[stale2002]

> Try to do something significant with the pirated content, as the AI startups do, and you'll immediately get trouble.

No they won't. 80% of the people I meet at tech events either have an AI startup or are working at one. Basically all of them are doing legally questionable stuff, when it comes to copyright law.

Anyone who has ever trained a model, or finetuned a model, is likely using other people's data. This is universal. It is almost everyone.

And yet, despite this being a gray area, basically nobody is getting into trouble for this. We are all getting away with it. And there is only a singular lawsuit about this against like 4 companies and nobody else. (midjourney, runwayML, Stable diffusion, and deviant art)

So the point stands. Almost nobody is getting into trouble, despite this behavior being widespead everywhere, and if you don't do the same thing then you are going to fall behind and fail.

[lucianbr]

> Anyone who has ever trained a model, or finetuned a model, is likely using other people's data. This is universal. It is almost everyone.

Citation needed.

The whole reason there's no legal trouble is that it cannot be proved conclusively that they used copyrighted data for training.

"when the CTO of OpenAI is asked if Sora was trained on YouTube videos, she says “actually I’m not sure” and refuses to discuss all further questions about the training data". Why do you think she's denying the obvious?

[stale2002]

Just look around at all the AI companies that are doing this, dude.

Its a lot of people.

So the point stands, a large number of people are doing this.

Meaning that, yes a lot of people are getting away with this, and only like 4 companies are being sued.

This means that the point about people immediately getting in trouble is wrong, given that a lot of people are doing this.

Can you directly address this point, instead of giving a 2 word response that doesn't really address the point, and instead tries to attack the argument on a technicality because I didn't include a 20 page research paper in my social media comment?

[lucianbr]

Honestly I'm having a really hard time arguing that regular joes are not on the same footing with megacorps as related to "legal gray areas". It's like asking me to prove the sky is blue. Whereas of course, anything you claim, is "obvious, just look around dude".

What's the claim here, that now torrenting is officially legal? People have been torrenting, there's no consequences, it's been hammered out, now it's a-ok?

[stale2002]

> What's the claim here

The claim is that people use models like stable diffusion, right now. Stable Diffusion was trained on copyrighted data. Lots of AI companies use that. And yet, they are all getting away with it. Only 4 companies are being sued for that.

Also, the claim here is that there are a whole lot of AI companies that are trained on other people's data, that are released publicly on the internet, and they are often made by regular joes.

> now it's a-ok?

People do seem to be getting away with it, seeing as there are only 4 companies that are being sued for this stuff, even though people have been using stable diffusion for 2 years now.

Do you really think that people aren't using stable diffusion? Do you believe that people are not making mario pictures, or copyrighted images? Clearly they are.

And a lot of AI startups, and regular people, train their models on data that they don't own.

> It's like asking me to prove the sky is blue.

If it obviously true, then you'd find more than 4 companies being sued over this stuff. But thats not happening.

Like, did you know that you can just google AI art stuff right now? Any of those models that pop up, right now, was trained on copyrighted data.

Stable Diffusion is used by all sorts of AI companies. And that model was all trained on copyrighted data.

Do you just think that nobody is using stable diffusion? As in, is it completely made up? Nobody is using that?

Clearly people are.

> there's no consequences

Apparently there isn't consequences because only like 4 companies have been sued in 1 lawsuit over this AI stuff.