Koray Kavukcuoglu on frontier models, coding agents, and building AGI

0:00There's always a lot of discussion about is there a test for AGI? There's no test for AGI. I don't think anyone has a test that they can say, okay, if you get this much on this test, that means you are at AGI. And I don't think that is— - If you do have that benchmark, send it to us, please, because we want to— we'll use it, we'll run our models on it. Hey, everyone, welcome back to Release Notes. My name is Logan Kilpatrick. I'm on the Google DeepMind team.

0:28Today we're joined by Koray Kavukcuoglu. Koray, I'm super excited for this conversation. We sat down probably three or four months ago, and so I'm excited to catch up, reflecting on a lot of the progress that's been made over the last few months, looking into the future, Gemini 4 on the horizon. The third iteration of the 3.5, post-3.5 series, lots of progress in literally only three weeks. We launched 3.5 at I/O, 3.6, and then three weeks later landed 3.7.

0:59Reception has been super positive. Congrats to you and the team. Yeah, it's been positive. Do you want to— I don't know if there's anything you want to say about what made that model. We teased a little bit of architectural innovation or something like that, which I'm not sure we'll spill the beans on, but— - Not really. - Yeah. - Going from the initial 3.0 launch, my reflection is we learned a lot in terms of understanding what it means to do coding and not just coding, right?

1:28Like what it means to do software engineering, what it means to work with tools or work with the functions that people use every day. Basically, turn this whole thing into an agent, from a model to an agent. And that transition, I think we were walking around it, and that was the thing that we wanted to directly tackle. And of course, software engineering is the most critical domain and environment that you want your systems to be successful at, because it's at the root of many different things that you can do.

1:58During that time, we learned a lot in terms of how to train a model, how to train an agent that can actually code with you. And after that, I think the steps have started becoming faster. And in any research project, there are many parallel tracks that is going on at the same time too. So, what we are trying to do right now is combine all these learnings, of course, add on top of them, understand, what it means to do agentic actions and agentic workflows better together with new architectural improvements, together with new ideas that we have been working on for a while.

2:34Many of these things that we have done in 3.6, 3.7 go back like a year or more, and then those start paying off and you converge them into the model. And it's great to see the quality impact that we got from that. We were very excited, in the run-up to 3.7, when we were doing 3.6, of course, we could see we could see what 3.7 could be or what the next step could be. And then it came together. And internally we started enjoying that model a lot. So that's the parallel track effect that you are going, that you are seeing.

3:07- There's a lot of parallel tracks. I don't know if we've done this in the past, but a few weeks ago we publicly said that Gemini 4, we're working on it. It's the most ambitious pre-training run that we've done so far, which is really exciting. People seem excited by it. I think folks are excited internally. Lots of research going into this. Yeah, I'm curious, this has historically been one of our strengths and the team, seems incredible and they've always done a great job.

3:32And so I don't know if there's anything top of mind or anything more you want to say about that. Yes, it's the most ambitious run, correct. And yes, so far, touch wood, it's going well. I am very excited. The team is very excited. I look at these as step by step, right? So, far, everything is looking good. We are excited. But it's all about realizing that potential until we get to a moment that we are actually putting the model in front of users, first internally, then externally.

4:04I'm always in a very cautiously optimistic state. - I think maybe to tee you up for a very specific question there's a meme about whether or not we care about being at the frontier. I see you in meetings all the time and know what your point of view is but what's your point of view as far as whether or not, you know, you think it's important for us as Google DeepMind and as a team to trying to be at the frontier? - I was surprised. I've seen these comments too, and I can understand obviously our quality of current models are a little bit below the frontier.

4:38- Yeah. - And that's fair enough. To put it very bluntly, there's nothing other than being at the frontier that is important for us. Yeah, that's it. All our goals, all our focus is always about that. And when we are making decisions, when we are making prioritizations in terms of ideas, in terms of what to do and what goals we should follow, it's all about being at the frontier, and I'm 100% certain that, we will be at the frontier. We have an amazing team.

5:09I have so much trust in the team. There is a great amount of creativity and dedication in the team, and being part of Google, we have amazing resources available to us. We have the whole full stack that we can and we are able to optimize to be able to achieve our goals. If you think about Google as a company, it is a company that has always been at the frontier of investing in the next big technology. It has never shied from investing in a long-term important project, right?

5:50AI investment in Google has been not the last five years or ten years, right? Fifteen years ago when there was no AI, Google was investing in AI chips. - Yeah. - Not chips only, the AI chips. We were not even talking about AI at the time. So, that kind of long-term vision and investment has always been part of Google, like the quantum computing, Waymo. We talk about these things as if it's normal, but these are very, technologically critical and heavily scientific and long-term ambitious endeavors that it's in the DNA of Google.

6:25So, that's why I feel I have the whole trust for us to be able to achieve that frontier. That's all our goal. That's why we were set up. That's what the whole team feels. Yeah, there's nothing less than that. - I love that. And I think in this role now, that you have leading all of Google DeepMind, the frontier AI folks, we actually have a frontier AI team, and the products and other stuff are now part of this story. And I'm curious— is there anything from this framing of like the frontier is what we're focused on that you're excited now that we're bringing these teams together?

6:59I know there'll be a lot of business as usual for a lot of folks, but anything particularly that you're excited about? - Gemini, of course, is very much focused on building AGI and delivering the goals. At the end of the day, it relies on two things. One, good execution, right? It's very detailed. It's very complex. To build AGI, something that you can put in front of people that they can work with. But also at the same time, it relies on a very large funnel to get ideas from everywhere.

7:28So by bringing together Google DeepMind this way, what I'm excited about is, as you said, there's the frontier AI, there's the products to be able to really orient and steer at least the idea, the knowledge and the focus towards this. You don't want to continuously streamline everything on one because the idea is at the end exploration. Success in building AGI still depends on innovation and exploration and channeling that in the right way through Gemini.

8:00I'm excited about being able to do that more. The path to building AGI then, really depends on our interaction with the users. And when we say users, yes, some people are using these models in their daily lives to go through their emails, but some scientists are using these models to be better scientists to do actual research to invent things that has never been invented and getting help from the models. So, the spectrum of usage of these models covers everything that is happening in the world.

8:33So that usage is our guide, right? That tells us what are the problems that are worth solving, that people want us to solve, and we need to solve those problems. And that leads us towards AGI. That's how we have something general that people like interacting with, that people rely on, that people trust. - Yeah, I think last time we were talking you had mentioned this co-building AGI with our customers framing, which I think has just continued to like it makes so much sense.

9:02It resonates. I think it makes it clear why the deployment matters so much, why this feedback flywheel matters so much. - There's always a lot of discussion about is there a test for AGI? There's no test for AGI. I don't anyone has a test that they can say, okay, if you get this much on this test, that means you are at AGI. - And I don't think that is— if you do have that benchmark, send it to us please, because we'll use it, we'll run our models on it.

9:29- All right. When you look at the history of AI, history of machine learning, it has always been like that. It's the progress. It's the journey that makes the difference, right? Go back like fifteen years ago, like twenty years ago or even more, there was always some notion, some concept of if we achieve these kinds of tasks, that means that is the true intelligence. And at every time period, each and every one of those were the legitimate targets to follow up on at that point.

10:00And that's what leads to this kind of progress that is happening. And that's the progress in itself, right? Picking the next goals is also a big part of that progress. And that's what we are doing, right? And you can see that the intelligence, from the very early days of AI being very narrow to now it's becoming very general. The goals that we are picking are very general goals. That's why we are building general intelligence. And I think it's going to continue like that.

10:26You go back twenty years ago, you look at what we are doing right now. It like, okay, maybe AI, it's already there. Like, what people were thinking twenty years ago in terms of what the goals was a lot of things are achieved. - Yeah. - But I think the more important thing is, at the end of the day, building an intelligent entity that you can trust, that you can work with. And if we can get there, and I don't think that there is a sudden threshold of that, it's more of a building up that trust and getting better and better at that.

10:58- Yeah, I'm with you. You mentioned twenty years ago, I want to go twenty years back. I don't know, maybe not twenty years back exactly, but you were Google DeepMind's first deep learning researcher, which is very interesting. And I want to double-click on that. You were at Yann's lab with a bunch of other folks. The deep learning community was quite small at that time. Previous iterations, DeepMind CTO leading a bunch of technical efforts. The story that you were telling about how during the Google DeepMind due diligence process, Jeff was asking you questions about random places in the Google DeepMind codebase because you literally reviewed all of the code.

11:32You had to, you had to be the one to explain random things, which seems like a crazy problem now in today's world where AI helps us do that. You've been at the helm and pushing on the frontier now for a long time specifically inside of Google DeepMind, before Google DeepMind. And so much of the world has changed. I'm curious if you want to reflect back on any of those moments, any of the papers or technical breakthroughs, obviously, Google DeepMind has done so much that stand out to you as you think back.

11:59- First, of all, thank you very much. Yes, I'm old. I think the journey in Google DeepMind has been very exciting. There was a time at which me and Karol Gregor, we were at Yann's lab, we joined Google DeepMind at the same time. I started the deep learning team and it grew from there. There was no big plan about how all this was going to happen. - And just to clarify, other folks were just doing, they were not doing deep learning, they were doing like what?

12:24Like just like RL stuff? - I think when you're in a startup, when you are doing research, everyone does everything, right? There were like three, four deep learning labs at the time. Bringing that deep learning background together with Karol into DeepMind, it was an amazing environment. I remember, I think in research there were ten, twelve people in research and everyone is working on some really interesting idea from their own background, like a really interesting group of people.

12:53At some point there was a discussion of, okay, let's pick an ambitious task that we can corral the whole research team around and try to show progress on that task. And that's how Atari started, the Atari games and trying to train an agent that can learn by itself playing those games and being successful in those games. It was an internal project because Atari at the time, there were RL researchers who were thinking that this can be a good environment to measure success.

13:22That was the first example of: we are known for these focus projects in DeepMind and we carry that over to now Google DeepMind as well, that we bring a diverse group of researchers and engineers together to solve a hard problem. That was the first example of where we did that. A large group of researchers that came together every week we would measure progress and that's what led to DQN. That's the first example of how deep learning and RL could come together and create a successful agent that can learn to achieve a task by itself.

13:54I think it convinced me and it convinced many people this is actually possible. That one, games are the right environment that you can actually study intelligence and it can scale. And two, at the time, deep learning was already very popular, deep learning and RL together is the basis of the recipe for building intelligent agents. It opened that door. - Was it that obvious? It's more obvious in hindsight, but like it really felt like the path was lit then, or at least there was a sliver of light at the end of the tunnel perhaps?

14:30- I'm very cautious about these things. I'm not sure about the path that was lit but we all felt like it was an important moment because it gave us confidence that we should continue. When you look at the history of DeepMind, history of Google DeepMind, bigger and bigger goals have been achieved in this trajectory, right? Like DQN was the first step and then every step after that was bigger and bigger and bigger, right? Like Go, Chess, being able to do those from scratch, and then using that knowledge for AlphaFold— all sorts of things, StarCraft.

15:04More and more we started building these agents. That was very exciting. - Yeah. I want to talk about agent stuff, but I'm actually curious if you contrast maybe some of the stuff that was happening back then to some of the stuff that's happening now. And one example of this is you mentioned this before about a lot of these ideas were purely in the research domain and it was, “Can you actually prove that this thing is possible?” And it feels like there's been at least some transition to we know a lot of it's possible.

15:34We have some amount of the recipe. It's really just a lot of really difficult research engineering work to actually bring the thing to life. And so, I'm curious if you want to talk about actually bringing the stuff to life now and how does it feel super different? - It does feel different. It's also the same in a strange way because we are still doing RL, we are still doing deep learning, we are still pre-training these models with very similar loss functions to what we used to do before.

16:05We are still doing RL pretty much using similar RL algorithms. The fundamentals of learning did not change. The fundamentals of optimization did not change, but the domains kept getting harder, right? Just as from Atari to chess to go to StarCraft, the domains kept getting harder. And the way they get harder is in real life you have ambiguity, you have depth. You have depth of ambiguity. That's the biggest change that you need to be able to deal with that.

16:34Language in itself is much richer, much complex, much more multi-domain, multifaceted than any constrained action space that you can work with. So, of course, like things became a lot more complex, a lot more larger scale. Like the basics of it is still the same. - Yeah. - But also at the same time, there's a lot more room to extract the information and be successful in this domain. But everything is a lot more open-ended, a lot more open to interpretation, and that requires a lot more intelligence to be able to deal with that to anticipate what someone is saying for the model to be able to partner with that person or to respond to that person.

17:17- Yeah, I want to talk about the future and stuff, but maybe one last question on this. Anything that's surprising to you as you look back and I'm sure there's lots of stuff, surprising new innovations that have come up, but anything that folks were working on before that has just turned out to be extremely successful that was not obvious it was going to be super successful or don't any way you want to take that. - I take a step back and if you look at it statistically, it is surprising that something has done this exponential takeoff.

17:50- Yeah. - Right? It's not like you have hundreds of these kinds of things where different scientific or technical domains are in this exponential takeoff stage. So, there's a little bit of an element of I feel lucky that we are living in this age and all the knowledge buildup has happened up until this point. All the technological progress in terms of chips and internet and data has happened up until this point that we are the ones living through this.

18:23I quite enjoy that. But also at the same time, like of course, it's intense and it's competitive. - Yeah. - I feel lucky. I feel surprised that a lot of the things that, twenty years ago, where we were exploring as students are now actually some of the biggest practical implications in the world. And I'm pretty sure a lot of people who were students at the time are feeling the same way. - Yeah. Somebody just describes you as somebody who loves the journey.

18:52A, is that actually true? And B, what that means in the context of what we're doing as Google DeepMind, because I think it's actually an important framing to capture your headspace and your belief about what we're doing. - I guess you have to like the journey to be able to achieve something. So yes, I like the journey. As I said, I think progress, especially in technology, in my experience sometimes happen through these stacked sigmoids where something happens, something gets unlocked, and then you have big progress, and then you put a lot of incremental stages on top which actually builds a lot of your understanding, which leads to another discovery, and then a little bit more of those, and then more understanding, and then more discovery.

19:37I think about that process a lot about what leads to progress, what leads to these jumps and how to achieve that in the context of any of the projects that we do. Because as much as I always like talking about, we are building AGI, we used to research AGI, we used to write papers about AI, we used to write papers about intelligence. And we did that, it was pure idea to pure science. Now, we are living in the world that we are actually building AGI because people are using it.

20:09So you have to build it, you have to do it with that responsibility. - Yeah. - Right? With that seriousness. But at the same time, you cannot let go of the fact that it is still research and innovation that fuels all this. That funnel is the critical thing, is the biggest differentiator. So I always talk a lot, think a lot about what makes it successful. What are the elements that will make it successful for us in the next step and the next step after that and the next step after that?

20:39And that's why I think that journey is very important, but the goal and the mission, that north star is extremely important, right? Like, journey only matters, you know that you are going somewhere good and valuable. And I feel very strongly and the whole team here feels very strongly that building AGI and bringing that positive impact to the world, to our users, to the world, I think is the best thing that we can do. - Yeah, And you made this comment, I don't know if this was yesterday, but there's something about doing it specifically at Google, I think is like a really interesting way.

21:14The mission is obviously important, but in context of Google and all these products we have and the broader work that's happening is such a unique way of doing this. - I think it's the right place. We talked about Google having the DNA of having, doing these technological investments, but the other part is Google has very large number of multi-billion people products. This is a company that really identifies itself by providing good services and good value for its users.

21:46Yeah. And understands that, and users like using our products, and that is really valuable. And as I said, I see that as a critical element of building AGI. The more we can put Gemini in a place where we understand and learn and get feedback from that interaction, the better we will be. And also, being in an organization where bringing value to the user is the utmost important thing. That's why I think it makes it the right place. - Yeah. In the context of us being very user-centric, lots of users asking, where's Gemini 3.5 Pro?

22:26So I'm curious if you can give us a quick spiel on where the model is. As much as we're willing to say. - Look, as I said, there are parallel tracks. In our model family, we always have the Pro, the Flash, and the Flash-Lite. But at the same time, we are seeing really fast progress with our Flash models. We are working on 3.5 Pro at the same time because it's one of those parallel tracks, because research happens in multiple scales at multiple times.

22:54Right? But also in terms of bringing the best thing forward for the users, we are seeing really good progress with Flash, we can iterate really fast. And I can see that like we talked about from 3.5 to 3.6 to 3.7, we have a Flash model that is getting better and better and more and more competitive, getting closer to that frontier. - Yeah. - So, that is very exciting for all the researchers internally as well. So, we are just using that as our progress path right now.

23:22And 3.5 Pro, it's still like, we are still working on it. But we are working on Gemini 4 as well at the same time. And there's a lot of exciting progress happening there too. - I love it. I'm excited. Another one that was top of mind, Gemini 3 felt like a huge moment for us. We were at the frontier. And then I think interestingly, it was almost like the frontier shifted and where there was a lot of new emergent capabilities and all that stuff happened to be in agentic coding and all these other things.

23:51And I think we were just starting to rev the engine and we were at the frontier in that moment and then the landscape shifted and how you think about that. - I think there are two things that we need to keep in mind. One is by definition in a very competitive environment, the frontier will always shift. There will be ebbs and flows of different things. And the cadence and the frequency of which lab is producing their most capable model is going to change.

24:21That's number one. But number two is a fair point. And we talked about that in the context of 3.5. I think we learned a lot about agentic actions and agentic workflows and being able to bring that to life, in a model. That's the process that we went through where I'm feeling right now, I'm feeling very comfortable and good right now where we are and our capability of understanding what users need when they are working with an agent that is partnering with them on any agentic task and workflow.

24:53But we went through that process of building that. - I love that. I'll ask you one more, which is a little bit hypothetical but if you could wave your magic wand and get the models to do anything and not need to spend a lot of time and energy, do you have anything on like the top of that list that you would want like a capability better behavior on something? It can be small or big. - I don't know. I if I had a magic wand I would just make them more intelligent.

25:23As the models get more intelligent, they do everything better, everything more intuitively. And I think that will be excellent. - Yeah, I love that. I sat down with Sergey at I/O, I think 2025 or something like that. I made this comment to him. This was after you and I were sitting next to each other listening to him and Demis do their talk. And I made the comment to Sergey that I/O and just like some of these moments of bringing people together, makes me feel the warmth of humanity.

25:51And I was saying that in context of some of the conversations that we were having. And so, I feel like this, it's just exciting to see, or I'm excited to see you take this leadership role because I feel that warmth of humanity from you at the same time that I feel the progress of like your focus on the frontier. And so on a personal level, I'm excited and I feel like we've got the things going in the right direction. I feel like you're well positioned to help us push. So...

26:19- Thank you very much. - Yeah. No fake, no free praise. I feel like you earn it. It's a lot of fun. Koray, this was an awesome conversation. Thank you for sitting down and chatting about all this stuff. I'm excited. Hopefully we'll have more to talk about soon. And thank you everyone for tuning in to this episode of Release Notes. We'll see you in the next one.