> It’s not impossible, obviously—humans do it

It's still not at all obvious to me that LLMs work in the same way as the human brain, beyond a surface level. Obviously the "neurons" in neural nets resemble our brains in a sense, but is the resemblance metaphorical or literal?