I’m mostly down on artificial intelligence (AI), or rather machine learning (ML), mostly for environmental reasons. Like plastic, the other great synthetic technology, ML is terribly adapted to our ecology. It consumes colossal amounts of water and energy to do things a human can manage on a bowl of oats, albeit slower. Many of the most prominent boosters of ML seem to hate our ecology; this is evident from their desire to transcend it to Mars or cyberspace. They also seem ignorant or contemptuous of, for lack of a better word, our souls. By that I mean emotions, relationships, shared meanings, culture, tradition, sacrifice, sustained struggle for indeterminate spiritual rewards – the stuff that wellbeing consists of. Whether from autism or trauma or psychopathy or left-brain or whatever it may be, they seem to have instead a black hole where their soul should be that has a rapacious appetite for more. Money, power, population, bedrooms in their house, employees in their firms, whatever, just as long as there is more of it. This is mostly bad, on balance, for the creative and emotional industries – the ‘arts’ – but in ways that I think are a bit complicated.
 |
| An AI generated cover for Huckleberry Finn |
Right now, the arts are being enshitified at an astounding rate. You can’t do a google image search without turning up ML generated images in the first 30 hits. Major publishers are talking about outsourcing much of their ‘content’ generation to ML. Advertisers have replaced most of their graphic design stable with ML. Word among my colleagues in ML is that Google is in copyright talks with the major record labels because they reckon they can pump out generic pop songs of Grammy-winning quality effortlessly.
The vast majority of this content is ‘shit’. Stale, soulless,
generic, repetitive, obvious. Like Cartman’s pitches to Hollywood media executives
when he needs to maintain his cover as the robot Awesimo 4000 in South
Park, e.g. “Adam Sandler is like, in love with some girl, but then it turns out
that the girl is actually a Golden retriever”. OMG amazing! We’ll call it puppy
love! The corporate executives keep asking for more awesome movie ideas, and
Cartman keeps pitching them Adam Sandler dreck until he collapses from
exhaustion. They can’t get enough. This is from season 8 (episode 5), in 2004!
A full twenty years ago, and the script perfectly captures how most non-creative
corporate types understand, value, and use ML to generate ‘content’.
ML art is like the Netflix algorithm. Thousands of items to
choose from but somehow it just keeps giving you shit. Oh you watched Studio
Ghibli films? You’ll love these other 12 generic animes. They’re all exactly
the same – A rag tag crew of x heroes with random powers assemble to fight a
rag tag crew of y villains with random powers. The genre is so generic that one
of its biggest recent hits is a spoof, namely One Punch Man. What the algorithm
misses in Ghibli is the genius, because it’s hard to meaningfully ‘tag’ genius.
Genius almost by definition reaches beyond the status quo upon which an
algorithm is trained into something new. It create, not just regurgitate.
The issue though, before I get ahead of myself, is that most
people want familiar dreck. Among the top 10 highest grossing films of 2023
were Wonka (a remake), Fast X (10!), Guardians of the Galaxy 3, and Mission
Impossible 7: Dead Reckoning. Some of the other films were also riffs on
established tropes, namely Barbie, Mario Bros, and Spider Man: Across the
Spiderverse. Some of the biggest commercial successes in terms of revenue >
costs were Bay Boys 4 and Inside Out 2. These things can be written by ML (maybe not
Barbie and Spiderman, but ML could write most of it – the nuances were in the
execution).
Now it’s possible that people would see more films if there were genuinely fresh new hits on offer, but I’m sceptical. Good art that is also mass market, like Sex Education or Breaking Bad, has always been elusive. And corporate financiers generally lack the skills to identify it. The problem with ML is not that it floods the zone with shit because shit is what most people want. They watch a whole season of Love Island and they come back for more. Generating such dreck with ML is actually less energy intensive than having a bunch of people make it over several weeks.
The problem is that ML’s automation of shit doesn’t free up a bunch of capital to take more speculative bets on things that might be genius because you’ve still got the same soulless people in charge of green lighting. What’s more, the genius is even harder to find because it is buried under a landslide of shit.
ML is also demoralising. Why go through all the (decades of)
trouble to learn how to make art if you can just vomit some shower thoughts
into a prompt and have an image generator do it for you? The obvious answer is
that the ML images are shit. But the marginal improvement YOU can make on the
algorithm seems small compared to the effort requires to get to the point where
you have the skills necessary to make that improvement. The margin is in fact
not small at all, it just seems small when you're so far from it. The ‘margin’ that human made art has over
ML is the spark of value, of life, that humans can give to something because we
are capable of determining value for ourselves rather than just being directed
to values prespecified by an algorithm programmer.
But then again so many people don’t really seem all that interested in being valuers. They just want to get through it or something. Feel generally pleasant in life, moving from one thin event to another until they die, like an ADHD kid with a tik tock connection. (Indeed, such a person is exactly the 'marginal user' that many social media companies are trying to attract to and retain on their platforms, which is part of why they're so rapidly becoming torrents of dreck).
This is what a lot of teachers are confronting in universities.
How can I get my students to slog through to understanding something if
they can just get an ML product to do it for them? It’s like a calculator. Except
I think here it’s so much worse than a calculator because what you’re
automating away is your agency, broadly conceived. You won’t mature into a
thoughtful, responsible, values-oriented individual aware of your community,
your creed, the significance of your moment in history. You’ll be perpetually
arrested at the intellectual and moral development equivalent of breast
feeding. And instead of being fed life-giving mother’s milk you’ll be on a diet of ML generated slop. The cyberpunk comic Tokyo Ghost is prescient on these themes. Calculators lift us up towards transcendence of
the mundane, ML casts us out, or even locks us in. As numerous tweeters and skeeters have noted, we
should want ML to do the laundry so we can do the art, not the other way
around.
This all might strike you as high-minded idealism, and much
of it is, but I am not talking exclusively about high art or epochal genius.
This demoralisation will infect everything that is artisanal rather than merely
cool. ML is very good at cool – things that look or sound neat but have little
substance behind them, like 90% of what instagram’s ‘explore’ tab recommends (because
it’s an algorithm!). ML is not good at things that require a great deal of care
and attention to detail. A reverence for the minutiae. Compare something like
the Chris Hemsworth action films Extraction and Extraction 2 – thoroughly
watchable on a long-haul flight, quite possibly something that ML will be able
to make by the end of the decade – to something like the first John Wick movie.
I picked this example deliberately because so much of John Wick was about ‘cool’
and the movie was so aggressively against having a point. It just wanted to be
stylish and violent. But there was so much dedication to those values,
and dedication across a large creative team of experts. The people who
made John Wick wanted to do craft, whether of martial arts, gun violence, cinematography, or suits. And my god did they deliver. John
Wick is arguably the greatest action film of all time, certainly top 10. ML is
going to severely undermine, at least in the short term, our capacity as
individuals and communities to generate things like this that require collective craft,
even when it’s just cool stuff without any allegory.
This is exactly what Marx was bitching about when he complained of the labour of the artisans being alienated by the factories. How this will shake out I don’t know. Making wheels in factories rather than by hand was certainly an improvement in most people's quality of life in the long term. Does anyone really want to make a wheel, as a calling? I doubt it. Factory clothing has made lots of beautiful things that were previously prohibitively expensive, like lace, popularly affordable. And what remains most expensive in clothing is design. Generic design slop like Shein remains mass market, and making it next day delivery or $1 doesn’t make it any less slop. People buy shine because their lives are empty. People who go on to become brilliant clothing designers can train more cheaply on mass produced materials. Perhaps something similar will happen with ML.
I guess what I fear is that the miserable people addicted to fast fashion will be even more addicted, even more lost, and even more miserable. ML may rot our brains, rob us of motivation, and confound our sense of style and value. I think a lot will depend on the people developing ML tools and what their values are. Right now it’s a bunch of soulless creatures who seem to like the look of cyberpunk dystopias. But there is also an enormous open source movement, so who knows. I can see a future where people, freed of drudge labour by humanistic application of ML technologies, gather together in metamodern communities collaboratively generating ‘art’ – games, movies, paintings, whatever – often with the help of ML tools to cut costs and increase the speed of brainstorming. Often times they will produce something that is barely better, or even worse, than something the ML can do. After all, the algorithms will be recycling what we produce with them. But every now and then we might get Horizon Zero Dawn, or Disco Elysium, or Beyonce’s Lemonade, or some other collectively produced masterpiece of value and allegory that is thoroughly human and thoroughly new. Something truly generative and not artificial.
Comments
Post a Comment