Google DeepMind CEO Demis Hassabis says solving mathematical benchmarks like Erdős problems does not constitute true AGI
DeepMind's AI systems recently solved eight Erdős problems.
#153Gary Marcus@GARYMARCUSwoah @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.
Noah Smith 🐇🇺🇸🇺🇦🇹🇼@NoahpinionHonestly this doesn't sound credible. Professional mathematicians tried for years to solve that problem and failed.
#153Gary Marcus@GARYMARCUSThis creates problems for the @Kevinroose narrative that AGI is near.
Valerio Capraro@ValerioCapraroFinally, a big name has the courage to tell it: we are nowhere near AGI. Demis Hassabis, CEO of Google DeepMind and Nobel laureate for AlphaFold, put it neat and clear: "Today's systems are nowhere near [AGI]. Doesn't matter how many Erdős problems you solve… I think it's far, far from what a true invention, or someone like Ramanujan, would have been able to do." This is the elephant in the room that many AI enthusiasts prefer not to see, or are actively trying to hide. Erdős problems are well defined, often combinatorial, on finite spaces. They are exactly the kind of problems on which current AI can achieve spectacular performance with a lot of compute and knowledge. 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. But true invention is something else. True invention is not only solving a problem. It is inventing new objects, new dimensions, new connections. It is inventing new problems. From resolving to inventing there is a discontinuity that we don't know how to bridge. We are making extraordinary tools. But we are nowhere close to AGI.
#153Gary Marcus@GARYMARCUSIMHO @Noahpinion is misinterpreting @demishassabis here, and I am with Demis 100% with what Demis 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 he should know since he runs probably the best team in the world working on that very question.
Noah Smith 🐇🇺🇸🇺🇦🇹🇼@NoahpinionHonestly this doesn't sound credible. Professional mathematicians tried for years to solve that problem and failed.
@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.
Noah Smith 🐇🇺🇸🇺🇦🇹🇼@NoahpinionHonestly this doesn't sound credible. Professional mathematicians tried for years to solve that problem and failed.
#156David@DAVIDSHOLZ@teortaxesTex instead of calling it AGI we can just call it AI haha
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTexthis lands harder after GDM's system solved 8 more, it's not Mistral-like "we have Mythos at home… in the aggregate". But sorry Demis. If this isn't going towards AGI, I guess we need some new label. It's pretty awesome.
#331François Fleuret@FRANCOISFLEURETIt 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.
Noah Smith 🐇🇺🇸🇺🇦🇹🇼@NoahpinionHonestly this doesn't sound credible. Professional mathematicians tried for years to solve that problem and failed.
@francoisfleuret +1, maths is basically search, automated systems should be pretty good at it (given the necessary compute)
François Fleuret@francoisfleuretIt 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.
@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.
NIK@ns123abc🚨 Google DeepMind CEO Sir Demis Hassabis: “Today’s systems, are nowhere near [AGI]. Doesn’t matter how many Erdős problems you solve… I think it’s far, far from what a true invention or someone like a Ramanujan would have been able to do” it’s over for the Erdős hype
#1446Shubhendu Trivedi@_ONIONESQUEThis 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.
NIK@ns123abc🚨 Google DeepMind CEO Sir Demis Hassabis: “Today’s systems, are nowhere near [AGI]. Doesn’t matter how many Erdős problems you solve… I think it’s far, far from what a true invention or someone like a Ramanujan would have been able to do” it’s over for the Erdős hype