Comment by raincole
10 days ago
> I can train insainly powerful models on my laptop.
What's with this hallucination? The thread is about GPT-5.6. Your laptop can't even run gemma 4 unquantized bfloat16, which is light years behind GPT-5.6, and running it is light years from training it. If something that a laptop can train is insanely powerful for you, you don't need to worry about this thread at all.
Indeed every AI enthusiast has to find within their heart their consistent position between the two extremes:
- Regulation like this is dangerous because the accessibility and capability of open-source AI training and inference are limited;
- Regulation like this is hollow because the accessibility and capability of open-source AI training and inference are unlimited.
Regulation like this is dangerous for the US because it can kneecap our country, but I don’t think it’ll affect the pace of innovation globally. China isn’t going to slow down just because the Trump administration asked for some “ballroom donations” or whatever nonsense is going on behind the scenes here.
Eh, they said models, which can mean hyperspecific domains, not necessarily a massive generalized LLM.
Has the US government limited access to any non-LLM ML models?