back to index

Timeline for AGI: 2030 with 50% chance


Whisper Transcript | Transcript Only Page

00:00:00.000 | You've estimated that we'll have AGI by 2030.
00:00:03.020 | How will we actually know that we got there?
00:00:06.300 | And what may be, quote, move 37 of AGI?
00:00:11.260 | So I think there's the sort of blanket testing
00:00:13.820 | to just make sure you've got the consistency.
00:00:15.900 | But I think there are the sort of lighthouse moments
00:00:20.220 | like the move 37 that I would be looking for.
00:00:23.140 | So one would be inventing a new conjecture
00:00:26.460 | or a new hypothesis about physics like Einstein did.
00:00:30.180 | So maybe you could even run the back test of that very rigorously,
00:00:33.380 | like have a cutoff of knowledge cutoff of 1900
00:00:36.220 | and then give the system everything that was, you know,
00:00:39.480 | that was written up to 1900 and then see if it could come up
00:00:43.060 | with special relativity and general relativity, right?
00:00:45.380 | Like Einstein did.
00:00:46.160 | That would be an interesting test.
00:00:47.780 | Another one would be, can it invent a game like Go?
00:00:51.800 | Not just come up with move 37, a new strategy,
00:00:54.760 | but can it invent a game that's as deep, as aesthetically beautiful,
00:00:57.680 | as elegant as Go?
00:00:59.140 | And those are the sorts of things I would be looking out for.
00:01:02.160 | And probably a system being able to do several of those things, right?
00:01:05.700 | For it to be very general, not just one domain.
00:01:08.180 | And so I think that would be the signs,
00:01:10.440 | at least that I would be looking for,
00:01:12.140 | that we've got a system that's AGI level.
00:01:14.920 | And then maybe to fill that out,
00:01:16.760 | you would also check the consistency,
00:01:18.340 | you know, make sure there's no holes in that system either.
00:01:21.900 | And you would be in the room for that probably two or three months
00:01:24.920 | before announcing it.
00:01:26.440 | And you would just be sitting there trying not to tweet.
00:01:30.160 | Something like that.
00:01:32.280 | Exactly.
00:01:33.000 | It's like, what is this amazing new, you know, physics idea?
00:01:36.340 | And then we would probably check it with world experts in that domain,
00:01:39.700 | right, and validate it and kind of go through its workings.
00:01:43.900 | And it would be explaining its workings to,
00:01:46.280 | yeah, be an amazing moment.
00:01:47.640 | Thank you for watching this clip.
00:01:50.920 | Please subscribe to the Lex Friedman YouTube channel
00:01:53.700 | and consider watching the full episode of the podcast.