Why do you think a UN-hosted process would work? Has the UN done similarly intrusive things before and been effective at it? How do you account for Chinese frontier and open weights models, which are just months behind American models and subsidized by a sovereign state that does not share any of these premises about intellectual property?
A lot of these posts seem subtextually premised on the idea that it's possible to put the genie back into the bottle; that if frontier labs in America didn't sign on to this tolling scheme, our recourse would be to halt the progression of AI completely. But that option does not exist, unless we're going to fight a world war to create it.
Well, I mean, the alternative is for a US-based process to be put into place, followed (or led) by an EU process. The premise of my comment is that no matter who hosts the process, China isn't going to comply, so it really just comes down to the US and Europe.
And the US isn't going to comply either. Just like it never recognized the International Criminal Court in The Hague. In that respect it's at least as bad as China.
The only jurisdictions where training data is going to matter are in the US, Europe, and China. I think we're just talking past each other. My subtextual point is that none of this is ever going to happen.
A lot of these posts seem subtextually premised on the idea that it's possible to put the genie back into the bottle; that if frontier labs in America didn't sign on to this tolling scheme, our recourse would be to halt the progression of AI completely. But that option does not exist, unless we're going to fight a world war to create it.