Defining AGI

0:00 WY sat down with Dennis Sabis, who's a CEO of Google Deep Mind, which is the engine of the company's artificial intelligence. He's a Nobel Prize winner and also a knight. We discussed AGI, the future of work, and how Google plans to compete in the age of AI. This is the big interview. [Music] Well, welcome to the big interview, Dennis. Thank you. Thanks for having me. So, let's start talking about AGI a little here. Now, you founded Deep Mind with the idea that you would solve intelligence and then use intelligence to solve everything else.

0:48 And I think it was like a 20-year mission. We're like 15 years into it and you're on track. I feel like yeah, we're pretty much dead on track actually is what would be our estimate. That means 5 years away from you know what I guess people will call AGI. Yeah. I think in the next 5 to 10 years that would be my you know maybe 50% chance that we'll have what we defined as AGI. Yes. Well, some of your peers are saying 2 years, 3 years and others say a little more.

1:11 But that that's really close. That's really soon. How do we know that we're that close? There's a bit of a debate going on at the moment in the field about definitions of AGI and and then of of course dependent on that there's different predictions for when it will happen. Uh we've been pretty consistent from the very beginning and actually Shane Le, one of my co-founders and our chief scientist, you know, he helped define the term AGI back in I think early, you know, 2001 type of time frame.

1:32 And we've always thought about it as you know a system that has the ability to exhibit sort of all the cognitive capabilities we have as humans. And the reason that's important the reference to the human mind is the human mind is the only existence proof we have maybe in the universe that general intelligence is possible. So if you want to claim sort of general intelligence AGI then you need to show that it it generalizes to all these domains is when everything's filled in all the all the check marks are are are filled in then when we have it we it's yes so I think there are missing capabilities right now you know that all of us who have used the latest sort of LLMs and chat bots will will know very well like on reasoning on planning on memory I don't think today's systems can invent you know true do true invention you know true creativity hypothesize new scientific theories.

2:18 They're extremely useful. They're impressive. Um, but they have holes. And actually, one of the main reasons I don't think we're we're at AGR yet is because of the consistency of responses. You know, in some domains, we have systems that are can do international math olympiad math problems, you know, to gold medal standard with our alpha proof system, but on the other hand, these systems sometimes still trip up on high school maths or even counting the number of letters in a word.

3:03 So that to me is not what you would expect. That level of sort of difference in performance across the board is is you know not consistent enough and therefore shows that these systems are not fully generalizing yet. But when we get it is then like a phase shift that you know then all of a sudden things are different all the check marks are checked. Yeah. You know and we have