‘Impossible’ to create AI tools like ChatGPT without copyrighted material, OpenAI says

There's 2 general groups when it comes to AI in my mind: Those whose work would benefit from the increased efficiency AI in various forms can bring, and those who want the rewards of work without putting in the effort of working.

The former include people like artists who could do stuff like creating iterations of concept sketches before choosing one to use for a piece to make that part of their job easier/faster.

Much of the opposition of AI comes from people worrying about/who have been harmed by the latter group. And it all comes down the way that the data sets are sourced.

These are people who want to use the hard work of others for their own benefit, without giving them compensation; and the corporations fall pretty squarely into this group. As does your comment about "small-time creators who suddenly have custom graphics to illustrate their videos, articles, etc." Before AI, they were free to hire an artist to do that for them. MidJourney, for example, falls into this same category - the developers were caught discussing various artists that they "launder through a fine tuned Codex" (their words, not mine, here for source) for prompts. If these sorts of generators were using opt-in data sets, paying licensing fees to the creators, or some other way to get permission to use their work, this tech could have tons of wonderful uses, like for those small-time creators. This is how music works. There are entire businesses that run on licensing copyright free music out to small-time creators for their videos and stuff, but they don't go out recording bands and then splicing their songs up to create synthesizers to sell. They pay musicians to create those songs.

Instead of doing what the guy behind IKEA did when he thought "people besides the rich deserve to be able to have furniture", they're cutting up Bob Ross paintings to sell as part of their collages to people who want to make art without having to actually learn how to make it or pay somebody to turn their idea into reality. Artists already struggle in a world that devalues creativity (I could make an entire rant on that, but the short is that the starving artist stereotype exists for a reason), and the way companies want to use AI like this is to turn the act of creating art into a commodity even more; to further divest the inherently human part of art from it. They don't want to give people more time to create and think and enjoy life; they merely want to wring even more value out of them more efficiently. They want to take the writings of their journalists and use them to train the AI that they're going to replace them with, like a video game journalism company did last fall with all of the writers they had on staff in their subsidiary companies. They think, "why keep 20 writers on staff when we can have a computer churn out articles for our 10 subsidiaries?" Last year, some guy took a screenshot of a piece of art that one of the artists for Genshin Impact was working on while livestreaming, ran it through some form of image generator, and then came back threatening to sue the artist for stealing his work.

Copyright laws don't favor the small guy, but they do help them protect their work as a byproduct of working for corporate interests. In the case of the Genshin artist, the fact that they were livestreaming their work and had undeniable, recorded proof that the work was theirs and not some rando in their stream meant that copyright law would've been on their side if it had actually gone anywhere rather than some asshole just being an asshole. Trademark isn't quite the same, but I always love telling the story of the time my dad got a cease and desist letter from a company in another state for the name of a product his small business made. So he did some research, found out that they didn't have the trademark for it in that state, got the trademark himself, and then sent them back their own letter with the names cut out and pasted in the opposite spots. He never heard from them again!

The Internet Archive has no ground to stand on at all. It would be one thing if they only allowed downloading of orphaned or unavailable works, but that’s not the case.

Wikipedia has had bots writing articles since the 2000 census information was first published. The 2000 census article writing bot was actually the impetus for Wikipedia to make the WP:bot policies.

It is free. They have a pair model with more stuff but the baseline model is more than enough for most things.

Of course it sounds bad when you using the word "steal", but I'm far from convinced that training is theft, and using inflammatory language just makes me less inclined to listen to what you have to say.

What if I do it myself? Do I still need to get permission? And if so, why should I?

I don't believe the legality of doing something should depend on who's doing it.

One name not a fucking click bait article. I want one single name of the artist who is now on food stamps because openai trained their model on their art.

I only discuss copyright on posts about AI copyright issues. Yes, brilliant observation. I also talk about privacy y issues on privacy relevant posts, labor issues on worker rights related articles and environmental justice on global warming pieces. Truly a brilliant and skewering observation. Youre a true internet private eye.

Fair use and pushing back against (corporate serving) copyright maximalism is an issue I am passionate about and engage in. Is that a problem for you?

now

It's called "machine learning", not "AI", and it's called that for a reason.

"AI" models are, essentially, solvers for mathematical system that we, humans, cannot describe and create solvers for ourselves, due to their complexity.

For example, a calculator for pure numbers is a pretty simple device all the logic of which can be designed by a human directly. For the device to be useful, however, the creator will have to analyze mathematical works of other people (to figure out how math works to begin with) and to test their creation against them. That is, they'd run formulas derived and solved by other people to verify that the results are correct.

With "AI" instead of designing all the logic manually, we create a system which can end up in a number of finite, yet still near infinite states, each of which defines behavior different from the other. By slowly tuning the model using existing data and checking its performance we (ideally) end up with a solver for some incredibly complex system. Such as languages or images.

If we were training a regular calculator this way, we might feed it things like "2+2=4", "3x3=9", "10/5=2", etc.

If, after we're done, the model can only solve those three expressions - we have failed. The model didn't learn the mathematical system, it just memorized the examples. That's called overfitting and that's what every single "AI" company in the world is trying to avoid. (And to do so, they need a lot of diverse data)

Of course, if instead of those expressions the training set consisted of Portrait of Dora Maar, Mona Lisa, and Girl with a Pearl Earring, the model would only generate those tree paintings.

However, if the training was successful, we can ask the model to solve 3x10/5+2 - an expression it has never seen before - and it'd give us the correct result - 8. Or, in case of paintings, if we ask for a "Portrait of Mona List with a Pearl Earring" it would give us a brand new image that contains elements and styles of the thee paintings from the training set merged into a new one.

Of course the architecture of a machine learning model and the architecture of the human brain doesn't match, but the things both can do are quite similar. Creating new works based on existing ones is not, by any means, a new invention. Here's a picture that merges elements of "Fear and Loathing in Las Vegas" and "My Little Pony", for example.

The major difference is that skills and knowledge of individual humans necessary to do things like that cannot be transferred or lend to other people. Machine learning models can be. This tech is probably the closest we'll even be to being able to shake skills and knowledge "telepathically", so to say.