Today, we're joined by Pushmeet Kohli, who is the head of our Science and Strategic Initiatives Team.
We are not looking for something which is an incremental improvement. We are really looking at goals which are going to have a transformative impact.
One of my general takeaways is the pace of science advancement inside DeepMind is crazy.
We will have more and more powerful models, which will be more general, which will be able to do many more interesting things. Yes, we will have this amazingly powerful general intelligence. The question for our team is, what will we use it for?
I love that framework as we get intelligence. How do you actually wield intelligence to solve these problems?
Our job is to leverage all that progress, add to it to solve the next impossible thing for the benefit of humanity.
Hey, everyone. Welcome back to "Release Notes." My name is Logan Kilpatrick. Today we're joined by Pushmeet Kohli, who is the head of our Science and Strategic Initiatives Team. Pushmeet, thanks for being here. Excited to chat all things science.
Yeah, thanks for having me, Logan.
Let's dive in. I feel like we have-- one of my general takeaways is the pace of science advancement inside DeepMind is crazy. We have all these new alpha programs always coming out. There's tons of stuff, and hopefully we'll be able to deep dive into a bunch of these today. But do you want to, for folks who have maybe seen some of the launches, do you want to give us an overview of perhaps maybe the three most recent Alpha Science launches that we've done?
Yeah, so very happy to. So I think even in the last three months, we have been having a number of launches come up. The first one was AlphaEvolve, our super optimization, code optimization agent, which really showed a lot of impact in solving a number of optimization problems, from optimizing data centers to speeding up Gemini training itself. Then came AlphaGenome, which was our model for deciphering the human genome. And that was very exciting. And then very recently, we had AlphaEarth, which was another geospatial model for shedding light on what's happening on the planetary scale.
I love that. I know you'll get a bunch of ignorant questions from me because I have a much better worldview of the Gemini ecosystem than I do of some of our Science work, but maybe to start, like for AlphaEarth, can you double-click as far as what-- if people missed the headlines, what is it, why should people be excited, what problem is it solving, all that stuff.
Yeah, so if you think about Google Earth, what Google Earth was able to do was able to integrate all this amazing information about our planet and making it accessible to everyone. What AlphaEarth does, is it basically, again, combines a lot of information that has been captured by remote-sensing satellites and so on into one single representation. And what this semantic representation is then able to do is you can use it to propagate information that you know about a specific attribute of a location. You might say, well, I'm looking at habitats of certain species. And I have found this species in this particular position on the map. Now propagate that and try to reason about where else will I find the same thing. So it essentially captures the semantics of what's happening at the geospatial level at the planetary scale.
I love that. How do you think about, as you describe just even those most recent three, there's such a-- there's a big range of that work that's happening. I think for your team, but also just generally for DeepMind, how do you think about what are those problems are we tackling versus what are they not? Because they feel it's very-- like, it's almost surprising to me that we'd be able to make progress in such different domains. So I'm curious to double-click on how you think about that framework.
Yeah. So I think-- if you think about the whole spectrum of intelligence, you can categorize it into maybe three broad categories. The first category is are intelligence competencies that everyone has, most humans have. Like all of us will be able to look at an image and say what is in that image. We will be able to recognize the characters in the image. We will be able to read handwriting and so forth. Then there is a second category of intelligence, which is more focused on expertise. If you see someone's symptoms, can you figure out what's happening to them? Can you make an accurate diagnosis? So that's expert-level intelligence. Similarly, if you are given a specification, can you write a program that implements that specification? Again, that requires a lot of expertise. So that's expert-level intelligence. Now, then there comes a point where you have certain tasks which maybe no human can solve. If you give a sequence of amino acids, which is a protein, and if you give that protein sequence to any human, like one of the smartest human being, and say, can you figure out what is its 3D structure? We will not be able to do that just by reasoning about it. It takes-- before we released AlphaFold, it used to take almost a million dollars and sometimes multiple years to figure out the structure of a single protein. And so with AlphaFold, we were able to do that in a matter of a second for the cost of a few cents. And so those are the kinds of competencies that our science program tries to aim for, things that are not possible today. And really building intelligent systems that can solve, really, the unknown.
We were talking before off-camera about what the different levels of impact from all these different initiatives. And I feel like actually, the three that we released most recently are great examples of the variance and the level of impact across different categories in the ecosystem, so I'm curious to also double-click on this framework of how we think about impact from all this different science work.
Yeah. So I think essentially the Science and Strategic Initiatives unit, we really focus on transformative impact. The important piece is transformative impact that will benefit humanity. And what is the impact type? The impact could be scientific, it could be commercial, or it could be social impact. I can give you examples of all these three different categories. So the exemplar for scientific impact is AlphaFold. I mean, everyone knows about it. It was essentially something that we pioneered, one of the great success stories of modern AI where this amazing fundamental root note problem in science, which has so many different applications, from drug discovery to understanding human health to understanding, frankly, life itself was open for many, many decades. And AlphaFold was this remarkable system that could, from a protein sequence, figure out what its 3D structure. And it made that big breakthrough at the height of the pandemic in 2020. And even in the last four years, there have been-- it has been used in so many different places and the scientific impact has been amazing. It has-- it was recognized as one of the most cited scientific papers and won the Nobel Prize last year for Demis and John. And so you can imagine, in such a short span of time, the amazing impact that AlphaFold had in the scientific community.
Now, moving on from AlphaFold-- I mean, of course it had commercial impact as well because we spun off Isomorphic Labs with the specific mission of really revolutionizing drug discovery, and I think they are doing a very good job at that. Now double-clicking on the commercial impact point, one of the agents that we released earlier this year is AlphaEvolve. Now AlphaEvolve is a Gemini-powered coding agent. It's not giving you a solution for something that you do every day. It's essentially trying to do things which no computer scientist has been able to do. So imagine, over the last two decades, people have tried to make sure that how we optimize our data centers is fine-tuned, and some of the smartest people at Google have figured have tried to find out, what are the right algorithms so that-- how do we distribute jobs to different compute clusters? And that is an incredibly important thing because a lot of compute and monetary impact is to be had. If you could refine this. And by running AlphaEvolve, not only did we make an improvement, but we made a very significant improvement. Almost 0.7% of the whole compute fleet we were able to save. And that's a huge saving when you actually multiply it with the amount of compute that we use. Similarly, it was able to speed up Gemini training itself, which is a very important element. And now, again, although it was focused on commercial impact, it had amazing scientific impact as well. We ran it on a bunch of open maths problems, and for 75% of them, it found the state-of-the-art solution which the best mathematicians had been able to find, and for 20%, It went beyond them. So these solutions were not known. So that was commercial impact.
Now coming to social impact, and a great example of social impact is SynthID. So SynthID is essentially our state-of-the-art watermarking system. Now, we have been one of the pioneers of generative AI. And if you think about both the VR launches as well as the Genie launch, which happened recently, you could see the impact this technology is going to have. But there are also risks associated with that. In the information ecosystem, the quality of these models and the quality of these results are so much that it's difficult for users to even see what is synthetically generated versus what is natural. And this was one of the key elements. Like for Google, it's really important that users understand what was synthetically generated and what was physically represented, the physical sort of world. And so SynthID was born out of that problem motivation. And I'm very happy and proud to say that we are one of the first of companies that every GenAI content across all modalities at Google, whether it's text, whether it's images, whether it's videos, is watermarked with this imperceptible signal, which is robust to different transformations of those signals, and can be detected.
Yeah. I don't think people appreciate how difficult technically of a problem it is to solve synthetic, so happy to double-click on that. But just in general, I feel part of the challenge, I imagine, for your team-- and I'm curious to get your reaction to this, is just, like, you have, on some ends of the extreme, AlphaFold and Nobel Prize science impact. On the other hand, you're probably saving Google billions of dollars by the 0.7% improvement in fleet efficiency and Gemini training in general. How do you-- from your team's perspective, is it just there's all of these huge opportunities, just, like, low-hanging fruit sitting everywhere and it's like, we can just go tackle a bunch of these? Or is it like, are you actually really having to turn over a bunch of stones and be really-- do a whole lot of exploration to find these places where AI can have a huge impact and Google is well-positioned to solve? I'm curious how that initial process works.
Yeah, so I think that's a very good question. And as you can see, there are a lot of problems that we could solve, but we think about it in a very specific way. I think in terms of what we try to solve, there is a very simple sort of algorithm that we follow. The first point of that algorithm is to make sure that the problem is going to have transformative impact, and is going to be feasible. So we are not looking for something which is an incremental improvement. We are really looking at goals which are going to have a transformative impact, whether it's scientific, whether it's commercial, whether it's social, but it needs to be transformative. And it needs to be feasible, in the sense that everyone in the community believes that it's going to be transformative, and it's going to be achieved at some point of time. We don't want to work on time travel. We want to work on things which are actually going to happen. So that's the first requirement. The second requirement is that there should be almost consensus that no one will be able to achieve it in the next five to 10 years. So it needs to be hard. If someone is working on it and thinking that it will be solved in the next six months, then it's not a problem for us. Our team is particularly focused on these things which are incredibly hard, and that require multidisciplinarity, which require the best AI research, coupled with the best engineering, coupled with the best compute or the best data to come together to solve this incredibly hard problem. And then the third requirement is we think that the consensus in the community that this is going to take five to 10 years is wrong, we can do it in half or one-third of the time. If those three conditions are met, we take it on.
Yeah, that makes complete sense. I'm actually curious, through this framework, how does AGI fit into that? Is there a world where-- and actually, as you solve-- as you think about these, you're solving very, it seems like, domain-specific problems. And is there a world where this intelligence generalization makes it easier to solve these problems? Or do you still think, even in a world where we have really advanced general purpose technology, from an AI perspective, you'd still need to go and build a bunch of bespoke stuff or leverage those systems in a very specific way? And the Gemini AlphaEvolve one is a great example of this, maybe, where the model is really good, and then you can supercharge it in a bunch of ways to help solve some domain-specific problem?
Exactly, it's exactly. The idea is that progress will happen in AI. And we will have more and more powerful models, which will be more general, which will be able to do many more interesting things. When we were thinking about reinforcement learning and being able to solve search problems, whether it was AlphaGo, whether it was AlphaChess and so on, there is development that is happening which is making things possible that was not possible earlier. And our job is to leverage all that progress, add to it to solve the next impossible thing for the benefit of humanity. And the key question is, yes, we will have this amazingly powerful general intelligence. The question for our team is, what will we use it for? The specification. The problem that we will apply this amazing technology that will eventually be created.
Yeah. I love that framework of as we get intelligence, how do you actually wield intelligence to solve these problems? And this is my plug for the DeepMind team in general, which is I think so many of the labs-- and I think this Demis said something like this. Like, so much of the narrative about AI right now is that in the future, it will be used for scientific impact, and I feel like DeepMind historically is the only place where it is actually, today, being used for scientific impact across the ecosystem. I feel a bunch of the examples you gave are the proof is in the pudding that DeepMind is doing this work, which is awesome to see. I'd love to, again, through my worldview of-- I spend much more time thinking about Gemini stuff than I do our Science initiatives just because it's closer to my world, how are those two worlds connected? Like the Science unit and the work that's happening across GenAI and across the Gemini world, how do you collaborate with those teams? How does the technology transfer happen? You talked about AlphaEvolve where some of the foundational components are the same. I'm curious to hear that story.
Yeah, so I think-- this is the other thing about DeepMind, I think, and Google DeepMind, which is the culture, which is extremely collaborative. We collaborate very, very closely with Gemini. I can give you many, many examples, whether it's the collaborations on the base architecture as to what is the next generations of architectures that we should be developing that would make our model-- which will make Gemini be able to understand and perform much better on scientific tasks? So this is one area that we collaborate on. We collaborate on evaluation. What are the next types of things that we should evaluate the system on to really see how that's making progress? And then we also collaborate on data. What kind of data should we train Gemini on so that it has a really good understanding of these amazing specific domains that we want to go after? Whether it's a biology, whether it's chemistry, materials, code, cybersecurity. We really want to make sure that Gemini has a good handle of all these different types of domains. And then, of course, there are projects that we work on together. For instance, the IMO Project has been a collaborative project from Gemini and the Science team. And as part of it, we first worked on AlphaProof and AlphaGeometry, and then this year, we basically learned a lot and then tried to integrate all of that into the Gemini deep learning model, and we have got this amazing result. Which really shows that both the generality, as well as the pioneering aspect of being able to go for that very ambitious challenge, can work for both kinds of models.
Yeah. Let's actually double-click on this example because I think it's perhaps one of the best versions of this story, which is I think the original-- so we got the IMO, I think it was silver medal, and we were one point away from gold last year. That was a domain-specific model that wasn't the base Gemini model that was available to the world. And that actually was-- I don't remember. Was that a custom alpha? Was that an alpha version of the model? Or do we say-- do we brand it in that way?
Yeah, so it actually was a collection of two models. One was called AlphaProof and the other one was called AlphaGeometry. So AlphaGeometry, by its name, you can guess that it was specifically focused on solving geometry problems. And then there was AlphaProof. And AlphaProof had this very interesting aspect, is that there was an underlying LLM in AlphaProof. But what it was doing is essentially searching in the space of all proofs. So it will formulate the problem, it will take your math problem, convert it into a formalized problem specification in a domain-specific language called Lean. And what this language is able to do is it's able to give you proofs. If you solve the problem, you will get a formal proof so that you know that it's actually correct. That if you find the solution, it is actually correct.
And just to clarify, for folks who aren't familiar with-- and I know pretty much nothing about formal math proofs, so I'll put myself in this camp, the nice thing is that it is like-- it has this verifiable component. So actually, it's not like a question of, if it runs, then essentially you know that this thing is like a true--
