korrents
Demis Hassabis

Demis Hassabis

@demis-hassabis · 24 positions · On ppll ↗

Co-founder and CEO of Google DeepMind, and a 2024 Nobel laureate in Chemistry for the protein-structure work behind AlphaFold.

Demis Hassabis did not write this page.

We collected these quotes from things they published elsewhere, and every quote links to where it was said. They have no account here and have not endorsed this site. Quotes are word for word; the short line under each one is our own restatement, not their wording. Their own site. Is this you? Claim it or ask us to remove it. Or tell us what is wrong here.

Filter All topics · Newest first
Topic
All 24AGI 6protein structure prediction 5AI alignment 2AI and jobs 2AI and science 2energy 2physics 2batteries 1chat interfaces 1
8 more topics
hiring 1neural networks 1recursive self-improvement 1robotics 1scaling laws 1solar 1taste 1world models 1
Sort
NewestOldest

  1. Our reading

    Anything nature shaped can be learned efficiently by a classical neural network, because evolutionary processes leave structure behind.

    neural networks

    Their wordsAnd I think the reason that's possible is that in nature, natural systems have structure because they were subject to evolutionary processes that shape them. And if that's true, then you can maybe learn what that structure is.
    spoken · official transcript hear it at 0:03:08 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 1st of 20 in this recording

  3. Our reading

    Information is the most fundamental unit of the universe, more fundamental than energy or matter.

    energyphysics

    Their wordsAnd it sort of fits with the way I think about physics in general, which is that I think information is primary, information is the most sort of fundamental unit of the universe, more fundamental than energy and matter.
    spoken · official transcript hear it at 0:06:17 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 2nd of 20 in this recording

  5. Our reading

    We have not scratched the surface of what classical, non-quantum computers can do; they can already go far further than anyone expected.

    protein structure prediction

    Their wordsI think we've proven, and the AI community in general that classical systems, Turing machines can go a lot further than we previously thought. They can do things like model the structures of proteins and play go to better than world champion level. And a lot of people would've thought maybe 10, 20 years ago that was decades away, or maybe you would need some sort of quantum machines to quantum systems to be able to do things like protein folding. And so I think we haven't really even sort of scratched the surface yet of what classical systems so-called could do.
    spoken · official transcript hear it at 0:08:53 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 3rd of 20 in this recording

  7. Our reading

    A model that can predict the next frames of a video coherently understands the world, in the only sense of the word that is doing any work.

    world models

    Their wordsI think to the extent that it can predict the next frames in a coherent way, that is a form of understanding, not in the anthropomorphic version of, it's not some kind of deep philosophical understanding of what's going on, I don't think these systems have that, but they certainly have modeled enough of the dynamics, put it that way
    spoken · official transcript hear it at 0:15:25 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 4th of 20 in this recording

  9. Our reading

    Intuitive physics can be learned from passive observation alone, so an embodied robot is not required to understand the physical world.

    physicsrobotics

    Their wordsBut it seems like you can understand it through passive observation, which is pretty surprising to me. And again, I think hints at something underlying about the nature of reality in my opinion, beyond just the cool videos that it generates.
    spoken · official transcript hear it at 0:18:06 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 5th of 20 in this recording

  11. Our reading

    Coming up with a good conjecture is harder than solving one, and that taste for the right question is what separates great scientists from good ones.

    taste

    Their wordsSo picking the right question is the hardest part of science and making the right hypothesis. And that's what today's systems definitely they can't do. So I often say it's harder to come up with a conjecture, a really good conjecture than it is to solve it.
    spoken · official transcript hear it at 0:38:03 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 6th of 20 in this recording

  13. Our reading

    There is a coin-flip chance that AGI arrives by 2030.

    AGI

    Their wordsMy estimate is sort of 50% chance by in the next five years, so by 2030 let's say. So I think there's a good chance that that could happen.
    spoken · official transcript hear it at 0:52:33 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 7th of 20 in this recording

  15. Our reading

    The real test of AGI is handing the system to a few hundred of the world's top experts and finding that none of them can locate an obvious flaw.

    AGI

    Their wordsAnd maybe also make the system available to a few hundred of the world's top experts, the Terence Taos of each subject area and give them a month or two and see if they can find an obvious flaw in the system. And if they can't, then I think you can be pretty confident we have a fully general system.
    spoken · official transcript hear it at 0:53:28 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 8th of 20 in this recording

  17. Our reading

    An end-to-end self-improving AI is probably possible, but it is not even desirable, because it is a hard-takeoff scenario.

    AI alignmentrecursive self-improvement

    Their wordsI mean potentially that's possible. I would say I'm not sure it's even desirable because that's a kind of hard takeoff scenario.
    spoken · official transcript hear it at 1:00:06 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 9th of 20 in this recording

  19. Our reading

    Scaling what we already have, with no further breakthroughs, will be enough to reach AGI.

    AGI

    Their wordsAnd so we don't know, I would say it's kind of 50/50 whether new things are needed or whether the scaling the existing stuff is going to be enough.
    spoken · official transcript hear it at 1:04:17 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 10th of 20 in this recording

  21. Our reading

    AI will not run out of data, because there is already enough real-world data to build the simulators that generate more of it.

    scaling laws

    Their wordsDo you have enough data to make simulations, so that you can create more synthetic data that are from the right distribution? Obviously that's the key. So you need enough real-world data in order to be able to create those kinds of data generators, and I think that we're at that step at the moment.
    spoken · official transcript hear it at 1:05:37 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 11th of 20 in this recording

  23. Our reading

    Fusion and solar are the two energy sources worth betting on, and solar's limiting problem is batteries and transmission.

    solarenergybatteries

    Their wordsI think fusion and solar are the two that I would bet on. Solar, I mean it's the fusion reactor in the sky of course, and I think really the problem there is batteries and transmission.
    spoken · official transcript hear it at 1:09:15 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 12th of 20 in this recording

  25. Our reading

    Today's text-box chat interfaces are too restricted to survive multimodal models, and will look archaic within a couple of years.

    chat interfaces

    Their wordsShouldn't it be something more like Minority Report where you are sort of vibing with it in a kind of collaborative way? It seems very restricted today. I think we'll look back on today's interfaces and products and systems as quite archaic in maybe in just a couple of years.
    spoken · official transcript hear it at 1:26:13 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 13th of 20 in this recording

  27. Our reading

    Meta is not at the AI frontier, and buying talent with enormous salaries is the rational move for a company that is behind.

    hiring

    Their wordsAnd Meta right now are not at the frontier. Maybe they'll manage to get back on there and it's probably rational what they're doing from their perspective because they're behind and they need to do something. But I think there's more important things than just money.
    spoken · official transcript hear it at 1:40:21 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 14th of 20 in this recording

  29. Our reading

    AI will not replace the best programmers; it will make them ten times more productive than they are today.

    AI and jobs

    Their wordsSo the great programmers will be even better, but there'll be even 10X even what they are today. And because there, you'll be able to use their skills to utilize the tools to the maximum, exploit them to the maximum.
    spoken · official transcript hear it at 1:43:27 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 15th of 20 in this recording

  31. Our reading

    AI will have ten times the impact of the Industrial Revolution and arrive ten times faster, which is what will make it hard for society to absorb.

    AGI

    Their wordsAnd the thing that's going to be harder to deal with this time around is that I think what we're going to see is something like probably 10 times the impact the Industrial Revolution had, but 10 times faster as well. So instead of a 100 years, it takes 10 years and so that's going to make, it's like a 100X, the impact and the speed combined.
    spoken · official transcript hear it at 1:45:37 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 16th of 20 in this recording

  33. Our reading

    The productivity AI creates should be shared out as a universal basic provision, with rare skills still buying you more than that.

    AI and jobs

    Their wordsIncluding things like universal basic provision or something like that where a lot of the increased productivity gets shared out and distributed to society and maybe in the form of services and other things where if you want more than that, you still go and get some incredibly rare skills and things like that and make yourself unique. But there's a basic provision that is provided.
    spoken · official transcript hear it at 1:46:31 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 17th of 20 in this recording

  35. Our reading

    The last steps to AGI should be taken by a CERN-style international research project, not a Manhattan Project.

    AGI

    Their wordsI hope we'll end up with something more collaborative if needed, more like a CERN project where it's research-focused and the best minds in the world come together to carefully complete the final steps and make sure it's responsibly done before deploying it to the world.
    spoken · official transcript hear it at 1:56:27 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 18th of 20 in this recording

  37. Our reading

    Quoting a number for P(doom) is a ridiculous notion, because it implies a precision that nobody actually has.

    AGIAI alignment

    Their wordsWell, look, I don't have a P-Doom number. The reason I don't is because I think it would imply a level of precision that is not there. So I don't know how people are getting their P-Doom numbers. I think it's a little bit of ridiculous notion because what I would say is it's definitely non-zero and it's probably non-negligible.
    spoken · official transcript hear it at 1:58:14 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 19th of 20 in this recording

  39. Our reading

    The brain is doing classical computation, not quantum, so everything it does can in principle be mimicked by a classical computer.

    Their wordsSo my betting is that it's mostly it is just classical computing that's going on in the brain, which suggests that all the phenomena are modelable or mimicable by a classical computer.
    spoken · official transcript hear it at 2:06:26 · all korrents from this recording

    ↗Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475youtube.com 20th of 20 in this recording

  40. 4 years earlier

  41. Our reading

    Accurate computational approaches are needed to close the protein structural coverage gap and enable large-scale structural bioinformatics.

    AI and scienceprotein structure prediction

    Their wordsAccurate computational approaches are needed to address this gap and to enable large-scale structural bioinformatics.

    ↗Highly accurate protein structure prediction with AlphaFold (with 33 co-authors)doi.org All 4 from this paper

  43. Our reading

    Existing protein structure prediction methods fall far short of atomic accuracy when no homologous structure is available.

    protein structure prediction

    Their wordsDespite recent progress10-14, existing methods fall far short of atomic accuracy, especially when no homologous structure is available.

    ↗Highly accurate protein structure prediction with AlphaFold (with 33 co-authors)doi.org All 4 from this paper

  45. Our reading

    A computational method can regularly predict protein structures with atomic accuracy even when no similar structure is known.

    AI and scienceprotein structure prediction

    Their wordsHere we provide the first computational method that can regularly predict protein structures with atomic accuracy even in cases in which no similar structure is known.

    ↗Highly accurate protein structure prediction with AlphaFold (with 33 co-authors)doi.org All 4 from this paper

  47. Our reading

    Contemporary physical and evolutionary structure-prediction approaches fall far short of experimental accuracy without a close experimental homologue, limiting biological utility.

    protein structure prediction

    Their wordsDespite these advances, contemporary physical and evolutionary-history-based approaches produce predictions that are far short of experimental accuracy in the majority of cases in which a close homologue has not been solved experimentally and this has limited their utility for many biological applications.

    ↗Highly accurate protein structure prediction with AlphaFold (with 33 co-authors)doi.org All 4 from this paper