Google DeepMind CEO Demis Hassabis says solving mathematical benchmarks like Erdős problems does not constitute true AGI

Ofc Demis is absolute goat no doubt but its kinda funny to think we pushed the goalpost to a degree that it takes Ramanujan to be considered AGI

This creates problems for the @Kevinroose narrative that AGI is near.

woah @Noahpinion is attacking @demishassabis here — calling what Demis said not “credible” — but misinterpreting what Hassabis is *actually* saying.

What Demis actually said is both completely credible and completely correct. He is not emphasizing a specific problem but rather the challenge of building systems that can innovate across domains.

We are not there yet. And Demis should know since he runs probably the best team in the world working on that very question.

@ATabarrok He didn’t move the goal posts. Those never were the posts.

There is no test for AGI. The goal posts are: can you explain how a person achieves the unique creative capacities they do?

It’s not just “replicate some things humans can do” (like arithmetic).

*Explain* creativity.

Ramanujan is my hero. I remember the Ramanujan mathematics competition and the scholarship which played a pivotal role in my interest in research.

@ValerioCapraro He says that the full version of AGI is only a few years away.

It can still be framed somehow as a glorious tree-search, admittedly with a very powerful branch sampler, that can only connect a pre-defined entrance to a pre-defined exit.

@Noahpinion due respect Noah but you are distorting what Demis said here and that’s undermining your credibility rather than his. He’s not focused on a specific problem; he is focused on the problem of domain general innovation. You have misread him.

@cloneofsimo My dude, the goalposts have been moved down. This was always the goalpost for DeepMind.

@Noahpinion Noah, there are thousands of problems mathematicians tried to solve for decades... Solving one a good proxy for saying AI can help advance mathematics — and a terrible proxy for comparing AI and human abilities. You have an incorrect, uneducated idea of what math is about.

@Oisin_FC @Noahpinion My view is that most professions are in an ~approximately equal amount of danger, but there are structural and institutional reasons to think academic mathematics will survive, in some form, for some time. I have no doubt there will be short-term shocks to the market though.

@teortaxesTex instead of calling it AGI we can just call it AI haha

@ValerioCapraro People think we're getting eternal life in the next 20 years. We haven't even cured male pattern baldness.

Or, math has structure and if you combine that with a powerful learning algorithm, one that is much more capable and tireless than the human brain at search, you end up with a math super-genius that is not as generally intelligent as a human

@oogway_defi @DavidDeutschOxf @ATabarrok No:

You missed this part: “A neural network can search a huge graph of possibilities. It can recombine existing knowledge at unprecedented scale. It can discover surprising solutions inside an already defined conceptual space.”

With current discoveries, many engineering problems hidden in the current knowledge space can be solved. Solving these can lead to new discoveries and transform the world into a different, better one. But what he is saying is that a discovery like Einstein’s relativity theory is not possible with the current system even with scaled computing, and no one is expecting that.

@ValerioCapraro Yes, and I've written an essay on one of the reasons we actually cannot get that close to true human AGI with transformers:

@ns123abc This a directionally respectable view, but many Erdős problems' solution will likely produce geniunely new and interesting mathematics, and probably require forging deep mathematical wormholes. I'd wager that the Erdős conjecture on arithmetic progressions is one of them.