[kelipso]

These things are just not going to go anywhere, big reason being AI is part of the technological race. If AI research gets constrained in the US, progress will happen in China. Since that can't happen, this won't go anywhere.

[TechBro8615]

I tend to agree with you, but I also recognize I could be unrealistically optimistic. This is the legal system we're talking about. I wouldn't expect every court case to be decided fairly, nor would I expect any new laws and regulations to necessarily be sensible. Frankly my biggest worry at this point is that regulatory capture from the first mover AI companies will stop me from purchasing more than one GPU.

I'm not too worried about copyright issues because regardless of whatever happens with upcoming case law and legislation, any regulation against the input data will be totally unenforceable. It's nearly impossible to detect whether or not an LLM was trained on some corpus of data (although maybe there is some "trap street" equivalent that could retroactively catch an LLM trained on data it wasn't allowed to read). And even if the weights of a model are found to be in violation of some copyright, it's still not enforceable to forbid them, because they're just a bag of numbers that can be torrented and used to surreptitiously power all sorts of black boxes. That's why I'm much more worried about legislative restrictions on hardware purchases.