maybe Israel for those last two, but they’re a terrorist entity.
Well if you discount the people China has killed and is killing within its borders, the number is definitely smaller. As well as another land empire called Russia.
Any books or articles you recommend?
not living there, for one. I don't care if they know where I live since realistically they can't do much of anything to me. If I were in china, I probably wouldn't trust them as much as I trust the US. If I were in Switzerland, I wouldn't trust the swiss government and might get my services from america or china.
You are free to hate capitalism (even if you benefit from it enormously). You are free to say that you hate capitalism and the U.S. as openly and as often as you like, without facing imprisonment or worse.
But if you were in China, could you say you hate the Chinese Communist Party and China openly and as often as you like without imprisonment or worse?
We know the answer to that. So go ahead and trust China more than the U.S., but I think that is pure foolishness.
Actually, Chinas free speech, while abysmal, is better than you're making it out to be. They only really care if they see you as a threat, which realistically isn't too far off from the US both currently and historically
I understand that there is some freedom of speech in China, but I don't think the implication is that US and China are functionally equivalent. In the U.S. you can criticize the president all day every day and have millions of followers. Could you do that in China? The more people listen to you, the more dissident opinions become a threat.
I'm not suggesting they are the same, rather, that if the US saw you as a real threat you'd be shut up pretty quickly. The threshold for what is considered a big enough threat is different, but they both have the ability to shut you up and have in the past. Any government would act the same, at least a competent one. The governments 0th priority is ensuring its continued existence, since all of the other functions rely on its existence
Can Comey?
I think you're missing their point, even if I can agree with part of your premise.
There was an outdated but relevant saying
'In America, you can criticize president Nixon anytime'
'Yes, but in Soviet Union you can also criticize Nixon anytime.'
The point is not that they're safer but that they're not a relevant concern in the same way. (According to OP)
I live in a Zionist country (the US) and will absolutely be canceled and blacklisted from my industry (tech) were I to publicly speak out against Zionism. They are attempting to put laws in place to make it illegal to be anti-Zionist. These laws already exist in countries like the UK and Germany. Some, such as anti-BDS laws, already exist in the US as well.
Many technological advances weren't driven by capitalism, early computers and the internet were literally developed by the government.
You can probably say 864#
I've been using the deepseek api (not for coding though) and have been getting great results and it's so cheap it may as well be free. Another reason I'm using it is because I like the license and I also have some hope of running it in my own hardware in the future.
But the thing is... I could be using any of the llms for my use - I'm using a middleware that lets me change providers only with a configuration change.
So it's going to be tough for USA ai companies to charge 5x to 20x (depending on what you're doing).
> where it matters most: commercialization.
It begs the question because both its premise and assertion are already wrong. Has AI improved the industrial capacity of the US in order to improve the lives of its citizens? No it hasn't. Has AI increased the wealth of its citizens by being able to do laundry or any household task in a generalized way? No it hasn't. The only thing it's really done is make very narrow slices of white-collar work more fungible. In what way has AI been able to address existing shortcomings of the US?
Amazing. Someone on the internet using 'begs the question' correctly...
Glad I'm not the only one that noticed this.
You better have some sources for declaring that industrial capacity hasn't increased. The Fed reports around 8% penetration of AI in manufacturing already, but in my opinion it's too early for grand declarations like that without data.
https://www.federalreserve.gov/econres/notes/feds-notes/moni...
> The Fed reports around 8% penetration of AI in manufacturing already, but in my opinion it's too early for grand declarations like that without data
Based on a survey if the business uses AI "in any of its business functions". And for all uses of what they consider to be AI, not just LLMs.
> ... it's too early for grand declarations like that without data.
You mean grand declarations like 'industrial capacity has increased'? Just because AI is present in the factory doesn't mean it's actually increased capacity.
> The Fed reports
Have you happened to purchase anything in the past 12 months, and looked at the Fed's inflation numbers?
> Have you happened to purchase anything in the past 12 months, and looked at the Fed's inflation numbers?
The Fed doesn't issue inflation numbers. The usually cited headline inflation numbers (CPI) are from the Department of Labor’s Bureau of Labor Statistics, the ones used by the Fed as an input to monetary policy decisions (PCE) are issued by the Department of Commerce’s Bureau of Economic Analysis.
AI has definitely improved the industrial capacity of the US
and of everyone else, right? what service or product is only available to the US? Even with Chinese models lagging behind, the difference in capabilities is not much.
Assertions made without evidence can be rejected without evidence.
t. literally works on AI for industrial applications
Computer vision certainly did. But LLMs? That needs citation.
What capacity?
How?
Such as?
The current commercialization isn't economically sustainable.
>> The current commercialization isn't economically sustainable.
And if it were, and the result were like Elon and Scam Altman say it would destroy the economy. Not sure any country wants to lead the race to self destruction.
I don't think so. From a nation state perspective, AI is a munition. Every advanced nation is going to have their own cyber division with their own AI hosted within its borders. Considering how xenophobic and belligerent the US is, nobody is going to want their national cyber defenses hosted in the US.
On a personal level, I simply do not trust the US anymore. I won't host any of my personal data in a US company. I don't want the US govt invading my personal privacy, and their corporations are constantly leaking and selling private data. I consider US to be rapidly approaching complete autocracy (on par with China) so US-hosted AI is a non-starter. And let's not forget local inference keeps getting more efficient, with higher context and TPS in the same amount of RAM. Within a year even small consumer machines will run local models good enough for basic coding, and in 3 years RAM prices will lower and everyone will be able to afford a decent rig.
Finally, open weight models are now good enough for daily work. They may never be as good as SOTA (SOTA will just keep increasing indefinitely), but that doesn't matter; my car may not be as fast as a Porsche but it still gets me to the grocery store and back. So I use non-US hosted model providers which provide open weights, which are both significantly cheaper than Anthropic/OpenAI, and actually allow me to use my subscriptions without a moat.
But yes, Anthropic/OpenAI are absolutely the new Oracle. They will win for US govt and Enterprise contracts. But that's far from the only users of AI.
The US is not xenophobic. That is ridiculous. Any time you say stuff like that, you discredit the things you say that actually make sense. I'm with you on the privacy aspect, but there are multiple dimensions of that which you're ignoring. I'd much prefer taking my chances in the US than in the EU, where they are constantly trying to push companies to weaken privacy.
Europe has way stronger data protection laws than the US. EU has GDPR, strict requirements, large fines. US only has a couple states protecting personal data, with HIPAA for health data, and that's it. We require you to unlock all your devices within 100 miles of a border (inland) so we can look at all your data. Of course our intelligence service also hoovers up the metadata of US citizens in contact with anyone overseas, which is borderline illegal. All our states are now passing "age verification" which is mass surveillance under a different name.
And US absolutely has been xenophobic for years, by official federal policy. I'm really surprised you're not aware of it, but here's a small selection of examples:
- Both our elected and appointed leaders are white nationalists. Our president called all Mexicans murderers and rapists, said African migrants were eating random pets in a rural US town (they weren't, obviously, but it was intended to exacerbate xenophobia)
- Our federal government has a mandate using ICE to try to eject anyone with a Hispanic name from the country (has already deported US citizens based on being hispanic/latino). We even boot people seeing asylum, often exporting them to foreign prisons even if they've never had a criminal record. We have concentration camps now, filled entirely with foreigners, and people who have lived here for decades but were foreigners.
- We stopped accepting new visas from 75 countries. We may even expel you for social media posts we don't like, or for attending a protest that our citizens can attend. We increased travel bans for people from majority Muslim countries. H1-B visas have been rolled back to only the highest paying jobs, and you may need to pay a $15,000 bond. We also now collect and store foreigners' biometric data indefinitely.
- Let's not forget the tariffs on virtually all other nations, to say nothing of "America First" and the new "Greater North America doctrine".
The US accepts immigrants from 200+ countries around the world with the top 5 being Mexico, Cuba, India, Dominican Republic and China. None of that has changed under Trump.
I think you got lost in the rhetoric somewhere.
Tariffs are just the US adjusting to reality which other countries are slow to do. Free trade died all on its own, because the pandemic showed that critical industries were hollowed out by free trade in a way that could be appreciated from a national security perspective. That situation was favoring China too much, so we need to unwind that some.
Tariffs already existed in many countries in practice, so it's not like the US reinvented modern tariffs.
Putting some numbers into the discussion census.gov [0] is tracking a sharp decline in net immigration due to both, a decrease in immigration and an increase in emmigration, from the start of 2025 to the present. Trending towards a net negative.
Pew [1] suggests that the changes around the start of 2025 were due increased restrictions on asylum applications under the previous admin and EOs by the current one to restrict new immigration. Given the rough numbers [2] of about 40k asylum grants per year in the early 2020s, I doubt the previous admin's actions are playing much of a role here.
Stating that none of it (immigration acceptance) changed under this administration might technically be true - with respect to the number of countries applying, but misses this point.
[0]: https://www.census.gov/newsroom/blogs/random-samplings/2026/...
[1]: https://www.pewresearch.org/short-reads/2025/08/21/key-findi...
[2]: https://usafacts.org/articles/how-many-people-seek-asylum-in...
Are you seriously trying to reframe the largest tariff war in 100 years, targeting 180 countries and territories, as a readjustment against China? And in both of Trump's terms he's radically changed immigration more than at any time since the 1960's. Either this is a great troll, or you need help, man.
Free trade isn't only a China issue, no. It's only the most important one partly as a function of China propping up massive state companies while also trying to avoid becoming a consumption led economy.
If you feel like formulating a good argument about immigration, I'll listen, but you haven't provided one.
Europe seems both better capable of sustaining democracy, privacy and rule of law. USA is on verge of being irreversibly done for in all three areas.
It can happen in Europe too, but the full fall is not that close.
Privacy is a concern everywhere, but the center of gravity of the issue moves further up or down the chain depending on the country.
The structure of the US makes it basically the single most secure democracy anywhere right now or in history. No country in Europe or Europe as a whole is even competitive by comparison. The main issue we're facing is that we are by far the primary target for foreign funded activism and systemic attacks, because China and Russia hated NGOs promoting color revolutions.
That is also part of the rule of law issue, but the system is overall managing quite well. It's all moving in slow motion, but many important metrics are going in the right direction, which we need as that's part of deterring China.
> The structure of the US makes it basically the single most secure democracy anywhere right now or in history. No country in Europe or Europe as a whole is even competitive by comparison.
How do you figure? I hear you have roving gangs of masked thugs beating up random citizens with the backing of your government, that doesn't sound very democratically secure, especially with what healthcare costs over there.
> The structure of the US makes it basically the single most secure democracy anywhere right now or in history.
This is just not true. It is failing visibly and loudly fast. It used to fail slowly but the process speeded up.
American administration supports Russia now. It praises Russian, Chinese, Belarus leaders again and again. It praises Orban. It hates last bastion of democracy - Europe.
China is not detered. Its power is growing while American one is going down. Trump openly admires its leader. China is celebrating current state of America.
> the single most secure democracy anywhere right now or in history.
So secure, in fact, that it has secured itself even against the influence of its own citizens.
That's not really accurate. The US is structured so that it is self-reinforcing from the bottom up and the top down simultaneously. State laws cannot violate the U.S. constitution and many types of elections cannot be gerrymandered. Even gerrymandered legislatures have limits on what they can do. You can't simply have one party change a state's constitution. Even congress can't be entirely gerrymandered.
Also, we have guns. LOTS of guns. The U.S. military's first and sole responsibility is to the constitution itself. If any state or the federal government tries to get rid of their constitutions, the military can rightfully take it over and re-establish a constitution.
There is no other country that's even remotely close to this secure.
Inclined to disagree.
The winner here will be whoever can move atoms with AI not take notes at the daily standup.
i.e. Think boston dynamics vs unitree
They're both doing well but I'd lean towards China is winning on atoms in light of a huge manufacturing base they can AI-ify.
We are _miles_ behind successful embodied AI. The demos are cool but the success rates are not high enough.
You can tell we're on the cusp when level 5 self driving cars are common an you have multiple companies deploying them on the street. Google is doing great work but the poured TONS of effort into it and the thing still needs intense stacks of perception and processing. Much more than I've seen any humanoids pour into it.
L5 SDV's are much easier to get than humanoids and the have tangible economic benefit. My thesis is that those will come first.
I'm really curious how quickly we would have huge numbers of L5 SDV if we societally accepted ~equal rates of injury and death, both of passengers and pedestrians. I want to be very clear, I'm not advocating for this (and even if I was, I haven't the faintest idea how one would go about getting society more broadly to go along), but part of me thinks that the primary hold up isn't actually capacity but instead standards.
This doesn't really argue against your point, because the standards are what they are, and like I said, I have no idea how one would go about changing them if one even decided they wanted to. And given what they are, it has taken, as you point out, enormous amounts of effort to reach those standards in a practical way.
That all being said, while I agree that SDV's are in many respects easier than other robotics tasks, they are also somewhat uniquely dangerous. Other categories of task, while potentially more complicated, won't have to worry nearly so much about safety, and so may be operating under a different constraint regime. I think this means that we may see adoption happen at a much more accelerated rate than we have seen in the automotive space.
Will we all be more or less flesh and bot in the future? Robocop style
Standards are not higher for self driving cars. Musk lied a lit about capability and safety of self driving, creating impression that it is safer then humans driving.
So far, they are not.
I have no idea where you get this impression. Tesla is no where close to the majority (or even plurality) of fully autonomous self driving miles. Waymo is dramatically safer (less injuries, not quite enough data yet to be certain about fatalities, but they are lower than average, we just can't yet claim statistical significance) than human drivers.
I haven't seen good stats on Tesla (they are less transparent than Waymo), but it would shock me if they weren't also at least slightly safer than the average human driver. Human drivers are really bad at driving.
But even if Tesla isn't safer, taken as a whole, the self driving industry as it currently exists still probably is, purely because it's mostly Waymo, and Waymo is dramatically safer.
Strongly agreed. AI powered drones will be the winning military strategy by 2030.
And why are atoms necessary? You're treating embodiment as the _only justifiable_ commercial path for AI. I don't think that's really close to true. Embodied AI is a subset of current LLM/agentic AI products (or perhaps intersection of something and this new AI?). No reason anything needs to move atoms _directly_ (e.g., via motors) to make a trillion dollars.
The winner is whoever can move the atoms for free, e.g. crack energy.
Mostly agree. I think there is a big time delay though.
If free cheap energy is unlocked today I reckon it would still take a good 30 years for that to ripple through properly.
It solves lots of problems (water!) but doesn't make the heavy machinery to consume it instantly appear.
K. Eric Drexler's "Engines of Creation" got a lot of favorable comment when it was published but then it kind of faded from view. Might be worthwhile to revisit it?
> They're both doing well but I'd lean towards China is winning on atoms in light of a huge manufacturing base they can AI-ify.
Why would an American company outsource manufacturing to China if the labor cost is the same in both places? The entire reason the Chinese manufacturing base exists is to exploit cheap labor.
What would be the point of shipping products across the ocean?
Labor is not the only cost in that equation though, there's business regulations, the cost of the operators/repair that troubleshoot and repair the bots when they break, etc. a lot of which could be cheaper still than the price of a container on a slow ship from China.
China doesn't have to ship parts across the ocean. US manufacturers do, because we gave up the whole bottom/middle of the manufacturing supply chain to China decades ago in pursuit of lower costs. In China, the maker of the parts you need to build your product is a few blocks away. In the US, the maker is in China.
And, if you need changes, you can go talk to them the same day you see a problem.
There are more competitive advantages such as expertise, supply chain, regulations, capable government, scale... If it were all price, there are countries that are much cheaper than China.
This para caught my eye
>Frontier cyber models may push states and defense firms toward the opposite logic: security by obscurity, with closed software, closed tooling, closed firmware, and closed chips. If a model cannot train on the code and architecture of a target stack, it will usually have less context and less speed. That does not make systems safe, but it does raise the value of proprietary stacks all the way down to hardware.
Is this really true. Are there any experts who can weigh in on this.
Should we interpret this to mean that in the new world Windows is more resistant to attacks than say Linux.
I think there’s some credence to the concept that more context == faster iteration cycles. Source code can be one major source of context.
I think “security through obscurity is no security” concept was aimed toward people not relying on obscurity alone as a security mechanism. And largely that message succeeded. But now we are in a rapid acceleration of capabilities (on both sides) where any advantage to one side will result in outsized gains, at least in the short term.
In general: less data = less "intelligence".
And basically all the security bugs I've read about were find looking on the source code.
But it doesn't mean windows is more secure, just image a scenario where someone is stealing windows source code and sell it to rogue actor, it will make it even less secure because no one (expect windows) would have had the chance to search for bugs in the source code.
> Should we interpret this to mean that in the new world Windows is more resistant to attacks than say Linux.
LLMs can read assembly better than most, so probably not. But reality has never stopped people from trying to obfuscate.
Alibaba's cloud is something that the author of this article seems to dismiss. It's being used massively in Asia and they're pretty close in services and offerings to what AWS, GCP and Azure provides. Once they start doing inference on their own custom chips it might be hard to compete with them due to the energy costs
No, Alibaba is excellent top-5 easily.
>Trump fits this moment well. He is a salesman at core, and Larry Ellison is too. That helps explain why AI infrastructure is an easy political product. Selling AI today is easier than selling Oracle databases in the 1980s.
I feel like the author (and perhaps many here on HN) are on a different planet than almost everyone I interact with.
Considering that you can easily swap out one AI for another and there is zero lock-in potential, does it really matter who is winning now?
Government, industry, and general integration into the business software stacks are not easily swapped out.
Most businesses are adding limitations on using open models.
> Government, industry, and general integration into the business software stacks are not easily swapped out.
My business's integration literally has a dropdown for which model you want to use. I think that's pretty standard.
couldn't you say the same about search?
I continue to be impressed by our collective willingness to engage with obvious AI slop, as long as it also talks about AI. Sincere question for any of the nearly 300 folks seem to be arguing about the article: why? The author couldn't be bothered to present their case, so they probably don't care about our opinion. They just want traffic and search ranking with the least amount of effort. The community is literally being played for clicks.
Is it just that the subject line alone is a springboard for casual discussion? If so, maybe that's fine, but then, it feels like we'd be better off cultivating these discussions as "ask HN" posts instead of boosting this kind of web content.
> Is it just that the subject line alone is a springboard for casual discussion?
I think this has been the case on many sites, for decades. Many people just want to read and write comments without engaging with the OP.
Have a look at this Reddit thread [0] about this Ars Technica article [1] - both are 15 years old.
I suppose in the 2010s this was an amusing detail of online discussion. In the 2020s it makes me feel a little uneasy - it suggests that the entire concept of people jumping from site to site, clicking links and understanding what they are writing about was flawed from the start. No wonder the internet became centralized and slopified.
And no, I didn’t read the OP, I found your comment to be more interesting to discuss. These days with AI articles flooding the internet it seems foolish to actually read articles before the comments.
Edit: although we have to contend with AI generated comments as well. I wonder how many of the comments on this page actually have original insights into the politico economics of AI.
[0] https://old.reddit.com/r/WTF/comments/gz9k7/the_internet_is_...
[1] https://arstechnica.com/science/2011/04/guns-in-the-home-lot...
This is HN equivalent of pretty girls changing clothes on TikTok. You would learn more about AI from reading a cereal box than this blog.
Same as software in general I guess? Lots of critical software has been developed by Europeans, but it's the US who build hyperscalers with it
Lol some of these western arm chair analysts actually need to visit China.
The whole "race" narrative is silly. It is all built on the assumption that one country (corporation, actually) somehow creates AGI and thus, essentially, the singularity. Great for raising VC, apparently, but at its core this is magical thinking.
Even if any of the US corporations would eventually end up in a scenario where their revenue is at least as high as their inference cost, what harm would that do to the other contenders? It's not as if there is any kind of network effect here that would exlude them from market participation.
No US has been "winning" or rather leading thus far, but there is no guarantees that it will ever "win". I do not subscribe to idea of omnipotent omnipresent AGI. China plays a long game, I think DeepSeek does not engage in platform building on purpose. DeepSeek was probably assigned with a role of a primer lab, the goal is to replace CUDA, align with Huawei chips to do cutting edge research and cross-pollinate other teams in China. They might even hide the best models on purpose. In a long run China will use its industrial capability to apply and use AI better than anyone else. And that would be a good thing for the World.
“It is not the same as profitable AI leadership”
Where are these profits of which you speak?
How does commercialisation matter when China will pump however much money they need to into this? This is a national security issue for them, just as it is for the US.
Not even gonna bother clicking through this one, the title is that egregious. And by the way, you can be damn sure that if Anthropic or whichever other American frontier model model is the best of its day was on the cusp of going under, the US gov would either pump it full of government contracts or (less likely) nationalise it.
I feel like the title of this post should have "for now" appended.
Why though? That could be true of any economic condition. Imagine if any time there was a race or competition in which one group is winning you had to just say "for now".
Michael Phelps is winning the race! ... for now
China is winning the EV race ... for now
It doesn't seem to add value to me, aside from being an opportunity to, as is the time-honored tradition of the haters, to sow doubt and create negative energy to anything related to American success.
+1
Of course US has a huge head start, but if AI keeps growing, what matters is how the market's gonna look like years from now.
Most of my clients using AI in the business workflows (in products) use Chinese LLMs, because after benchmarking for a specific use case you nearly always end up finding that you pay half or a tenth.
That's not a new phenomenon. I've adapted Gemini Flash 2.5 years and years ago when people were dissing it as "crap", yet it was the best budget and quality fit for the task I had at hand back then (translating and summarizing tons of documents). It was both faster and around 100 times cheaper than the best GPT 4 model available.
Needless to say, medium-sized Chinese models are far better than those LLMs and a perfect fit for countless applications.
Even with all the Qwens and Kimis and GLMs etc, the latest Gemini Flash models are still an insane value! I recently settled on 3.1 Flash Lite after testing basically everything on offer on OpenRouter, and it was not a close call - cheaper, faster, and better (for translation and visual understanding tasks).
It’s interesting to see short term business interests again undermining American sovereignty.
Just as business exported strategically critical manufacturing to China, now it is helping funding China’s race to take over the US in AI and beyond.
Lesson is pure free trade doesn’t work if (a) not everyone is playing by the same rules and (b) the trading territories are or may become opposed.
American economic policy gave the world an authoritarian super power and Trump. Not a great track record.
> Many people use the wrong scorecard.
Correct. "Revenue" is the wrong scorecard when they're selling 20$ bills for 15$. I too can make a bajillion dollars in revenue with that strategy.
Show me a company not speed running the uber/doordash playbook and we can talk.
If it is a war, then China is delighted.
The USA is very good at loosing very, very expensively....
Matters the most to whom? I certainly will not care about expensive models that do about the same thing cheap non-american models do.
It's like the USA Librem 5 vs PinePhone. About the same HW for $1600 vs $150.
Sure will not pay 10x for "US" thing just because it's a US thing.
That's like saying that Louis Vuitton is monetizing shoes the best. Sure, but it's not winning shoes
Commercialization is not enough. The US is built on financialization.
Cultivating an ecosysyem of strong capital protections, wealth creation through extraction, and tax advantages for AI finance is what we should be looking for. Commercialzation may be a step towards that, but isn't the destination. We have to create a system where those with money can multiply it, not simple add to it.
Just spitballing here but I think the financial system’s already set up for this.
Whatever derivative structures and equity and options need to exist will be easily created.
I don’t think we need any additional motivation or incentives to cultivate this for AI. We need to keep some in the tank to handle the fallout.
As a more personal aside: the US would do well to put up some sensible barriers to outrageous financialization and reduce moral contagion risk. Otherwise all these folks trying to multiply their money end up leaving the bag with the folks that don’t have it in the first place - and then the folks with money end up, uh… well, it won’t end well.
Betting your entire country's future on usury seems like a terrible idea to me, but what do I know.
You're really not going to like what happened in 1971.
https://wtfhappenedin1971.com/
Don't look now, but if you've got a 401k then your pension is already dependent on perverse financialization.
“A system where those with money can multiply is” sure sounds dystopian to me.
It's (beautifully) dry humor making fun of OP, whose post is rather dystopian already.
Oop you’re totally right! Bit close to home. :)
"Where it matters most": accuracy and repeatability?
Sorry, nobody's winning that AI race.
GLM, Kimi, MiMo, Minimax, Deepseek, Qwen would like to have a word. :)
This.
I dont know what the benchmarks are supposed to represent, but to me Kimi K2.6 is indistinguishable from e.g Opus 4.6.
How much would one have to invest in hardware to feasibly run a usefully large Kimi K2.6?
There's no set-in-stone answer to these kinds of questions, it all depends what "useful" performance means to you.
Yeah, I know. Was just throwing it out to see what folks would say based on their idea of useful.
Around $35-45k if willing to use it comfortably (fast to answer, large context) as coding agent.
Even when other countries won, the teams would have reloated to US.
This sounds like an ad for US than anything else.
Does any of the US companies earn money on LLMs? No, they bleed money. Github Copilot is switching to token based pricing, which will be costlier than hiring juniors.
Anthropic also is switching enterprises to token based pricing from their subscription one.
From the big three only Codex is still in somekind of subscription pricing, but they'll shift eventually (usage limits are a kind of that, but they have them less stricter than Claude ones)
There is one winner in this race - China. Trump with his agendas and wars makes it even more likely that China will lead this new market.
> Does any of the US companies earn money on LLMs?
Inference? Yes.
Infrastructure build and training? Not yet.
US AI models would not survive a free market re: this metric.
Same can be said of healthcare.
And, yet, the US AI companies are not actually making a profit, right? They're selling at a loss and trying to make it up in volume (or lock in some kind of monopoly position, in a currently non-sticky product). We're all currently enjoying investor-subsidized tokens from the big guys, and that pushes out the reckoning for US AI. But, I think they're beginning to think maybe they need to ring the cash register. Copilot dramatically reducing usage limits and what models are available on its plans, Anthropic playing games with what's included in the Pro plan, etc. I think they're starting to feel the bleeding.
Not only is the investment that keeps US AI companies flying high slowing, I suspect in two or three years, we'll all mostly be using open models and the people making money will be the hardware manufacturers. Even the small models will keep getting more capable. I'd guess a model you can run on a high end, but not outrageously overbuilt, developer desktop or laptop (something like 128GB of unified RAM), will be competitive with the current frontier when it's allowed to search the web and do research and write test code. You can't fit as much knowledge in a small model (80GB of weights can't store the world's knowledge), but I don't have the world's knowledge in my head, either, and yet I can figure out most problems with a little googling and experimentation. The reasoning and tool use abilities of smaller models is where the gap is closing, and that's what will make the huge models obsolete for huge classes of problem.
Already, there are many classes of problem that the easily self-hostable Qwen 3.6 27B can solve that required a frontier model a year ago. When the self-hosted options reach Opus 4.5-ish levels of capability, the argument for paying for tokens for most work begins to look a lot less compelling. And, looking forward, 1.58 bit models are coming. Incredible intelligence density, and still a lot of improvements happening.
>> And, yet, the US AI companies are not actually making a profit, right?
I think they already, actually making profits especially Antropic. But think how important it's from a business standpoint - the entire software stack from OS to Databases to browsers will be rewritten in the near future, for a company such as Oracle or IBM it means their bread and butter/cash cow can be replaced. It's worth almost any kind of Capex. And from Washington standpoint it's more important than F-35 program or even Apollo mission.
I'm glad we went to space, truly. Racing the USSR might have been the wrong reason but it got us there. We've benefited immensely as a species from exploring the solar system and looking deep into the universe.
I'm not certain that racing China in AI is the right reason but it might get us... somewhere.
It is good for China.
>The US is winning the AI race where it matters most: commercialization
Mass unemployment and an eventually collasping economy is winning?
it's an exhibition, not a competition
> Trump fits this moment well. He is a salesman at core, and Larry Ellison is too.
Larry just fired 30% of his people at Oracle because, apparently, he is in an immediate need for cash. Because Oracle's early AI bets aren't paying off.
Ppl don't understand Commercialization is not incidental to the Western system, it's why we beat out Communism. Commercialization incentivizes ppl to build, bc ownership and control.
The FSF was not an attack on commercialization, it was about giving users more freedom with their own copy.
AI commercialization is why we will always be a few steps ahead in AI.
The Chinese and Russians are free to join us. It's a pickup game.
This is both true and insightful, but the "its us capitalists vs. communists" framing obscures some very important details.
For one, "Communism" is presented as a single monolith, but it's not: it's socialism PLUS despotism. The despotism part is really important! China/Russia/etc. fail because they try and control things top-down, instead of letting the market decide.
However, you can have socialism without despotism! Tons of European countries are far more socialist, but no less democratic than America (many are more democratic).
So yes, America vs. Russia/China and Capitalist vs. Communist are relevant frames ... but don't let them obscure the fact that you can have a successful, democratic country .. without doing what America does (and giving all control to corporations).
They're not independent. They are the systematic consequences of a naive intelligentsia pressing their ideals that are not grounded in precedent. They are susceptible to manipulative despotic takeover. The pattern keeps repeating. Across geographies large and small.
China has pretty much market economy and is not socialist at all. These economic changes happened years ago. Russia has no communism either.
China is despotic in its treatment of political dissent and human rights, but not in economy.
Well we're certainly losing in terms of public sentiment. There's a real anti-technology mindset that has taken hold. Basically the equivalent of people protesting the electrification of the country early last century.
Chinese culture is quick to embrace the benefits.
> The US is winning the AI race where it matters most: commercialization.
puke
Yeah, go ahead and run your country into the ground because of hypercapitalism and hypercommercialization, you're almost at the end game now! While the rest of us try to figure out how to actually build societies worthwhile to live in and experience, with healthcare and not waging war on our neighbors.
I don't know how people can seriously publish stuff like this and not feel like they're actively trying to make the world worse. Is money really the single thing y'all can focus on? Is there nothing better in life you can chase, even if it's also a number? So sad to see stuff like this.
They are trapped in the copium den
Hahaha, but no.
It's like people forget the entire point, perhaps even definition of technology is "doing more with less."
The "brute force" of power and cycles is almost certainly the least important thing, perhaps even a hinderance.
… and it’s substantially due to foreign born researchers and engineers. US will win as long as smart and driven people will want to move over.
The US is winning the AI race in all matters.
Back in the dot-com bubble, people started inventing new metrics to "value" dot-com companies that lost money hand over fist. My favorite was "revenue multiples". So instead of a a P/E ratio, it was just a multiple of revenue no matter how much money you lost.
We've invented a new term here too: revenue backlog. OpenAI and Anthropic in particular need to recover probably at least $2 trillion to recoup their capex investments. Now Claude code has had an impact on software engineering but for a lot of AI uses you're just not going to recover $2T on $20/month subscriptions. It reminds me of Twitter trying to dig itself out of a $44B hole and losing half their ad revenue with $8/month blue ticks.
The only commercial product AI sells is labor displacement and the resulting wage suppression. You lay off 10-20% of your staff and nobody is asking for raises. The people left are happyt o still have jobs (and thus a house). They'll work even harder doing unpaid labor of the displaced workers to keep those jobs. That's what OpenAI and Anthropic are selling.
The problem is that if these companies get their way, 10-20% of the population is going to be out-of-work and society is going to fall apart. Data centers are going to be the targets of increased societal desperation and anger as this gets worse.
There was a report this week that roughly 50% of singles in the US aren't dating because they can't afford to [1]. This goes well beyond the well-understood problems of not being able to afford a house let alone start a family. This is a birth rate death spiral in the making.
So, back to OpenAI and Anthropic, the only way they justify their valuations and can make up the "revenue backlog" is if they have a moat. And I don't think that's going to happen. Hardware will get cheaper. Nobody is talking about how the generation of AI hardware will write of trillions in investments for some reason. I don't know why.
But the dark horse here is China. DeepSeek when it was first released (early last year?) was a shot across the bow. We have it and toher models (eg Qwen) that will close the gap with whatever OpenAI and Anthropic produce such that no company will "own" AI in the way that OpenAI and Anthropic need to. In the coming years, China's chipmaking is rapidly closing the EUV gap and Western companies have zero penetration into this market. China doesn't want to be dependent on foreign tech that can be withheld at any moment.
Don't believe me? Just listen to the NVidia CEO say the exact same thing [2][3]. Huang realizes this is such a problem that he's gone on Air Force One to this week's Trump summit in China to try and convince the Chinese to buy NVidia chips.
[1]: https://parade.com/living/nearly-50-of-single-americans-not-...
[2]: https://www.youtube.com/watch?v=aJmHfmrRMUE
[3]: https://www.youtube.com/watch?v=Hrbq66XqtCo
Labor displacement is also called rising productivity.
No it doesn’t - it means cost efficiency. Incremental productivity can remain flat - unless you’re a bozo who counts LOC.
It's not "AI bubble" - at this point it's a software bubble. It's Antropic or OpenAI that should justify their valuations, they have close to billion customers at this point. It's non-AI software companies without strong cloud business that must justify why their core product is not going to be replaced with the help of LLMs. It's not "fair" but that is how it's seen by the Wall Street.
> There was a report this week that roughly 50% of singles in the US aren't dating because they can't afford to [1].
Poor people with nothing date when they want to. If people have interest in having partners, they can date and socialize for free.
No, it’s not.
It’s all about adoption and the bigger picture. The US is an untrustworthy, isolated island in the AI future if you vote another idiot into office in a few years. If you’ll still be able to vote at all, that is.
The largest part of the world is not the US. The cutting-edge US models are way too expensive for most parts of the world, and that also shows in adoption.
China is building an ecosystem of open-source models that are both cheap and good enough for most use cases. While most of the US AI sphere will collapse under the pressure of making profits, which means having their models and infrastructure adopted by as many enterprises and individuals across the world, China’s models will have become global standards and hard to displace.
If Beijing’s AI pitch centers on universal access and cost-effectiveness, then Chinese AI firms do not need the latest chips to win the global AI race. They also don’t need the expensive US-run infrastructure. If you watch Chinese AI adoption closely, they already want as many Chinese people as possible to be able to build and try with AI, whereas for most Americans, US models for productive use are already too expensive.
Kimi K2.6 sits within touching distance of Opus 4.7 and GPT-5.5 while costing about $4 per mil output tokens. That is six to eight times cheaper than cutting-edge US models. If you run hundreds of agents, that’s a significant opportunity to get the same work done for a lot less.
Even early adopters like Singapore, ditching US models, the government kicked Zuckerberg in the nuts and went for Qwen instead to build its sovereign AI models.
To understand why the US is at a severe disadvantage in this race, you need to understand China’s Belt and Road Initiative (BRI). BRI entails Chinese firms delivering fully financed infrastructure projects in a bid to lock third countries into China’s economic orbit. They use the same approach for their open source ai models, but this time the infrastructure is both invisible and free.
No need to build power plants or buy /build ports. AI dependency is invisible to both policymakers and the population, limiting pushback. No pesky activists in Germany nagging about China buying parts of ports. No African nutbags questioning why the humble Xi is building hospitals in areas Chinese mining companies take things out of the ground for pennies on the dollar.
China is going for a marathon here while the US tries to push their ai tech by sheer force into the throats of the world. As soon as Chinese ai models have become global standards, it’s game over for us ai companies. And China is way better at this game than the US. They have proven this over and over again in the past 50 years.
I recommend reading the China Standards 2035 strategy to get a better understanding of their approach and how smart this is.
https://www.china-briefing.com/news/china-standards-2035-str...
AI is not as revolutionary as you think in terms of our experiences with previous technological advances in terms of trade and economics.
Western economies are locked into U.S. models, while China runs on Chinese ones. It’s the age-old game. But the real war of the AI race will be fought in the global south.
I will give you three examples.
Can you really imagine, if you look at what AI needs to cost to make a profit, that even at the current prices, US models and infrastructure, which are heavily subsidised already, being used in cost-sensitive countries? I am not talking about coders, think really big here for a second.
Secondly, US ai models are trained on Western data. How do you expect them to grasp local contexts in the Southern Hemisphere? Chinese open-source models, on the other hand, can be downloaded and finetuned with country-specific data.
Want an example? Check out AfriqueQwen-14B, which is adapted to the top twenty African languages.
So I think this author is wrong. The ai race to be won is not hardware or cloud infrastructure, my money is on it will be a contest to decide which models and standards become the default infrastructure in countries that are up for grabs.
China neither needs the best models nor does it need the best cloud infrastructure, it just, like so often, only needs to be affordable and good enough to become the default choice in emerging markets.
The right choice would be for everyone to step off the gas pedal and think about whether we are willing to become China in order to beat China. Our ancestors worked really hard to get us here, our rights, our ways of life, culture, all the blood, sweat, and tears.
AI better be worth it in the long run for all of humanity if we go back to survival of the fittest. Because that is what it will take to beat China at their game.
> While most of the US AI sphere will collapse under the pressure of making profits
I think deep down, sama knows this and that's why he's pushing for "Universal Basic Compute", which really means forcing every US citizen to become an OpenAI subscriber.
That's nothing new, we had BASIC computers back in the 1980s.
>If you’ll still be able to vote at all, that is.
Stopped reading here. What a ridiculous statement and I can only assume the rest of your post is just as ridiculous.
Why do you... seem so sure that this is a ridiculous statement?
The US is running itself, and humanity, into the ground by massively increasing the amount of electricity it uses, instead of reducing electricity and fossil fuel usage. US residents have already started to feel the crunch in terms of water and power in some areas; and the entire world is experiencing the (admittedly less critical) shortages in RAM and SSDs.
And that's not to mention the warping of US economic life by the concentration of capital around this bizarre endeavor, with the circular multi-hundred-billion-dollar deals and such.
Unfortunately, the detrimental effects of global warming arrive gradually, and are spread out over the entire globe, so the "AI barons"/tech magnates will probably suffer the least, while island countries will be completely wiped out, whole regions will become too hot to sustainabily live in, tens of not hundreds of millions will have to migrate, biological diversity will suffer, etc. They will look back on these times in a 100 years and will think of us, or at least of US, as the people boarding the Titanic. Hopefully not as the people who board the Hindenburg.
"Yet like Musk the ouster wounded his ego". So the journalist believes that reacting to rejection with emotions like a biological person makes him like Musk. Err okay.
How about the obesity and fall of democracy race?
This is patently absurd. US AI companies are investing non-existent money on huge infrastructure with negligible income. This cannot be sustained. And if/when it fails it will take down the economy of the US and probably any other country touching us business.
Almost as if concentrating all of the wealth and decision-making into the hands of a few billionaires is a bad idea...
They kept telling us Soviet central planning was inefficient and ineffective, while the lean rationality of the free market something something...
Hopefully it's a race worth winning and not a race to global disaster.
Even if it leads to global disaster, you'd rather be the country with the best AI. You won't be saved just because your AI is worse.
> Even if it leads to global disaster, you'd rather be the country with the best AI
Depends no? If the "Best AI" means "The AI decides when you wake up, go to work, and go to bed", then I probably want to live in the country with the worst AI or even without.
If it instead means "UBI and healthcare for everyone, money lost all meaning and we're all just having fun while AI does all the boring stuff" then yes. But since capitalism still exists, that's a pipe-dream, and "Best AI" won't lead to that for the average person, only for the 0.1%.
This is a false dilemma created by the model companies to try to convince us we must invest or else.
What does the term "AI race" even mean, beyond wooing clueless VCs and soon retail investors ? It's not like the LLMs are some super-secret technology. Any economy willing to sink in copious amounts of money and resources can get it to some level - the question, what's the actual payoff? We have yet to see anything really useful, on the level of step change, besides Johnny who can now spin up demo projects quicker.
Not if they cannot get the GPUs.
If China can't get the GPU's they'll build them. If they can't build them they'll smuggle them, and either way, eventually they will take over Taiwan, and then the US will be wondering what they can cook with leftovers.
Opinions are easy.
The US has destroyed the PC market perhaps irrevocably, and made getting small single-board-computers for new products extraordinarily difficult. It's enraged its own populus with skyrocketing energy prices causing wild consumer shock. It's waged a trade war over chips that has lead China to develop their own, which they have done astoundingly quickly with phenomenal success, far far faster than anyone could have guessed.
As with another recent example, sometimes in war there is no winning: just loss. This is obviously for us programmers an incredible and wild age, filled with nothing short of miracles. It's incredible. But the prices we are paying, the extreme tensions we are creating, the stress and strain of this all has been incredibly unpleasant, and very very very few people feel like they are seeing upsides to this worrisome menacing age, that promises very few people on the planet anything better coming, and which. Has already made life considerably worse, which no nation has yet directed towards helping its people.
The trade war and tariffs are bringing inflation, consumer prices will soar, but from a geoeconomical standpoint this will hurt China (And EU) more than US. US consumer on average has the deepest pockets in the World and people the top will make money on insider trading, stock anyways. If AI tokens will become like US dollar, it will be under total control of the Fed.
>The US has destroyed the PC market perhaps irrevocably, and made getting small >single-board-computers for new products extraordinarily difficult. It's enraged >its own populus with skyrocketing energy prices causing wild consumer shock.
Strikes me as the real outcome: the end of "personal" computing, "local" anything.
so much winning https://layoffs.fyi/
"Methinks the lady doth protest too much"
The word "winning" implies there is an upside.
By this article, "the US" is not winning. Sam, Dario, Elon, Zuck, etc. are.
It remains what benefit, if any, Americans will see from all this...
It's a race straight to the bottom. Anyone who gets excited about being a more efficient and productive corporate wage slave that gets to train their future AI replacement is either a shill or not very intelligent.
> Winning the AI Race
Which one of them all?
If you mean "building models that are very good at coding and as substitutes for search engines", then yeah, sure.
But if you mean: "applying AI to industrial applications and robotics", then China is far ahead: https://time.com/7382151/china-dominates-the-physical-ai-rac...
I mean you can argue the same about Telsa, but look at BYD now.
Just because you are first to do x, doesn't mean you are going to be the winner.
I reject the premise that AI can be "won" at all. It's just another piece of software.
Yes, and it's doing so primarily because of immigrant nerds, H1B's and F1 bros who chose America and may not have this avenue in the future. Potentially, making this the last race USA wins.
It's a myth. IBM, Xerox, HP, DEC was innovating long before H1B's.
There is nothing in the H1B program that makes these immigrants different from the immigrants that ran IBM, Xerox and HP. Other than the country of origin of course.
Jobs, Wozniak, Gates - it's a myth that you need poor migrants pulling themselves by their bootstraps to innovate. Sometimes a nazi scientist like Wernher von Braun is what it takes.
The whole country? Really?
Define winning. They integrate AI everywhere. I hate it. No money shall come from me into AI anywhere. Not sure if I can maintain it, but right now I can.
As an American, we may be winning this race but we are still struggling to define why this is the race to win.
The cost of winning this race has been telling our citizen s we will replace them with robots and there is no hope for their children’s future employment.
The cost has been destroying trust as we tell citizens water and power should go to server farms and not them.
The cost has been naked power telling democracy it’s wrong and dying
I think when we discover the limits of LLM tech and tally its benefits over its cost — we may regret this win.
But don’t let me contradict a bunch of fake techno oligarchs wrapping themselves in war like patriotism to get the investments they need to keep this going.
The goal is global domination as always, unfortunately. DARPA and the Pentagon helped create the Internet, and Silicon Valley later turned it into a major commercial success.
Likewise, the goal is to not get globally dominated by China. You cannot point to one without the other. It's not like the U.S. is the only country that values national security and geopolitical power.
What does it mean "dominated"?
How would your life change if your country became the second wealthiest instead of the first?
This is a ranking and competition no other country in the world gives ... about.
What would the world be like if China led? Whatever Xi wanted it to be like. I do not want a non-democratic, dictatorial power leading the world. You might think that most of the world does not care, and maybe they don't right now. But this is, I suspect, due to the relative benevolence of historical U.S. policy of international law and order. New boss same as the old boss is a motto that is only as informative as the extent to which the new boss acts like the old boss.
Why would the world care? Take Trump's threats against Greenland...actions that run completely contrary to our historical policies and treatment of our Western allies. They were alarmed, because when the leader of the most powerful military in the world makes a threat, you have to treat it seriously. Despite Trump's hubris, such an invasion did not occur because Americans, Congress, made it very clear that Trump would be impeached if he invaded our NATO ally.
Let's say China is ascendant. It is now the dominant military and economic power in the world. China is under Xi's complete dictatorial rule. Xi decides that invading Greenland is a good idea. Stopping that internally would not involve democratic processes, it would need to involve a coup.
Let's make it even more stark. If Germany had won WWII and had become the ascendant world power, would it make a difference to most countries? ABSOLUTELY YES. If there is going to be a dominant world power, the character of that nation matters.
Are there other nations with the character and institutions that could do as well or better than the U.S. has done? Sure. I can think of several nations. But let us not pretend that all nations are equally bad/good for the world.
uh, what? I'm pretty sure the AI race matters most is improving society. That absolutely does not equate to making money off of things.
There will be no winners in this once the jobs start to disappear. There is no 1:1 with new human-in-the-loop jobs. Its basically in the definition. We go from being the loop to be "in the loop" .. huge difference.
Ever since I was a small child it was my sincerest wish to build so much value for others by becoming a human-in-the-loop. I am so happy my wish has finally come true.