[zamalek]

> t's not impossible that even an LLM, which totally isn't designed with the goal of having it, might nevertheless have it.

I was referring to the idea that a brain scan of a human could be copied into an LLM.

[ben_w]

Oh, in that case I agree. My understanding is LLMs (or at least GPT-architectures) aren't Turing machines, so they can't simulate arbitrary other systems even if you made them very big, and because of this my guess is that it would be extraordinarily unlikely for them to just happen to have the right shape and power for simulating a full human brain.