[fstokesman]

> ...remote biometric identification systems in publicly accessible spaces, biometric categorisation systems (e.g. categorizing by gender, race, ethnicity, citizenship status, religion, political orientation) and the use of AI for predictive policing.

> AI systems which can influence voters in political campaigns and by use of suggestion systems on very large platforms...

> New transparency and risk assessment requirements for providers of (generative) foundation models like GTP.

> Clarified exemptions for research.

Putting these kinds of restrictions in place is absolutely a good thing. While they might not get everything right, this is a step in the right direction. Our laws and understanding as a society has been lagging behind technological development for decades now. That fact has enabled a large amount of exploitation to take place, which has (in the last decade especially) had a large hand in massively undermining our democracies.

[logicchains]

>New transparency and risk assessment requirements for providers of (generative) foundation models like GTP

This is absurd. For a relatively small sum in the grand scheme of things, I could rent a few A100s, download a free dataset and train a model like LLaMA 30B, which is comparable to GPT3 (and indeed there already are such efforts popping up). Such a law could potentially make it illegal to upload such a thing if you live in the EU without going through a potentially expensive and bureaucratic process. It will completely stifle AI development the same way requiring people to going through a bunch of paperwork to upload a new library would stifle web development.