Yaaay!

The problem is that CSAM is also in LAION (used for Stable Diff) training dataset and in others too most likely.

Aaaand we have the same issue today with how genAI companies source their training material! Let's just hope it wont take 14 years to make the obvious decision!

It should be illegal for companies to sell your informations at all. But wait, selling personal information is already illegal just as selling copyrighted ones from which you have no rights to either! You'll need some kind of license for that and these needs an actual signed arrangement with the individual right holder! ToS are not considered as such! We don't really need new regulations or laws, we need to enforce the ones that already exist and cover most of these cases!

And it actually is a thing, at least in EU. Machine readable opt-out needs to be respected per the EU copyright directive, this include robots.txt but also ToS and metadata.

EU directives require any and all machine readable opt-out to be respected. It include robot.txt of course but also ToS.

Well, strip mining people of their data and work without consent while disregarding existing legal framework is already wreaking havoc.

Machine and humans are not working the same way this argument is tiresome. LLM need to copy their training data. Also, on top of preventing competition in the same market, the doctrine of fair use does not exist outside US and I doubt that openAI only used US based materials.

What about respecting people's ownership over their data/work first?

I'll have to disagree with you on that. While you raise good points, I do believe that regulations that protect small businesses and hobbyist are good, not perfect but good. A lot of the people impacted by this data gluttony are exactly these, small businesses, freelances, hobbyist, independents, etc. Impacted by monopolies in a bid to replace them. You take the side of AI entrepreneurs without acknowledging that the tech and the way it is deployed right now is relying on lawless exploitation of others. We already have a legal framework that is supposed to prevent this kind of harms, yet most if not all companies and startups decided to widely ignore it in the pursuit of profit and dominance in an emerging market. If your business rely on breaking laws and exploiting people in the hope that you might change the regulations later on, then maybe it should not exist in the first place.

..without exploiting the whole world in the process? "optimistic that AI could contribute to helping humans get along with each other." heh? This is rich, Sam don't care about others.

So now instead of the "Don't text while driving" we'll have the "Don't talk to your car while driving" campaigns, truly an improvement!

Great, so it'll be able to hallucinate parts and procedures that don't exist and lead you to damage your vehicle even more!

They're not looking to protect the underlying information but the articles/literary work that has been scraped indiscriminately and that's a good thing. As much as I don't really enjoy monolithic corporations, the way "AI" models are done currently is atrocious for human culture and privacy.

That "Machines learn just like humans" argument is getting old and has been debunked numerous times