| Not checkable |
Not checkable |
23 Jul 2025 |
It’s a multi objective optimization problem.
Not checkable · 23 Jul 2025 |
|
| Not checkable |
Not checkable |
23 Jul 2025 |
Because for the first time in human history we wouldn’t be resource constrained.
Not checkable · 23 Jul 2025 |
|
| Not checkable |
Not checkable |
21 May 2025 |
And at the same time, you want to spend a bunch of effort on what's coming next.
Not checkable · 21 May 2025 |
|
| Not checkable |
Not checkable |
21 May 2025 |
You need to scale to the maximum the techniques that you know about and exploit them to the limit
Not checkable · 21 May 2025 |
|
| Not checkable |
Not checkable |
Dec 2024 |
So it means I’m not that good at celebrating successes.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
AI is potentially the perfect description language for biology
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I think I’ve used gaming at least in three different ways.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I think some of the ambitions that we had for the game design and the technologies were a little bit too ahead of where the technology was.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I think that would be one of the most consequential inventions humanity will ever invent.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
My original ambition in AI is to build what we sometimes call artificial general intelligence, which is a system that can exhibit all the…
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I play chess sometimes late at night online, just as a little bit like going to the gym for the mind, you know, to keep the mind sharp.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
For me, there’s nothing more fascinating than a phenomenon of intelligence and trying to understand what it is.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I’ve always believed it’s quite hard to describe that system just with a few mathematical equations, so it’s not like physics.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
it could potentially have big implications for complexity theory including P=NP
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I don’t even think of it as failure.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
So games have been a critical part of my entire career.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
Technology as transformative as AGI requires exceptional care and foresight
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
the reason I spent my whole career on AI is because I believe it could be the ultimate tool to help with science.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
The technology itself is kind of neutral.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
It’s a geopolitical question.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I’m a really big believer in interdisciplinary research and interdisciplinary science.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
Either lots of data and/or an accurate and efficient simulator
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I don’t really regard it as failure because it’s so fundamental to advancing anything in science.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
My aim was, as always, to build AI algorithms that were general enough to eventually be able to be applied to scientific problems like…
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I think probably the most important teacher for me was my high school maths teacher for my A Levels
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
Turns out this is a very general solution and many problems fit this approach
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
Biology at its most fundamental level can be thought of as an information processing system albeit a phenomenally complex and emergent one
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
that convinced me that AI was what I wanted to spend my whole career on.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I’ve always had unbelievable drive since I can remember.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I think science is the most important, and the scientific method is probably the greatest, invention humanity has ever come up with.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
AlphaFold is accelerating progress on a huge range of problems
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
AGI - the ultimate general-purpose tool to help us understand the universe
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
We sometimes think of this as doing Science at Digital Speed
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
Go is the most complex game ever - the pinnacle for games AI
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I think science is what drives all progress.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
Actually, I think AI systems and AI algorithms are quite a good match, a good description language, for biology.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
Any pattern that can be generated or found in nature can be efficiently discovered and modelled by a classical learning algorithm
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
AI has incredible potential to help with humanity’s greatest challenges
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
Classical Turing Machines can do much more than we previously thought
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
Finding the optimal solution in an enormous combinatorial space
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
Critical threshold: predictions had to be accurate down to the width of an atom (<1.0Å average error) to be of use to experimentalists
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I don’t really think of things as work-life balance.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I hope that AI as a field will be part of helping humanity solve some of our greatest challenges.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
For me, maybe it would be to believe in your ideas, even though they seem quite outlandish.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I think I’ve thought of biology as an information processing system at a fundamental level.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
I just think of it as a necessary part of the scientific process.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
Actually, what I realised was that timing is very important.
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
Dec 2024 |
But AI must be built responsibly and safely, and be used for the benefit of everyone
Not checkable · Dec 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
But it's become too popular for that.
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
always interested in the big questions
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
Physics was my favorite subject,
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
So if we could create that [intelligence] artificially and make that abundant and have even super-intelligence
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
If you want to understand the fabric of reality or the nature of time or any of these big questions or just the universe, you study physics.
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
I never cared about that,
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
It's sort of surprising to me that more people haven't realized that a lot earlier.
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
I'd rather it would have stayed more of a scientific level,
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
Obviously I didn't know it could be done in my lifetime.
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
I've been thinking about this for decades.
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
It was so obvious to me this was the biggest thing,
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
We don't have to wait
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
People still don't really understand what I've lived with and sat with for 30 years.
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
that would change the whole world,
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
Maybe it's a watershed moment for AI that it's now mature enough, and it's advanced enough, that it can really help with scientific…
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
we were lacking some tools to tackle such momentous questions.
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
overhyped in the near term
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
So it seems an obvious logical progression.
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
still massively underrated in the long term
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
21 Oct 2024 |
I was going to do this no matter what.
Not checkable · 21 Oct 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
The amount of information and models that can be built up just from language is surprising.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
Yes, exactly.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I think we get some grounding through the RLHF feedback systems because obviously the human raters are by definition, grounded people.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
In my opinion, they’ve clearly got some form of concepts and abstractions and things like that.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I’m obviously a huge techno-optimist but I want us to be cautious given the transformative power of what we’re bringing into the world…
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I think if it’s not properly grounded, the system won’t be able to achieve those goals properly.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
They’re not at the level where they can create the hypothesis themselves or ask the right question.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
The first part is making sure you detect it ahead of time.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
So I think we are going to need some automated testing.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
That’s what I think true multimodality is.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I’m very optimistic about that.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I think we’re still in the early days of those kinds of systems being at risk.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
You don’t want to be live A/B testing out in the world with these very consequential systems because unintended consequences may be quite…
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I think that maybe in the next three, four, five years, we would also want air gaps and various other things that are known in the security…
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I think no one knows.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
The brain is an existence proof that general intelligence is possible at all.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
One order of magnitude is probably about the maximum that you want to do between each era.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
In a way, I look at the large models today and I think they’re almost unreasonably effective for what they are.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
It’s been fantastic actually, over the last year.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I think we’ve got to push scaling as hard as we can and that’s what we’re doing here.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
It is more challenging because of the data problem.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I haven’t heard a compelling, clear answer to that from proponents of just open sourcing everything.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
So that’s very useful already for many of the problems we deal with today, but not the most high-level creative problems.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I think they’re not reliable yet enough to do things like science with them.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
So I think the world’s about to become very exciting in the next few years as we start getting used to the idea of what true multimodality…
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I think that suggests there must be high-level common algorithmic themes around how the brain processes the world around us.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I think we’re just at the beginning of actually understanding how exciting that might be to interact with a full multimodal model system.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
That pushes us in very interesting research directions that we think are going to be useful anyway: sampling efficiency and data efficiency…
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I think this has to be a big collaboration with many stakeholders from civil society, academia, government, etc.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I think that we’re making great progress with math and things like theorem proving and coding.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
We’re very excited about our progress with things like Gato and RT-2.
Not checkable · 28 Feb 2024 |
|
| Not checkable |
Not checkable |
28 Feb 2024 |
I sometimes like to call it virtual brain analytics.
Not checkable · 28 Feb 2024 |
|