Interesting analogy. The obvious reason so far is that GenAI has very generic output, and struggles massively with specific prompts. Errors aside, even when it gets it right it tends to be like some kind of average of the art style, character or scene. This still has its uses but for any interesting art project or product, the expectation is something unique and novel.
If GenAI can overcome its generic output issue and somehow can tap into some algorithm that gives it more creativity, then I think it will not be like polyester, and it will revolutionize art like photography or film did in the early 20th century.
These past 2 days I have seen some very entertaining short videos made with veo3 in my feed, I'm surprised. Brazilians are creating some really funny mashups of old historic moments with current brazilian culture and slangs. I can see a creative editor making a top quality movie with it.
pirate speak was only fun for a very particular small subset of LLM users, the early adopters. Those videos I'm talking about are as good as any high quality movie scene.
True, it's definitely another tool for creative people already and will at least create a new genre in many categories. The question is whether it will become the dominant force for creative works across the board.
Reminds me of the Transcendental Painting Group in the early 20th century around Santa Fe, New Mexico, whose members explored the nature of substance and form and spirituality within the total abstract.
It was a reaction to the fact that too many contemporary painters in the region focused on landscapes, capturing the beauty of the area through rigid photorealism. This was seen to TPG as derivative and completely missing the essence of painting; after all, cameras were now able to take increasingly vivid shots of natural landscapes, and as such the value of such paintings began to decline.
An excerpt from their manifesto:
The Transcendental Painting Group is composed of artists who are concerned with the development and presentation of various types of non-representational painting; painting that finds its source in the creative imagination and does not depend upon the objective approach.
The word Transcendental has been chosen as a name for the group because it best expresses its aim, which is to carry painting beyond the appearance of the physical world, through new concepts of space, color, light and design, to imaginative realms that are idealistic and spiritual. The work does not concern itself with political, economic, or other social problems. Methods may vary.
Your observation is astute: The recent revolution in generative art and human interfaces is here to stay, and it is the next disruptive and contemporarily misunderstood evolution in art, just as film was to painting. Regardless of how current-generation artists feel, next-generation artists will be born into these tools and adopt them without question, whether directly or through subversion.
One future is in realtime hypercontextualization... Art installations which prize subjectivity more than TPG could ever hope to achieve in their time, creating the abstract not from the mind of the artist, but the observer. Art which is not just observed subjectively, but created subjectively, the observer being able to fully experience themselves from new angles, guided by the hand of the artist.
These installations may be physical or digital, and will use all sorts of signals as input. Local signals, remote signals, colors, shapes, sounds, brain waves, weather patterns, HUMINT, data dumps, trending topics... you name it. Any information will be fair game for integration and resynthesization. Observers will weep, will walk away with a new feeling or realization about themselves, will stay up that night staring at the ceiling and contemplating deep, unearthed aspects of themselves. And for some installations, the observer may continue to participate with the installation over a period of time, whether in person or digitally, and sometimes in a way which incorporates the interactions between observers. This kind of experience is only possible at scale with generative art.
And while still a pipe dream 10 years ago, it's become an increasingly viable reality, especially with the recent upgrade to GPT 4o's generation capabilities, or tools like Sora, or the incredible community tooling around Stable Diffusion, etc.
If I am not mistaken, was not cotton itself - while known and used outside Europe for thousands of years - the polyester, a kind or wonder textile, of the 18th Century Europe that drove the British colonial expansion, the global slave trade and ultimately, in the form of Spinning Jenny, the industrial revolution?
Strangely, while I disagree completely with the writer, I actually think he is completely correct in his analogy, although not the way he intended. I don’t believe AI will be a revolution the way Spinning Jenny, electronic computer or the Internet were. I believe it will be exactly like polyester, in a sense that it will be hyped for a bit, then people find the hype a bit of a silly fad, and after a while just get bored of it while it has become ubiquitous and plain common.
Same way the oat milk carton will be designed by an AI bot directed by some senior advertising agent - as it probably is already - and some video game turned blockbuster video is slapped together by a supervisor and bunch of bots - as it probably is already - but people will flog to see something that is masterful and make them feel genuine human emotions.
AI is a tool like any other and while I’m certain some future creative master will use it to make something truly stunning, without human creativity and innovation it will just churn out some boring slag that while vaguely useful nevertheless make us feel kinda bored and empty inside - kinda like my polyester sweatpants.
I like the comparison with plastics: it is often only used because it is cheaper than the alternatives with worse results. Also there’s a pollution problem.
That seems like a bad analogy, and reasoning by analogy is usually not helpful because no new technology is like what came before it. And there is no backlash for polyester, it still exists everywhere and is part of most fabric mix. If anything Polyester has won by being invisible.
Polyester solved minor issues in fabrics and was generally less comfortable and less odor-resistant than cotton. AI, by contrast, is dynamic and steadily removing pain points with no serious competition from non-AI solutions. The title is catchy, but the comparison doesn’t hold up.
> Polyester is still used extensively but it isn’t the replacement industry wanted
This is almost always the case. It's very rare that an improvement is better than the original in everything.
The new thing is always sold as universally better, and some times it takes a while for the mass consciousness of people to figure out that the new is better in some respects but the old is still better in others.
I disagree with you on AI, the results are clearly poor.
I agree on polyester, though. I sweat and itch wearing polyester clothing, even relatively low blends. Maybe not an issue for everyone, but to me it’s clearly an inferior fabric.
"Our era's particular neoliberal hyper-connected, hyper-capitalist economy is creating a lot of profit for a few people.."
I feel is off base: the last 20 years has seen the absolute carpet-bombing of the very definition of American middle class. An entire new layer of workers (tech workers in their tens upon tens upon tens of thousands) make salaries that blow the old-trend / old-trajectory middle class out of the water.
I understand that the modern tech world has created billionaires in astonishing numbers, but I dislike skipping over the part where vast numbers of people in America are making in some cases full-on double what a without-them income trajectory would see for our middle class. FAANG salaries aren't representative - I get it. But those FAANG salaries are being paid to vast numbers of people - and salaries that what I'll call a "conventional" middle class person blanches at.
Tech cities are experiencing massive inflation in no small part due to this entirely new explosion of what we have to now call the new "upper middle class," at a scale that is unprecedented in the moderate-range past.
Tech has created a wild number of billionaires - I understand that. But it has also created an entirely new, and substantially large, additional income-demographic slice that didn't exit much before.
So to disagree with the author, tech has created unprecedented wealth for a very large number of people.
The class of well-paid FAANG employees is not that large: I'd guessestimate that all five combined employ ~100,000 engineers in the US, give or take. Yes, they have many more employees in total, but many of these are not well-paid engineering roles and/or are not in the US.
If we generously assume that there are 3x as many engineers employed at various non-FAANGs that pay competitive salaries, we still only get ~400,000 people, or around 0.1% of the population of the US.
There's one, maybe two major cities in the US where this has visibly impacted the entire local market: SF and Seattle. Everywhere else tech employment is more or less a rounding error.
Just to confirm I'm not pulling these figures entirely out of my ass, Pew Research says the "upper-income tier" has grown from 14% in 1971 to 21% in 2021 (+7%), but the "middle class" (median income $90k in 2021 dollars) has shrunk from 61% in 1971 to 50% (-11%) in 2021.
So the [middle + upper] tier has shifted from 75% to 71%, for a 4% net drift downward? While 7% have drifted upward from [middle + lower] classes to upper?
The Pew link is fascinating me, thank you for sharing it! One thing that stands out to me is just how extreme the upper-class skew is—its median being 10x the median lower-class income. That helps me contextualize the types of graph that describe the “decline of the middle class” as a share of total wealth: it makes it seem more like a story of “lots of middle-class families’ wealth is wealthy in global terms and also growing, but a shocking number of families have shockingly much money too.” To some degree a story of middle class decline relative to the people richer than them—but to an even greater degree, a story of a burgeoning upper class.
Certainly you’re right that highly-paid tech workers are a drop in that bucket. To me, though, the overall numbers frustrate the “decline of the middle class” narrative to some extent. I guess it’s considerably worse to sink from prosperous classes to lower-class than it is to rise from the middle to the upper class—still, the net slice of the population affected by the change is smaller than I would have guessed.
>An entire new layer of workers (tech workers in their tens upon tens upon tens of thousands) make salaries that blow the old-trend / old-trajectory middle class out of the water.
You make a good point. Tens of thousands is a huge number, that’s gotta be like what, most people? Probably at least half of the total number of people.
Now I personally don’t have the time or attention span to count all of the people but I don’t reckon that the number of them is so big that “tens of thousands” doesn’t make up an enormous percentage of everybody. There would have to be so many people that I wouldn’t be able to picture them all, which is preposterous, so tens of thousands has to be astronomical in the grand scheme of things
It has made a LOT for a few people and sure a lot for many. The pay increases exponentially up the corporate hierarchy. Maybe, the author is trying to highlight that.
Agreed. Depends on your definition of wealth, but there are a lot of people who are wealthier because of tech. Even the aircon installers going around replacing analogue controls with digital are making bank.
The question of whether these developments are a net positive for meritocracy seems unclear. It's tho, nothing compared to crypto-- many got propelled into a different class whom I have trouble wanting to give my precious attention to...
(I also get the robotech ref, thanks for the nostalgia!)
Interesting analogy. The obvious reason so far is that GenAI has very generic output, and struggles massively with specific prompts. Errors aside, even when it gets it right it tends to be like some kind of average of the art style, character or scene. This still has its uses but for any interesting art project or product, the expectation is something unique and novel.
If GenAI can overcome its generic output issue and somehow can tap into some algorithm that gives it more creativity, then I think it will not be like polyester, and it will revolutionize art like photography or film did in the early 20th century.
These past 2 days I have seen some very entertaining short videos made with veo3 in my feed, I'm surprised. Brazilians are creating some really funny mashups of old historic moments with current brazilian culture and slangs. I can see a creative editor making a top quality movie with it.
> Brazilians are creating some really funny mashups of old historic moments with current brazilian culture and slangs.
For a time, it was popular to use ChatGPT to translate text into "pirate speak," or something similar. You don't seem much of that anymore.
pirate speak was only fun for a very particular small subset of LLM users, the early adopters. Those videos I'm talking about are as good as any high quality movie scene.
True, it's definitely another tool for creative people already and will at least create a new genre in many categories. The question is whether it will become the dominant force for creative works across the board.
Reminds me of the Transcendental Painting Group in the early 20th century around Santa Fe, New Mexico, whose members explored the nature of substance and form and spirituality within the total abstract.
It was a reaction to the fact that too many contemporary painters in the region focused on landscapes, capturing the beauty of the area through rigid photorealism. This was seen to TPG as derivative and completely missing the essence of painting; after all, cameras were now able to take increasingly vivid shots of natural landscapes, and as such the value of such paintings began to decline.
An excerpt from their manifesto:
Your observation is astute: The recent revolution in generative art and human interfaces is here to stay, and it is the next disruptive and contemporarily misunderstood evolution in art, just as film was to painting. Regardless of how current-generation artists feel, next-generation artists will be born into these tools and adopt them without question, whether directly or through subversion.One future is in realtime hypercontextualization... Art installations which prize subjectivity more than TPG could ever hope to achieve in their time, creating the abstract not from the mind of the artist, but the observer. Art which is not just observed subjectively, but created subjectively, the observer being able to fully experience themselves from new angles, guided by the hand of the artist.
These installations may be physical or digital, and will use all sorts of signals as input. Local signals, remote signals, colors, shapes, sounds, brain waves, weather patterns, HUMINT, data dumps, trending topics... you name it. Any information will be fair game for integration and resynthesization. Observers will weep, will walk away with a new feeling or realization about themselves, will stay up that night staring at the ceiling and contemplating deep, unearthed aspects of themselves. And for some installations, the observer may continue to participate with the installation over a period of time, whether in person or digitally, and sometimes in a way which incorporates the interactions between observers. This kind of experience is only possible at scale with generative art.
And while still a pipe dream 10 years ago, it's become an increasingly viable reality, especially with the recent upgrade to GPT 4o's generation capabilities, or tools like Sora, or the incredible community tooling around Stable Diffusion, etc.
If I am not mistaken, was not cotton itself - while known and used outside Europe for thousands of years - the polyester, a kind or wonder textile, of the 18th Century Europe that drove the British colonial expansion, the global slave trade and ultimately, in the form of Spinning Jenny, the industrial revolution?
Strangely, while I disagree completely with the writer, I actually think he is completely correct in his analogy, although not the way he intended. I don’t believe AI will be a revolution the way Spinning Jenny, electronic computer or the Internet were. I believe it will be exactly like polyester, in a sense that it will be hyped for a bit, then people find the hype a bit of a silly fad, and after a while just get bored of it while it has become ubiquitous and plain common.
Same way the oat milk carton will be designed by an AI bot directed by some senior advertising agent - as it probably is already - and some video game turned blockbuster video is slapped together by a supervisor and bunch of bots - as it probably is already - but people will flog to see something that is masterful and make them feel genuine human emotions.
AI is a tool like any other and while I’m certain some future creative master will use it to make something truly stunning, without human creativity and innovation it will just churn out some boring slag that while vaguely useful nevertheless make us feel kinda bored and empty inside - kinda like my polyester sweatpants.
I like the comparison with plastics: it is often only used because it is cheaper than the alternatives with worse results. Also there’s a pollution problem.
> John Waters could conjure up an intense feeling of kitsch by just naming his film Polyester.
Looking forward to scratch-n-sniff Odorama GenAI!
https://www.youtube.com/watch?v=k3WPpbEIYSs
https://www.youtube.com/watch?v=2B2HE-e8wYY
That seems like a bad analogy, and reasoning by analogy is usually not helpful because no new technology is like what came before it. And there is no backlash for polyester, it still exists everywhere and is part of most fabric mix. If anything Polyester has won by being invisible.
> The stigma against polyester persists even now.
Not in active wear / sports clothing!
Oh, look: https://news.ycombinator.com/item?id=31114188
cheap clothing is still sold..
Yup, Shein etc use a lot of polyester and polyester blends because it's cheap.
Genai could, in principle, evolve/change to a point where people cannot detect it, rendering the entire argument flawed.
So a metaphorical soft plastic whose external poisoning costs at almost every stage of existence are delusionally externalized.
Nailed it.
Polyester solved minor issues in fabrics and was generally less comfortable and less odor-resistant than cotton. AI, by contrast, is dynamic and steadily removing pain points with no serious competition from non-AI solutions. The title is catchy, but the comparison doesn’t hold up.
Polyester is still used extensively but it isn’t the replacement industry wanted. Folks overdid it and found a good balance.
We will overshoot with GenAI and over use it. Eventually rolling back and finding a better balance.
Lewis Mumford had an essay about this - how we don’t need turn the speakers up to 10 just because we can.
> Polyester is still used extensively but it isn’t the replacement industry wanted
This is almost always the case. It's very rare that an improvement is better than the original in everything.
The new thing is always sold as universally better, and some times it takes a while for the mass consciousness of people to figure out that the new is better in some respects but the old is still better in others.
Maybe not but it led to much higher quality synthetic and hybrid clothing. It's insane how much cheaper and better clothing is than even 20 years ago.
I disagree with you on AI, the results are clearly poor.
I agree on polyester, though. I sweat and itch wearing polyester clothing, even relatively low blends. Maybe not an issue for everyone, but to me it’s clearly an inferior fabric.
https://newsocialist.org.uk/transmissions/ai-the-new-aesthet...
I like the polyester metaphor, but this phrase:
"Our era's particular neoliberal hyper-connected, hyper-capitalist economy is creating a lot of profit for a few people.."
I feel is off base: the last 20 years has seen the absolute carpet-bombing of the very definition of American middle class. An entire new layer of workers (tech workers in their tens upon tens upon tens of thousands) make salaries that blow the old-trend / old-trajectory middle class out of the water.
I understand that the modern tech world has created billionaires in astonishing numbers, but I dislike skipping over the part where vast numbers of people in America are making in some cases full-on double what a without-them income trajectory would see for our middle class. FAANG salaries aren't representative - I get it. But those FAANG salaries are being paid to vast numbers of people - and salaries that what I'll call a "conventional" middle class person blanches at.
Tech cities are experiencing massive inflation in no small part due to this entirely new explosion of what we have to now call the new "upper middle class," at a scale that is unprecedented in the moderate-range past.
Tech has created a wild number of billionaires - I understand that. But it has also created an entirely new, and substantially large, additional income-demographic slice that didn't exit much before.
So to disagree with the author, tech has created unprecedented wealth for a very large number of people.
The class of well-paid FAANG employees is not that large: I'd guessestimate that all five combined employ ~100,000 engineers in the US, give or take. Yes, they have many more employees in total, but many of these are not well-paid engineering roles and/or are not in the US.
If we generously assume that there are 3x as many engineers employed at various non-FAANGs that pay competitive salaries, we still only get ~400,000 people, or around 0.1% of the population of the US.
There's one, maybe two major cities in the US where this has visibly impacted the entire local market: SF and Seattle. Everywhere else tech employment is more or less a rounding error.
Just to confirm I'm not pulling these figures entirely out of my ass, Pew Research says the "upper-income tier" has grown from 14% in 1971 to 21% in 2021 (+7%), but the "middle class" (median income $90k in 2021 dollars) has shrunk from 61% in 1971 to 50% (-11%) in 2021.
https://www.pewresearch.org/short-reads/2022/04/20/how-the-a...
So the [middle + upper] tier has shifted from 75% to 71%, for a 4% net drift downward? While 7% have drifted upward from [middle + lower] classes to upper?
The Pew link is fascinating me, thank you for sharing it! One thing that stands out to me is just how extreme the upper-class skew is—its median being 10x the median lower-class income. That helps me contextualize the types of graph that describe the “decline of the middle class” as a share of total wealth: it makes it seem more like a story of “lots of middle-class families’ wealth is wealthy in global terms and also growing, but a shocking number of families have shockingly much money too.” To some degree a story of middle class decline relative to the people richer than them—but to an even greater degree, a story of a burgeoning upper class.
Certainly you’re right that highly-paid tech workers are a drop in that bucket. To me, though, the overall numbers frustrate the “decline of the middle class” narrative to some extent. I guess it’s considerably worse to sink from prosperous classes to lower-class than it is to rise from the middle to the upper class—still, the net slice of the population affected by the change is smaller than I would have guessed.
It's also visible in Austin, your broader point notwithstanding.
>An entire new layer of workers (tech workers in their tens upon tens upon tens of thousands) make salaries that blow the old-trend / old-trajectory middle class out of the water.
You make a good point. Tens of thousands is a huge number, that’s gotta be like what, most people? Probably at least half of the total number of people.
Now I personally don’t have the time or attention span to count all of the people but I don’t reckon that the number of them is so big that “tens of thousands” doesn’t make up an enormous percentage of everybody. There would have to be so many people that I wouldn’t be able to picture them all, which is preposterous, so tens of thousands has to be astronomical in the grand scheme of things
It's all relative.
It has made a LOT for a few people and sure a lot for many. The pay increases exponentially up the corporate hierarchy. Maybe, the author is trying to highlight that.
Agreed. Depends on your definition of wealth, but there are a lot of people who are wealthier because of tech. Even the aircon installers going around replacing analogue controls with digital are making bank.
The question of whether these developments are a net positive for meritocracy seems unclear. It's tho, nothing compared to crypto-- many got propelled into a different class whom I have trouble wanting to give my precious attention to...
(I also get the robotech ref, thanks for the nostalgia!)