Is pop culture a form of "model collapse?"
Disclaimer: I do not like LLMs. I am not going to fight you on if you say LLMs are shit.
One of the things I find interesting about conversations on LLMs is when have a critique about them, and someone says, "Well, it's no different than people." People are only as good as their training data, people misremember / misspeak / make mistakes all the time, people will listen to you and affirm you as you think terrible things. My thought is that not being reliably consistent is a verifiable issue for automation. Still, I think it's excellent food for thought.
I was looking for new music venues the other day. I happened upon several, and as I looked at their menu and layout, it occurred to me that I had eaten there before. Not there, but in my city, and in others. The Stylish-Expensive-Small-Plates-Record-Bar was an international phenomenon. And more than that, I couldn't help but shake that it was a perversion of the original, alluring concept-- to be in a somewhat secretive record bar in Tokyo where you'll be glared into the ground if you speak over the music.
It's not a bad idea. And what's wrong with evoking a good idea, especially if the similarity is just unintentional? Isn't it helpful to be able to signal to people that you're like-that-thing instead of having to explain to people how you're different? Still, the idea of going just made me assume it'd be not simply like something I had experienced before, but played out and "fake." We're not in Tokyo, and people do talk over the music. And even if they didn't, they have silverware and such clanging. It makes me wonder if this permutation is a lossy estimation of the original concept, just chewed up, spat out, slurped, regurgitated, and expensively funded.
other forms of conceptual perversion:
- Matters of Body Image - is it a sort of collapse when we go from wanting 'conventional beauty' to frankensteining features onto ourselves? Think fox eye surgeries, buccal fat removal, etc. Rather than wanting to be conventionally attractive, we aim for the related concept of looking like people who are famous.
- (still thinking)
I guess this is analogous to your example, in the sense that a copy of a copy of a copy has produced a poor-quality imitation of the original. The difference is, people aren't only as good as their training data. So we're equally capable of taking an idea and improving on it, innovating new things from its core concepts etc.. That's how the original, alluring concept of The Stylish-Expensive-Small-Plates-Record-Bar was created in the first place.
So with humans, the idea degrades if it's copied badly and improved if it's copied well. Presumably this is because a person can draw on their own experiences and feelings to contextualise the concept, identify problems and innovate solutions. Whereas an AI has no concept of the real world it's representing, it has to rely on its training data to stay "tethered" to a sensible concept space.
I agree about the human portion - humans aren't just photocopiers, they modify and iterate; they see what works and what doesn't; and they try to improve. Right now, AI is basically a photocopier that can collate different ideas together, but they are getting better as well.
It's not out of the question for an AI to be able to iterate and improve upon an idea. Where they currently fall short is that they simply haven't contextualized the "experience" of the world correctly, so sometimes their suggested "improvements" get stuck in a loop of switching back and forth between two states, or go off in a weird and unexpected direction. But I think that human-adjacent contextualization is coming, and I wouldn't be surprised to see it within a year. They already have all of the information, they just need to be pushed in just the right way to "figure it out".
No it's not out of the question, in fact collating ideas together can arguably be an improvement itself. The problem is that the output isn't constrained to things that make physical practical sense in the real world, so even small absurdities in an output, if fed into the training data, will accumulate over generations like feedback in speakers. And AI do not have "all the information". Quick example, humans get all sorts of information from our senses that don't fit into the text-only data an LLM is working with.
Why is this presented so negatively? It’s not like Japan had western-style bars to begin with that idea was copied from American influence during the American occupation. If it weren’t for the transmission and evolution and development of ideas across cultures, you’d just have Izakaya-style establishments in Japan.
All ideas are remixes, and that’s fine.
That's a fair point, all ideas start somewhere!
As for the negativity, I supposed it's unearned as I would still totally go to one of these aforementioned places to give it a shot. I think the negativity from me comes in part from my last experience at one of these, and finding it more like a pretty expensive restaurant with a so-so DJ (obviously subjective), rather than a place where music was really centered. It had an impressive look and presentation, but I think it mislabeled itself when calling itself a "listening bar."
Hmm, this is something I've been thinking about for a while now.
I saw a leftist art crtic on YouTube (Morbid Zoo, I think?) essentially say that our culture is stagnant, that nothing new is being created. I've been exploring this idea, writing and rewriting a piece called "The End of Art History."
Their argument is basically that all art now is post-something, postmodern, poststructural, whatever. All it is is engagement with what came before -- through allusion or through metatext or adaptation, essentially all art that is now produced is in conversation with what came before, or else blandly recreating and repackaging it. The Lion King is just Hamlet. Limbus Company is just The Divine Comedy meets The Stranger. Every contemporary work of surrealist art is just cribbing from Lynch.
I don't really think I agree with this depicition of modern culture. The feeling of cultural stagnation is certainly facilitated by what appears to be a cessation of innovation from capitalists, be it remaking movies or essentially rereleasing the same iPhone every year. Nothing is new, nothing is fresh, nothing is changing. But I think it's a naive view of history and culture to narrativize things in this way. Maybe it feels like culture is stagnant now, but it will not feel that way for long. The upheaval we are experiencing -- a global rise in neofascism; labor alienation and disempowerment through AI -- might give rise to new art, change culture, at once sharpen the divide between people and create new ground for us to walk on. Or these trends might fizzle, and new ones emerge to have that transformative effect. But human societies have all, at all times, existed in a process of transformation, of nonlinear, unpredictable change, governed by the fluid dynamics of amassed human ideology. If things seem stagnant now, it's only because you're ignoring the people standing at the bank throwing pebbles; ignoring the social shifts, the outsider art movements, the emergent changes in global thought that will eventually metastasize, and render the world unrecognizable.
Everything seems like a regurgitation of something that came before, yes, but they were saying this back in Biblical times. King Solomon wrote "There is nothing new under the sun." I think we inherit, from the way history is taught, the childish idea that change happens suddenly, when great men, or great societies, suddenly have the idea to produce something radical, new, miraculous. But in reality, I think every "new" idea that has changed the world: industrialization, Communism, American democracy, was assembled out of things that came before it; was then, later, co-opted and transformed into something else. This is how society has always evolved, and likely, always will.
That's because it's what the people are buying. Franchises are big business, and many people will only get into something they're already familiar with.
Why does one of my friends only buy Godzilla coffee and soap with Marvel characters? Companies make things that people want to buy.
If we wanted to shake things up, we'd eliminate or massively reduce intellectual property. This would encourage more innovation and allow artists to do more than pay homage or draw inspiration. They could write their own stories with familiar characters and evolve mediums.
I've seen the same video. One important aspect to keep in mind is the amount of media that is made as explicit nostalgia bait first and entertainment for the current generation of kids second.
I can't resist a cheap shot at unconstrained capitalism. Innovation is risky, so it makes sense that capitalists will only innovate where there's also a high opportunity for reward. Art is probably one of the domains most challenged by this, because (at least, this is my thinking) highly artistic value requires originality. Highly economic value can often be very unoriginal and formulaic. It seems like a side effect of the design that capitalists operating only on capitalist values will very rarely contribute meaningfully to art.
I realize that these aren't mutually exclusive, but I have the feeling lately (purely non scientific) that modern businesses are becoming more cynical about artistic value as purses around the world go tighter. I'm not sure art was ever meant to be big business, anyhow, so maybe it's a good thing.
You might be interested in https://en.m.wikipedia.org/wiki/The_Death_of_the_Author by Roland Barthes.
His essay is tough to summarize, but he talks a bit about nothing is really new and everything is simply a product of its time.
Had to look that one up. good source material for soap I’ve been told.
This part at least is a feature, not a bug. Self control and/or wealth are adaptive traits for mate attraction in large swaths of the here and now and that’s kinda how biology works.
As far as the other issue, there’s the famous pithy quote, “good artists copy, great artists steal.”
Also, there’s nothing wrong in my mind with drawing inspiration from stuff in high cultural centers to model in more low-key, local environments.
There's a science fiction book set in the future where there are teleporter disks that can take you to any city on Earth in a few seconds. This wasn't the main plot of the book but still the idea stayed with me. I'm pretty sure the book is Ringworld by Larry Niven. If so, this is the item, Stepping Disk.
Anyway the book mentions that once these were created it ruined the uniqueness of different cities. Paris and Rome and Beijing all would become the same because there is no barrier to moving people or things between them quickly.
Maybe the internet is doing that to culture. It's nice that people from different cultures can so easily share and talk to each other, but it does homogenize the differences a lot. Also, it doesn't seem to have reduced conflict between cultures, but that's being done intentionally by evil people who get power and money from conflict.
Copying ideas is central to human culture. Sometimes they’re bad imitations. Sometimes they’re just different, or even an improvement. I don’t think calling it “model collapse” adds anything?
Sometimes people talk about memes or cultural evolution, which uses biology as metaphor.