Firstly - we start off by listing several ways that AI could be used for various creative pursuits, focused mainly on analysis. Yet, beyond the anecdote of “feeding AI samples of my own work, I can generate style guides that highlight my quirks and recurring themes”, this isn’t pursued any further. I’m certainly curious about this, but the author doesn’t go into any further detail about how to actually use tooling for this.
Unfortunately, after that first section, I land in this regrettably common trope of “person A (who likes AI) tries to convince person B (who doesn’t like AI) that AI is useful, completely missing the issues that person B has with AI”. Personally, my issues are all about the ethics of training models, which the author completely glosses over:
“I have a well-founded and well-tested belief that the training for these models isn’t stealing in a legal sense”
So it’s stealing in a non-legal sense, then? Or what?
It seems obvious to me that the current crop of generative AI is only as impressive as it is because of the breadth and quality of the training set, which in large part (if not totality) comes from uncredited and uncompensated labour from the very artists who the companies selling access to the model are attempting to undercut and obsolete – and the undercutting is already happening.
Call me “the chorus” all you want. Yeah, “the people pushing this stuff early on were assholes”, but the people pushing it now are assholes too.
It seems obvious to me that the current crop of generative AI is only as impressive as it is because of the breadth and quality of the training set, which in large part (if not totality) comes from uncredited and uncompensated labour from the very artists who the companies selling access to the model are attempting to undercut and obsolete – and the undercutting is already happening.
The thing that I’ll point out is that it’s incredibly unlikely that those creators are going to be any better compensated under some other operating regime: the alternative to a stupid anime diffusion for a blog post isn’t that I’m going to commission a piece, and the alternative to a silly little anthropomorphic dhole on a slide deck is usually just gonna be “okay fuck it, stock clipart, whatevs”. Similarly, the people who want to support an artist will both find that art and fund that artist. That isn’t changing.
The thing I think should be more widely understood, using the music industry’s attacks on post-scarcity as a guide, is that there will be a few companies stepping forward to agitate for copyright enforcement, and then coincidentally selling access to those datasets with some kind of “reasonable split” with the authors of the content. You need only look at how good Spotify and similar schemes have worked out for the independent and small musicians, whose main purpose is to provide a veneer of legitimizing outrage.
In my opinion, we’re better off normalizing the use of generative AI, making it a valid option that isn’t just to the benefit of the large companies with legal departments who can exercise regulatory capture, and setting a societal expectation that slop is low art but you can totally pay somebody to make good art instead! Remixes are okay! Sampling is okay! Transformative work and curation is okay!
The alternative, of course, is the ghoulish creatorwashing by distributors and publishers to pretend that they’re sticking up for the little guy while eating all of the licensing fees for these models. It’s so painfully obvious a ploy that I’m genuinely baffled people keep advocating on their behalf.
My main disagreement on this one is the dominance of Western centric companies that are based out of the US, and China. The Arab world has a tremendous amount of interesting AI things going on and I think it will play much larger than this particular prediction story makes.
Unfortunately, I don’t know much about what’s happening outside US companies. My main thing was to persuade the American Creative Precariat to relax a little and start thinking of generative AI as a tool rather than a force coming to destroy them.
I would love to explore stuff outside my currently too-limited view though so if you have some names, projects, etc to share that would be great.
The Arab world has a tremendous amount of interesting AI things going on and I think it will play much larger than this particular prediction story makes.
I didn’t really come away with the impression that the article made much of a comment on that at all? ¯\_(ツ)_/¯ Can you clarify why you think that’s contradicting anything there?
But I’m curious to know more about that… I’m not specifically aware of a lot that has come out of the Arab AI community recently, but can’t say I’ve specifically paid attention to where a lot of what I’ve looked at originated. Are there things I’ve likely seen in this bucket? Or others you can point to?
I’m only part way through, but I appreciate that his perspective is broad in both time and space. Thinking that can see past ideology, borders. and the current moment makes all the difference.
I have a few problems with this post.
Firstly - we start off by listing several ways that AI could be used for various creative pursuits, focused mainly on analysis. Yet, beyond the anecdote of “feeding AI samples of my own work, I can generate style guides that highlight my quirks and recurring themes”, this isn’t pursued any further. I’m certainly curious about this, but the author doesn’t go into any further detail about how to actually use tooling for this.
Unfortunately, after that first section, I land in this regrettably common trope of “person A (who likes AI) tries to convince person B (who doesn’t like AI) that AI is useful, completely missing the issues that person B has with AI”. Personally, my issues are all about the ethics of training models, which the author completely glosses over:
So it’s stealing in a non-legal sense, then? Or what?
It seems obvious to me that the current crop of generative AI is only as impressive as it is because of the breadth and quality of the training set, which in large part (if not totality) comes from uncredited and uncompensated labour from the very artists who the companies selling access to the model are attempting to undercut and obsolete – and the undercutting is already happening.
Call me “the chorus” all you want. Yeah, “the people pushing this stuff early on were assholes”, but the people pushing it now are assholes too.
The thing that I’ll point out is that it’s incredibly unlikely that those creators are going to be any better compensated under some other operating regime: the alternative to a stupid anime diffusion for a blog post isn’t that I’m going to commission a piece, and the alternative to a silly little anthropomorphic dhole on a slide deck is usually just gonna be “okay fuck it, stock clipart, whatevs”. Similarly, the people who want to support an artist will both find that art and fund that artist. That isn’t changing.
The thing I think should be more widely understood, using the music industry’s attacks on post-scarcity as a guide, is that there will be a few companies stepping forward to agitate for copyright enforcement, and then coincidentally selling access to those datasets with some kind of “reasonable split” with the authors of the content. You need only look at how good Spotify and similar schemes have worked out for the independent and small musicians, whose main purpose is to provide a veneer of legitimizing outrage.
In my opinion, we’re better off normalizing the use of generative AI, making it a valid option that isn’t just to the benefit of the large companies with legal departments who can exercise regulatory capture, and setting a societal expectation that slop is low art but you can totally pay somebody to make good art instead! Remixes are okay! Sampling is okay! Transformative work and curation is okay!
The alternative, of course, is the ghoulish creatorwashing by distributors and publishers to pretend that they’re sticking up for the little guy while eating all of the licensing fees for these models. It’s so painfully obvious a ploy that I’m genuinely baffled people keep advocating on their behalf.
Well-founded and well-tested? A little early for that. Still waiting on the verdicts of some big lawsuits.
My main disagreement on this one is the dominance of Western centric companies that are based out of the US, and China. The Arab world has a tremendous amount of interesting AI things going on and I think it will play much larger than this particular prediction story makes.
Unfortunately, I don’t know much about what’s happening outside US companies. My main thing was to persuade the American Creative Precariat to relax a little and start thinking of generative AI as a tool rather than a force coming to destroy them.
I would love to explore stuff outside my currently too-limited view though so if you have some names, projects, etc to share that would be great.
I didn’t really come away with the impression that the article made much of a comment on that at all? ¯\_(ツ)_/¯ Can you clarify why you think that’s contradicting anything there?
But I’m curious to know more about that… I’m not specifically aware of a lot that has come out of the Arab AI community recently, but can’t say I’ve specifically paid attention to where a lot of what I’ve looked at originated. Are there things I’ve likely seen in this bucket? Or others you can point to?
https://restofworld.org/2025/juergen-schmidhuber-ai-saudi-arabia-tech/
I’m only part way through, but I appreciate that his perspective is broad in both time and space. Thinking that can see past ideology, borders. and the current moment makes all the difference.
[Comment removed by author]