[sixtyj]

> The rulebook sets out rules for all models that lack a specific purpose but can be adapted to a variety of use cases, requiring transparency on how a model was built, disclosure of any copyright-protected content used for training, and enough information for downstream users to understand the model's capabilities.

Re. disclosure: How do they want to do it? It seems to be similar to a “disclosure” used in students’ works: “Source: internet”…

[m4rtink]

Seems quite sensible - good to see an institution woth some weight finally enforcing such basic principles that should have been universal - but money is of course more important for AI companies, making such regulation necessary.

[Retr0id]

No copyright intended

[ArekDymalski]

> Re. disclosure: How do they want to do it? It seems to be similar to a “disclosure” used in students’ works: “Source: internet”…

I think enforcing compliance with the law will be rather simple, just like it happened with GDPR, food labeling and many other regulations. But I wonder how will the disclaimers/declarations even be verified? The more I think about it the more I see open sourced (both dataset and weights) models as the only truly verifiable solution. And that makes it less attractive as a business foundation if. So we might end up with state-funded models working in similar fashion to museums it other culture/art institutions ensuring that original material creators are treated fair. But that will bring whole other set of challenges...

[sixtyj]

Do you think that states have enough skillful ppl to run such a comparison model?

I remember European efforts to create search engine, digital library. And a lot of EU-funded projects in Horizon 2000 are useless, just an expensive bureaucracy…

GDPR and cookie consent bars are pain in the a* and ux/ui failures.

What I mean - intention is good, realisation is often bad. (But I don’t think that rest of world would be better btw :)

/end-of-rant