I don't think AGI is right around the corner

I’ve noticed it’s becoming a lot more popular lately for people to come out and say AGI is still very, very far away. Is the hype cycle ending somewhat? Have we passed peak LLM?

Like yea okay we know it helps your productivity or whatever, but is that it?

Maybe - anecdotally, HN at least is not very receptive to the idea that transformers are not (or with more data will never be) sentient somehow, and almost every post I see about this is followed up by the idiotic "how do we know human intelligence isn't the same thing", as if there's some body of commentators whose personal experience with consciousness somehow leads them to believe it might be achievable with matrix math.

Anyway, I don't think we're over the peak yet, the tech adjacent pseudo intellectuals that feed these bubbles (VCs etc) still very much think that math that generates a plausible transcript is alive.

> experience with consciousness somehow leads them to believe it might be achievable with matrix math

That's trivially true if you subscribe to materialism; QM is "just matrix math".

You're not making the point you think you're making.

Patel isn't saying that AGI is still very, very far away. He says his best estimate is 2032 and "ASI" in 2028 is "a totally plausible outcome".

(He thinks it might be quite a long way away: "the 2030s or even the 2040s", and it seems to me that the "2040s" scenarios are ones in which substantially longer than that is also plausible.)

>QM is "just matrix math".

Err no. You can solve QM without using matrices. Matrix math is just a tool.