Discussion
…Artists have always stolen from other artists. I have done so. In the creative world it’s one facet of what we call ‘inspiration’.
Remixing is what culture does.
The key differences in the era of degenerative LLMs are scale and speed. It’s an accelerant.
It’s a heightening of the mismatch between what capitalism wants and the grain of the internet.
Capitalism profits from controlled scarcity. The internet wants everything connected: engineered abundance
…Artists have always stolen from other artists. I have done so. In the creative world it’s one facet of what we call ‘inspiration’.
Remixing is what culture does.
The key differences in the era of degenerative LLMs are scale and speed. It’s an accelerant.
It’s a heightening of the mismatch between what capitalism wants and the grain of the internet.
Capitalism profits from controlled scarcity. The internet wants everything connected: engineered abundance
@urlyman What these models do, and what humans do, is nothing alike. We’re way more complex, and capable of innovation, these models can’t. They do statistical permutations. Feed them Van Gogh, and they can do images that look close to those of Van Gogh. But feed them the painters Van Gogh used as inspiration, and they can't generate something as a new Van Gogh. That’s the main difference. When humans do what these machines do, we’re normally accused of copying, or stealing… 😝
@dmian I think it’s fuzzy. There are parallels and differences, and they’re not clear cut.
The key thing is that when humans borrow/steal ideas from other humans that’s relational and the relations are complex and, well, human. And the speed that happens at has historically been something that culture can metabolise.
What these models do is *transactional* and in service to degenerate acceleration. In that respect, there’s a lot of overlap with crypto https://mastodon.social/@urlyman/111946915022302081
@urlyman I had to correct my post because I meant "they CAN'T generate", not "they CAN generate". My mistake, sorry... 😝
My point is: these models can't innovate. They can do permutations on what they're fed. But nothing novel comes from them.
Important thoughts on what genAI / LLMs do to us, thank you.
Love the term "degenerative LLM"!
I'm still looking for the right hashtags to identify this strand of reflection. #noAI#noLLM#antiAI#antiLLM#antiGenAI#degenerativeAI#degenerativeLLM ?
I have a few I recycle:
Large Larceny Model
Degenerative AI
Lack of care economy
I may have others I can’t remember rn! I’m a medium-sized excoriation model
…We can discern the following probable facets:
‘Creators’ rent degenerate AI from Spreadshirt. The latter will take a cut from all sales they make.
Spreadshirt rent the underlying LLMs from OpenAI or equivalent. The latter take a cut from every ‘token’ generated.
And the whole edifice, including the people buying the shirts, have a token attentiveness to the impacts.
The rentier economy, with both tenants and landlords steadily deconstructing the house they live in
…It’s the distance.
The distance between creator and appropriator.
The distance between appropriator and LLM engineer.
The distance between T-shirt manufacturer and printer.
The distance between the price paid for the shirt and its cost.
Connection via the web engineering disconnection at dizzying speed. Serious fucking consequences with barely any attentiveness to consequences
we’re now a few years into this bullshit, and we’ve seen many, many, many different kinds of attempts at use of them. including some people in earnest trying to coax novelty from them (something that used to be better when they were “less polished” and more weird/fun as @davidgerard likes to say)
this has a direct implication as an “output cap” - the trash can only ever be _that_ good
…I am a part of this. If you think I’m throwing mud, I’m fucking caked in it
…It’s not new. It’s just MOOOOOAARRR.
At the time when what we need more than anything else is to learn to be happier with less stuff.
I did loads of T-shirt design in the 80s and 90s. It became a sideline for our Ultimate team. Our shirts were hot property.
When Oz and I started out we were painstakingly making layered acetate artwork on Grant Enlargers.
Then things sped up a bit using Illustrator on a Mac.
And now you can ask an LLM to magic stuff up in seconds
A space for Bonfire maintainers and contributors to communicate
