[00:00] Hello and welcome to the Tech Download, a new CNBC original podcast where we unpack the tech stories that matter most. Each season, we dive into one big theme and what it means for your money with insights from the industry's most influential voices. I've always thought that in the end, it would be the most important technology we'll ever invent. And it's sort of the natural progression really of the computer age. This season, we're looking at Google DeepMind, the powerhouse driving the tech giant's AI push. We've been given rare access to
[00:32] key figures at the company, including our guest for this episode, DeepMind co-founder and CEO Deis. I think it's going to be like the industrial revolution, but maybe 10 times bigger, 10 times faster. So, it's incredible amount of transformation, but also disruption that's going to happen. Hey everyone and welcome to the tech download. Allow me to reintroduce myself. I'm Arjent Karple, senior technology correspondent at CNBC based
[01:02] in London. Um, and I've got a very special new co-host with me. Hey there, Arjent. Yeah, Steve Kovac here. Um, I cover tech over here in New York. I mostly focus on Apple and Microsoft, but look, I've been covering the tech industry for over 15 years now. I kind of have a good grasp on everything and I'm so excited to be here with you Arjun because I've just admired your work from across the ocean for so long and now we actually get to kind of collaborate and do this thing together. Uh I think it's going to be a good time. It's going to be so fun Steve. So between us, we think
[01:33] we've got nearly three decades of experience covering tech. And the crazy thing is we've got so much to learn. And I think over the course of us doing this podcast, we're going to learn so much, speak to so many uh interesting people. I'm so excited that this first series we're kicking off with an insight into Google DeepMind, one of the world's leading AI labs as well. Um and just for our listeners and our viewers, a a quick intro, I guess, to uh Google DeepMind. It was a company founded in 2010 here in London where where I sit as well. Very small company founded by three people.
[02:04] Deisabis, Shane Leg, and Mustafa Sleman who who's at Microsoft now, right? >> Yeah. And in fact, I interviewed him uh god nearly a year ago now. Uh Mustafa Sullean. Uh he's basically doing what Deis is doing over at Google. And it's just kind of interesting to see how Google was like this incubator, so to speak, for all of this top AI talent around the world. Demis obviously stuck around. He's running deep mind over there. Uh what I also think is really interesting though is just this AI moment Arjun we've been living through
[02:35] for the last three years and how three years ago chat GPG comes on the scene and Google was kind of seen as under threat. They went through this code red. They had to go through a bunch of reorganizations internally. Eventually Demis came out on top as the leader of AI. And guess what it 2025 was a really interesting year for AI over at Google. they kind of caught up and in some ways even surpassed what chatbt was already doing. And this is really interesting because the fundamental technology for
[03:05] all this these large language models we've been talking about for so many years started at Google and the perception was Google let chatbt kind of take that technology and run away with it. But now in my view at least Gemini is pretty much on par if not better than Chat GBT >> and Google DeepMind is integral for this. I mentioned it was found in 2010. Google actually acquired DeepMind in 2014. I was very new into my career as a tech reporter as well. Google paid around 400 million pounds uh for Deep
[03:37] Mind at the time in 2014. About $540 million. It's a stake this day that could be worth tens of billions, maybe hundreds of billions of dollars according to some estimates today. And DeepMind really is very much responsible for Google's AI. We talk about Gemini, the the the chatbot, the the AI um that that Google's released to consumers. This is powered so much by the technology coming out of DeepMind. But even before all of this, DeepMind was having some big breakthroughs. There was a a big moment a few years ago when they
[04:09] released a system called Alph Go. This was the first computer program that was able to defeat a world champion in a game called Go. This is a very complex uh game and it was seen at the time as one of the grand challenges uh of AI because it was such a complex game with so many different combinations available. The other big breakthrough of course was was something called Alphafold. This was another AI system developed at deep mind that could accurately predict 3D models of protein structures and the idea is here is if you could do that this may lead to some
[04:40] medical breakthroughs. So this advancement of science uh has been pretty core to what uh DeepMind's been up to and clearly it was a significant bet from Google more than 10 years ago because it's helped turn Google into an AI world leader today. >> Yeah and that that's exactly right and what really struck me about Deep Mind having watched them for so many years is how rooted in science they were. They weren't necessarily trying to build consumer products like they do now. They were really trying to solve fundamental
[05:10] problems in science and really usher in this era of AI powered drug discovery of other big complex problems like climate change. I know Demis talks about that a lot and he's going to talk about that in your conversation as well Arjun. >> Absolutely Steve um look it's a great scene setter for Deep Mind. So let's get into the conversation with its CEO Demis. >> Demis thanks for joining me on the tech download. Appreciate it. Thanks for having us. >> Uh Dis, we're going to try to get through a lot in our time here, but I
[05:40] want to start first with the technology itself. And we've been talking about AI and we've been talking about the the capabilities and how they've been continuously improving um as well. Now, in the tech world, I know there's a lot of conversations about how good can these models get, how good can these systems get, and there's a lot of debate around this idea of of scaling laws. Um for our for our listeners, you know, it's this idea of of more compute, more data, bigger models. uh eventually will lead to bigger systems as well. You said we need to push scaling laws to the
[06:10] maximum. Um >> there's questions over now. Are we hitting any kind of walls in terms of progress of those scaling laws in terms of the ability for these models to get better? And just from you know what you've been developing here at DeepMine, what are you seeing? >> Well, look, I think scaling laws um are going very well. So we're definitely seeing increased capabilities by putting in more compute, more data, uh, and making these models generally larger. So that trend is continuing. Um, may be not as fast as it was a couple of years ago.
[06:42] So um, there's some talk of diminishing returns. Uh, and and but but there's a big difference between sort of no returns and exponential. And I think we're somewhere in the middle where there's very good returns and that's worth doing. Um on top of that if I to you know in terms of like getting all the way to AGI artificial general intelligence um you know maybe that there's one or two uh big innovations still needed as well and maybe missing in addition to the scaling up of um kind of the existing ideas. >> We'll get on to AGI very shortly but
[07:13] what what are missing in your view? Well, if you look at I mean we've all you know played around with different chat bots and you can see that uh you know they can do very impressive things um in some dimensions but they're kind of like jagged intelligences I like calling them in the sense of like they're very good at certain things but there are other things that they don't do they're not capable of at all and um and if you pose a question in a certain way you you know you find that they're flawed um and they they can't do some relatively simple things and so for a
[07:43] true general intelligence you shouldn't see that inconsistency should be consistent across the board. Um, and also there are things like it can't continually learn. It can't learn new things online. It can't truly create original things. So there's quite a few capabilities that you would like to see and you would need for general intelligence that are missing from today's systems. That's really interesting. So what what would be the sort of unlock to get to those intelligent systems? I just want to quickly discuss a conversation I had with with Thomas Wolf who's the co-founder over at Hugging Face and he was talking to me um a few months back
[08:13] about his view on LLMs in particular large language models and just saying they're really great and you know you use these chat bots and the chat bots say hey great question great idea um and here's here's all the information you need to know but what's missing is the ability for these systems to come up with new and novel ideas perhaps and particularly I know you're so interested in science and what AI could do to unlock new drugs or discover new diseases etc. Um that actually maybe the LLM's limitations are there that you can't come up with these Nobel Prize
[08:43] winning ideas, these novel ideas. >> So perhaps there needs to be some sort of new architecture. What's your what's your thinking on that at the moment? >> Yeah. Well, look, my passion for and my whole reason I I've spent my whole career on AI is I think eventually it will be the ultimate tool for science. And of course, we've shown that with things like Alphafold and all of the science work we've been doing over the last decade, but there's still a long way to go uh in terms of uh can an AI actually come up with a new hypothesis itself? not just solve a conjecture that is already out there which would be
[09:14] already useful and impressive but can it actually come up with a new conjecture a new a new idea about how the world might work and so far um these systems can't do that they don't really have the capability to do that so there seems to be something missing um I think uh uh uh some of the capabilities that are required are kind of long-term planning better reasoning maybe also the idea of a world model uh this idea of like you know the system actually understanding better the physics of the world so that it can run simulations um you know kind
[09:44] of in its mind uh to test its own hypothesis you know these are uh things that you know the best scientists do human scientists do uh and so far our AI systems you know are not able to do that >> can you just help us understand a bit more of this idea of world models because it may be a term people are hearing for the first time you know how that guess they differ from LLMs language >> so LLMs and and the models we use at the moment are you know mostly around text um Of course, things like Gemini, our our foundation model can also cope with
[10:14] images and video and audio. So, different modalities. Um, but it's still actually understanding the physics of the world, the causality of the world. You know, how one thing affects another thing. Um, can you plan a long time into the future? These are all related concepts. And if you really want to understand how the world works so that maybe you can invent something new in the world or explain something about the world that was not known before which is basically what a scientific theory does then you have to have uh this this accurate model of how the world works.
[10:46] Um you know starting with intuitive physics and and and how the physics of the world works but all the way up to biology you know and and and economics. >> Yeah. And and do you envision the world if we get to this idea of artificial general intelligence this sort of human level of intelligence that that there will be a combination of LLMs and world models working together or will sort of world models supersede in some sense LLMs? >> No, I think there will be some convergence of these technologies. That's at least my betting is is um uh there will be these LLMs or foundation models you know like Gemini under the
[11:17] hood that will be a key component. I think the question I think there's almost no doubt about that in my mind which is why uh we must try and scale those systems as as as big and as as powerful as we can but the question is is it the only component that's needed for an AGI and um that's where I think I suspect uh other types of technologies and other types of capabilities will be needed and I think these world model capabilities and we're working on our versions called Genie uh and uh and we
[11:47] have video models like VO state-of-the-art video models that you can generate videos from from text and you can think of video models and and interactive models like Genie as kind of uh you know early embionic uh world models where if you can generate something that's realistic about the world then in a sense your model understands that about the world otherwise how could it have generated it demos you mentioned this AGI artificial general intelligence I know there's various definitions of it floating around you've previously said you believe that reaching AGI could
[12:18] somewhere in the in the realm of 5 to 10 years away. Um, is this still your view given, I guess, some of the profound developments we've seen in 2025? >> Yes, I think we're right on track from that. Actually, when we started Deep Mind back in 2010, we thought this would be a 20-year kind of mission to to build AGI, uh, you know, a system that's capable of exhibiting all the cognitive capabilities we we we have, including, you know, things like, uh, uh, uh, true innovation and creativity, um, and planning and reasoning and things like that. And I think we're about 5 to 10
[12:50] years away from that. Um, but that's, you know, pretty incredible if you think about how transformative a technology this is. >> You mentioned there might need to be some more technology breakthroughs. We're seeing things like the models advancing. We're seeing the semiconductors advancing rapidly as well. Are there any currently bottlenecks and things you need to figure out? I know energy is something that's been bought up so much saying, well, look, we can keep advancing chips. we can keep advancing models, but at some point >> we're just not going to have enough energy to run these data centers, to run these AI models. >> Um, >> well, look, look, there's there's lots
[13:20] of physical constraints. So, um, of course, there's, you know, no one ever has enough chips and, you know, we're lucky that we have, you know, our own TPU range in addition to GPUs and um, but there just aren't enough uh, compute chips in the world really for the demand. Uh, and of course, in the end, that comes down to energy as well. there's this idea of energy will be effectively is synonymous with intelligence as we get into the era towards AGI. Um now the interesting thing is I think that AI itself will help here in the sense of getting more efficiencies out of existing
[13:50] infrastructure but helping with things like material design better better solar materials but it could also help with new breakthrough technologies like fusion. you know, we have a collaboration with Commonwealth Fusion uh in the US to help contain plasma and fusion reactors and um one of my pet projects is can we come up with a room temperature superconductor uh material using AI. So I think there are multiple breakthroughs that AI could come up with and help uh us come up with that would help with the energy uh situation. In fact, indeed that's I think that's one
[14:20] of the most promising use cases of AI. Um and then the other thing is as these systems are getting better, they're also getting, you know, 10x more efficient per year. So if you look at our range of models, we have our kind of lighthouse model, our pro versions of Gemini, but then we have our flash versions which are way more efficient and the sort of workhorse models that are used for everything. And um they use techniques like distillation where you have a big model that teaches a smaller model and the smaller model is really really efficient. And I think there are more and more innovations and techniques like that that will keep bringing the
[14:51] efficiency curve uh down and so you get you know much better performance per per watt. We hear a lot about sort of AGI and I think there's a lot of people wondering technology sounds amazing sounds great but there's al also a lot of fear right around uh the proliferation of this technology and the impact it's going to have on on people every day and their lives. Um I guess for you what what are some of the the things we need to consider? Yeah. From from that perspective in terms of the impact on society, whether it's around jobs, whether it's around kind of what we're going to do with our time if if we reach this goal versus I guess the
[15:21] benefits that you believe this technology is going to bring for humanity. >> Well, of course, you know, I I believe that overall AI is going to be one of the most beneficial technologies and humanity's ever uh uh invented. Uh that's why I spent my whole career working on it. But it's only, you know, it's not a given. It's a dual-purpose technology. Um, I dream about using AI for things like curing diseases. We have a spin out called isomorphic that builds on on alphafold work on protein folding work that we did a few years ago to accelerate drug discovery and try and
[15:52] solve all disease. I think that's now you know within reach that type of thing in the next decade or two. Um, we've discussed energy. There's many benefits I think AI is going incredible benefits AI is going to bring. Um, but there are also risks. Obviously there's kind of economic disruption. Um and I think there it's going to be like the industrial revolution but maybe 10 times bigger 10 times faster. So you know it's incredible amount of transformation but also disruption that's going to happen. And you know we need some uh new economic models probably for that. Um
[16:24] and then on terms of the the worries about the usage of AI I have two which I think are are worth worrying about. one is bad actors repurposing these general purpose technologies AI technologies for harmful ends. Um and then the second one is AI itself as it get we get towards AGI and agentbased systems. So these are systems that are able to do things more autonomously than than today's systems. Um they can you know what are the guard rails around that? How do we make sure we can keep them uh uh doing the things that we want them to do and not veer off
[16:55] into uh something that we didn't expect. And so those are the two kind of risks that are kind of uh that I foresee. >> Do you feel that you're developing systems that you can be in control of? >> I think we're we're very confident about that. You know, we we've had uh and thought about responsibility and safety and security of these systems from the very beginning. Um you know, we started Demi back in 2010. Almost no one was working on AI back then, but we planned for success and we knew success would mean these extremely powerful systems.
[17:25] So we also understood the the the the other side of the coin of that. So from the very beginning we've tried to be very thoughtful use the scientific method and scientific approach to try and understand as much about our systems we're building before we deploy them. Um of course that doesn't mean we won't make any mistakes. There's too it's too it's it's such a incredible and fastmoving technology. But I think with with something like AI we need to be you know I call myself a kind of cautious optimist. I'm I'm I'm very uh big believer in human ingenuity. I think given enough time and care, we'll get
[17:56] this right as scientists and as a society, but it's it's not a given. And so, um, we shouldn't be sort of rushing into this. Um, and and we need to go into it with our eyes open because I I guess the reason I ask it because I know you've spoken to people like Joshua Benjio and Max Tegmark and and these are people I've also spoken to and and they're of this cohort that believes do do we need to be rushing so quickly into a world of AGI and agentic systems? Maybe we need more uh toolbased uh AI, AI to solve specific things rather than these allpurpose or general purpose kind
[18:27] of systems and I know they've called for for perhaps a slowdown to the development of of these AGI systems. >> In your view, do you think you should be slowing down? Well, I've I've had lots of you know, I know them very well. Yoshua and Max we've had many discussions and many others and and actually I have some sympathy for that view that you know building a tool based AI is you know thinking of AI as a tool or the ultimate tool for say science is the right way to build AI in the initial stages um and uh certainly that's the
[18:58] way we're viewing it and the kinds of things we apply AI to uh like AlphaFold but um the thing is you know it's a very complex geopolitical and corporate uh system that we're in And it isn't just about you know there are many companies trying to build this there also many nations trying to build it and um it's there's a sort of race dynamic which I ideally wouldn't be there. So in an ideal case this would be a scientific endeavor and it would be very carefully uh each step would be carefully considered but unfortunately the the the
[19:29] pra the real world isn't isn't like that and we have to kind of be pragmatic about uh where we are. So what we're trying to do is be good role models for um yes being on the frontier pushing that uh the benefits of that as quickly as we can and as broadly as we can um but also try and be as responsible as possible with that along the way and thoughtful as possible and I think we've got that balance pretty pretty good right now and hopefully that's a bit of a role model uh to the rest of the field in the industry too. Yeah, I want to address some of those dynamics as well, but just just first I guess just from a
[20:00] personal point of view, have you ever you said you sort of started this mission of deep mind, you know, you believe in the technology, but has there ever been any moments in your career when you g like should we be doing this? Um, look, you when you look at how powerful the technology is. Um, I really think there that there are so many challenges confronting society today, not to do with AI, climate, poverty, you know, the access to water. There's a there's just so many uh issues um health uh aging, population, uh disease. So
[20:31] like uh uh um you know energy we talked about earlier. So if a some if I if there wasn't a technology transformative as AI coming down the road, I'd be really worried about uh society's ability to deal with these challenges. So, interestingly, AI itself is one of those challenges, maybe one of the greatest ones, but it's also one which can help us um cope with and resolve and solve some of these other big grand challenges. So, it's a very interesting one, right? It's it's it's sort of double-edged and I've always believed in
[21:02] that. I've always thought that um uh in the end it would be the the the the most important technology uh we'll ever invent. And um I think it's sort of the natural progression really of the computer age. >> Dennis, you just just a quick aside, you started uh life in gaming, which is amazing. Co-developing theme park. Fantastic. >> Fantastic game as well. Um did you ever do you still play games? >> Yes, I love games. It's my main and only hobby really. Well, like these days like League of Legends with my two two boys
[21:32] and my brother and we have a little team. We've done it since lockdown. Um but yeah, I love games in all its forms from from football to >> such a high impact stressful role as you have potentially. Is that your unwind? >> It is. It is. I would say so. And it's also, you know, it's a it's a kind of in the past as well as being a great creative endeavor for me, you know, and it's how I learned programming and other things was was through making games. >> I have nowhere near as a stressful a job as you, but that's my unwind, too. >> Yes, for sure. >> Get home, turn the console on.
[22:02] >> Exactly. Exactly. Just in that small segment alone, Steve, there's so much to unpack and I want to focus on on two kind of big buzzwords right now. The first is artificial general intelligence or AGI. This idea, and I know there's so many different definitions of it, but broadly speaking, this idea of AI that that is as smart or smarter than humans. And I think that so many of these big AI labs including Open AI, including Deep Mind, are pushing and
[22:32] hoping to get to this stage of AGI. A and so far they've approached this with a a technique called large language models. These AI models that are trained on huge amounts of data, but mainly text. But there's this other buzzword, right? World models. This idea of these AI models that understand the physical world. And this is this this buzzword is really growing in popularity, right? >> Yeah. And I think this is going to be a big theme of AI going into uh the rest of 2026 and even into next year because
[23:03] the idea here is that LLMs, sure, we got the language part down. It can mimic the way humans talk and and speak and and write and things like that. Uh but when it comes to the physical world, you know, we talk so much about robotics and AI and physical AI. Well, they need to understand how the physical world works, how water flows, how air moves, and things of that nature. And what really struck out to me when you brought this up to Demis, he he said, "Yeah, we do need to start exploring that more." And in fact, he sees a world in which the
[23:33] LLM and those world models start to converge. I think that was the word he used, converge uh into something uh more unique and and more powerful and capable. This is also a debate that's been playing out among AI leaders like on social media. You could fire up X or your favorite social media site and uh what really struck out to me is Yan Lun. He was the head of AI for many years over at Meta. He recently left uh to start his own thing because he kind of got superseded by Alexander Wang and that whole big talent wars that happened
[24:04] over last summer. um he had a really interesting interview in the Financial Times. He doesn't think LLMs are what's going to get us to AGI. To your point, that's what everyone's chasing, the super intelligence, AGI, whatever you want to call it. His thing is LLMs can only get you part of the way. You need world models and all sorts of other things. And he kind of uh harshly criticized Meta for not thinking beyond the LLM. Um and that seems to be part of the reason why he left to do his own thing. And it's really interesting to see one of Meta's big competitors,
[24:35] Gemini, just talk openly about it and say them is saying, "Yeah, we we need to do this. We need to start thinking about this." Uh it enables so many things from robotics, autonomous driving um and just a better understanding for these AI models and um intelligent systems that we're chatting with to get you that right answer. >> Steve, do you ever use a a chatbot um and you put something in and it will say, "Hey, Steve, great question. That's a really clever thought." >> All the time. That's the sickop fancy of all these chatbots, right? Where they're like, "Oh, you're so smart and great at
[25:06] asking me these questions." Yeah. All the time. >> Exactly. Because the reason I bring that up is is partly to this point, this growing criticism of LLMs is that actually, yes, they're great and they'll give you the information and but actually when it comes to LLM as a foundation for being able to create new ideas, novel ideas, there's limitations there. And I think that's partly what uh Demis was speaking to and why this idea of world models is really growing in popularity. Um it's going to be interesting to see how this plays out as you mentioned into this next phase of AI where uh it's key for things like
[25:37] robotics, driverless cars, and many other use cases, too. >> Yeah. And I'm you'll you'll notice as we continue this podcast, I'm incredibly cynical about the robotics angle of this AI moment we're living in. All that so many of the robots we're seeing, they're literally puppets. They're teleoperated. The best example, of course, is the Tesla Optimus robot, uh which started out as a man in a bodysuit dancing around. Now it's a real robot, but again it's tea operated. There are literally people in a control room controlling it over the internet and even using their
[26:08] voice to talk to you and things like that. So we are the robotics people I talked to, we had one in the office just a couple weeks ago and they said the hardest part isn't building the actual robot, it's training it and that's where these world models are going to come in so they can actually operate autonomously like we've been promised. Deus, you mentioned some of the dynamics at play, right? And competition >> commercially, of course, is one of those. We've got Open AI, we've got Anthropic, we've got all these different
[26:38] AI labs um out there. It's intense. Uh and Gemini 3 has had such good reception uh so far. Um but there was a point people were doubting >> Google as a whole and its ability to compete and I say a point it was at some point in 2025 and it wasn't that long ago and then you know, Gemini 3 really came out and and impressed a lot of people as well. Um but it's a space that's ever changing. Uh so how are how would you assess right now the competitive environment? How do you feel it? >> Yeah. Well look it's a ferocious uh uh competitive environment at the moment. I
[27:09] mean many people who are telling me you know been in tech for 20 30 years say it's the it's the most intense environment they've ever seen perhaps you know ever in the technology industry. and uh and and and you know all the I guess most capable players whether it's individual you know tech titans or big tech companies or and or the best startups they're all involved in this space now cuz I think everyone has understood what we've known for 20 plus years now that this is really the most important technology um so that's sort of to be expected but it's tough
[27:39] but it's also exciting and um you know going back to games I uh I sort of I've started playing chess when I was very young for the England and junior chess team. So I've kind of been brought up in in competition. So you know I love competition fortunately. In fact many ways I live for competition. So a lot of a big part of me sort of like likes to lean into this. But on the other hand the only thing I would say is at the back of my mind I know there's something much more important than individual competition between companies or even countries which is overall getting stewarding AGI well for the world for
[28:10] the whole you know for all of humanity. And I think that's incumbent of all of us who are leaders of the AI labs. um and uh and and can have an influence over this is to have that at the sort of in the front of their minds in amongst this sort of ferocious capitalist competition that we're in as well. So both are true at the same time. >> I mentioned kind of the moment people were questioning what Google was going to do with with AI earlier in in the year. >> Did you do anything different? Yeah, I think look I I feel like you know if we go back over the last decade actually
[28:41] you know Google Google brain specifically uh the research division in Google and deep mine as it was uh sort of fairly independent we kind of invented about 90% of the technologies uh that everybody's using today you know whether it's transformers of course most famously the architecture behind all the LLMs or AlphaGo you know sort of introduced reinforcement learning at scale uh on a really hard problem so we've invented all this technology but then um maybe we were in hindsight we were a little bit slow to commercialize it and scale it and um you know that's
[29:12] what open and others did very well and then the last 2 three years I think we've had to come back to almost our startup or entrepreneurial roots and um be scrappier be faster ship things really quickly and um and and sort of make really rapid progress and I think what you're seeing over the last couple years culminating in Gemini the Gemini series which we're very happy with Gemini three is as as you mentioned our latest version um has sort of put us back at you know near the top of you
[29:42] know the top of the leaderboards where we feel we belong and you feel like you can stay there >> I I I feel like we can stay there of course yeah >> amid all this competition there's obviously a lot of talk about >> bubbles in AI uh particularly around valuations of certain companies companies raing raising astronomical sums of money the tech giant spending hundreds of billions on infrastructure uh and companies out there quite frankly raising large sums of money with very little product or or or even very little profitability if any. And so where do
[30:14] you think we are right now in terms of this this kind of bubble discussion? Do you think we're in a financial bubble when it comes to AI industry? >> I think it's not a binary thing this bubble discussion. I don't I think um some parts of the industry might be in a bubble to me that's what it looks like and and others probably not. you know, fundamentally AI is going to be the most transformative technology ever invented. So that's there's that part that underpins everything. So in the end, it's a bit like the internet bubble in the end. The internet was critical and there were some generational companies
[30:44] that were created in during that time, right? Um so I think you know that's sort of almost inevitable. There'll be overexuberance once everyone realizes how transformative a specific technology is. uh and then there'll be probably a reckoning and then the the things that are real will survive and and flourish. Um where it seems to me is you know maybe like in the private markets where there sort of seed rounds at tens of billions of dollars where basically there's just almost nothing there yet and that seems a little bit unsustainable over the long run. As far
[31:15] as I'm concerned I don't really worry about bubbles from my my point of view is sort of leading Google deep mind. I've got to make sure that what whichever way it goes, whether um it continues to go all rosy and exponential like it is now or there's a bubble, you know, there's some kind of bubble bursting, that we're in the right position to to to win either way and to take advantage of that either way. And I think we've got a good position given Google's underlying business and how AI fits with that. Um uh to to to benefit uh whichever way it goes from here. some I guess some of your biggest competitors
[31:46] are the ones who have managed to raise huge sums of money in the private markets at this point. So do you feel confident that even if there is some sort of correction at some point that you know you'll be able to weather it out I guess? Yeah, I mean look, you know, that's the whole point of uh Google's balance sheet and and also all the incredible products that um and surfaces that that we have. You know, I think it's you know, dozens of multi-billion user products and and AI kind of naturally fits into uh all of those products, whether it's um you know, email workspace or or you know,
[32:17] new things like the Gemini app. >> Yeah, you mentioned dynamics at play as well. We talked competition. And the other one is geopolitics which you mentioned as well when huge discussions around China of course in this kind of competition battle between China and the US. But you know there was a point where people were discounting the ability of China and it and its companies to come up with strong AI um models and and and technologies. But actually we saw with kind of what Deep Seek did um it kind of brought a bit of shock to world but actually more than that some of the big
[32:47] tech companies like Alibaba coming up with some very competitive open-source models. So China's not out this game, right? >> Not at all. And actually, you know, I think they are closer to the US front, you know, US and West frontier models than maybe we thought one or two years ago. Um maybe they're only a matter of months behind at this point. Um the interesting thing is and they're very there's some very capable teams of course like the Deep Seeking and Alibaba you mentioned. Um and uh the question is is can they innovate um something new
[33:17] beyond the frontier? So, I think they've shown they can catch up, you know, and and be very close to the frontier and catch up very quickly. Uh, but can they actually innovate something new like a new transformers uh, you know, that gets beyond the frontier? I don't think that's been shown yet. >> Is that going to be in your view difficult because of restrictions on access to technology like leading edge chips for example? >> No, I think it's more a mentality issue you know. So I think it's something that at least the leading labs the leading frontier labs in the west have uh
[33:47] nurtured I can say for ourselves you know we you can think of deep mind as a bit like a try to be a modernday bell labs and encourage uh innovation and exploratory innovation not just scaling out what's what's known and and uh today and of course that's already very difficult because you need world-class engineering already to be able to do that um and and China definitely have that the question is uh is the scientific innovation part that's a lot harder to you know to invent something is about 100 times harder than it is to
[34:17] to copy ed is and I haven't seen evidence of that yet but it's very difficult so one of the most striking parts of that part of the conversation for me Steve was uh around China um I used to live in China for just over 3 years report out of China uh for CNBC covering the tech sector there and there was this growing view recently that actually China is so far behind the US when it comes to AI uh for for multiple reasons.
[34:49] One of those is that oh it may not be able to get its hands on the most advanced chips so its industry could fall behind. One view is that it's just not innovating and it doesn't have the capital the way US companies do. But actually what was really interesting from Demis is he said that he believes Chinese AI models are are just months behind uh where the U US is. So actually not far behind and remember when uh last year we had uh Deepseek really shock the world and markets. Um it showed I think China is in the game and since then whilst Deepseek hasn't quite made the
[35:20] waves it did uh when it first kind of came out um Alibaba one of the world's biggest or one of China's biggest tech companies um has been a leader there. gets developed some really interesting models which if you look at the open- source community such as uh on a site called hugging face you see Alibaba's models are amongst some of the most popular experts who I've spoken to in the space say they're amongst some of the most advanced in the world so you are seeing there and one of the things I can tell you just from living and
[35:50] working out there is Chinese companies move fast they have the expertise and they can innovate so you can't discount them out of this kind of AI race but also take Demis' point that he said whilst the Chinese companies are sort of catching up and and and are very much in this race, one thing they haven't proven is their abilities to kind of make these big breakthroughs. So, you know, I thought that was a really interesting and nuance for you. I guess the other part here, Steve, is something you picked up on is Dis's comments on bubbles and AI bubbles.
[36:20] >> Yeah. And that and by the way, just talking, let's go back to what he said first about the months thing. uh Deepseek a year ago. It wasn't just about the fact that China can do it and make a really good large language model or a chatbot. It was also the idea that they did it without the most powerful Nvidia chips that kind of rattled the markets as well. And that's what we're seeing here in the United States now, Arjuna, is trying to limit China's ability to get those NVIDIA chips. There's all this talk about maybe they'll get those H200 chips, which aren't the best chips, but they're better probably than what China has
[36:51] access to. And then you get into the whole smuggling thing. But to Dennis's point, you know, if they really are months behind without full access to these chips, you know, that kind of questions Nvidia's prominence and dominance uh in the chip space as well. But yes, what what you said about the bubble is also super uh interesting too because you asked him about that. Are we in a bubble? What do you think? All this sort of things. And he basically said we're Google, we're rich. It doesn't matter. We have the money. We have the free cash flow to spend this. Our
[37:21] balance sheet is our superpower. If for some reason we need to rein back the spending, we can do it and we'll be fine. But guess who can't do that? That's OpenAI and Enthropic. The other two leaders, XAI, we can throw them in here, too. Their whole thing is they have to raise money indefinitely in order to get to the point where they can finally show some revenue and and revenue growth to uh sustain themselves without continuous fundraising. If things start to dry up, OpenAI and Enthropic are at extreme risk. Google,
[37:52] Microsoft, Meta, they have the cash flow to move on to another project. Meta's already done it with the metaverse. These companies can pivot very easily because they have these big high margin businesses already. De um a lot of people I guess forget how much of Google's AI capabilities come out from DeepMind and and yourself and your teams. Um how do you work with Google? There's a lot of fascination around that. call you up one day and say, "Hey, Deis,
[38:22] we need this thing or we have this idea for Gemini or for some other AI product. Um, can you build it?" How was that relationship? >> Yeah. So, the last three years we've combined everything together as into Google Deep Mind. This this one entity that that that all the AI research at Google goes on in and it's a kind of combination of uh Google research, Google Brain and and and DeepMind. And I run that group and it's it's like the engine room of Google. You should think of it like that. So uh all the AI technologies is is done by this group by our group and then it's diffused across
[38:52] you know all of these incredible products uh right across Google and the last couple of years we've been building that backbone so not just the models but also almost rearchitecting the entire infrastructure of Google so that it can you know these things can ship incredibly quickly these models it's almost sim shipped to all the main surfaces so you know when we release a new Gemini model it's there the next day or the same day in in search and uh and that's been going really well. I think I would say we've really got into our groove uh with the 2.5 Gemini models and
[39:22] and for the last sort of year uh that's been coming really uh a smooth process now and I think you'll see that more uh over the next next 12 months. Um and so you know we think of ourselves as the and describe ourselves sort of as the engine room for that and you know Sundar and I pretty much talk every day about strategic things and where should the technology go and what does uh uh the wider Google need um and then you know we adjust the road maps and the plans uh you know on a daily basis whilst keeping in mind the long-term goals of uh you
[39:53] know getting to AGI first fast and safely. So we should we should expect more of the ability to come up with with new things, new AI tools and that be shipped across the Google portfolio etc. because of that kind of change you've made in that relationship. >> That's right. So it's an incredibly uh tight sort of uh iteration loop and and and and you know we're all on the same tech stack and so on. A lot of what you're building is going into Google products, but I know kind of covering companies like Samsung, you help companies like Samsung to build out some of the AI tools um within, you know,
[40:23] their smartphones for example and that kind of thing as well. Well, look, we work with a lot of partners as you as you mentioned um you know, we're very proud of the fact that our technology selected by those partners um because they see how capable it is and and actually you know, it comes to Samsung and other devices. Um, I think there's really interesting way I'm very interested in the idea of uh uh edge compute and and faster versions of these models working on these edge devices be those phones, but also new devices like glasses that we're working on um and you
[40:53] know partners like WBY Parker and the idea of smart glasses and I think um Google's worked on smart glasses for a long time as you know but I think the day you know finally we have the killer app I would say for it which is this idea of a universal assistant and um and uh and and sort of helping you in your everyday life. And I think all the all the all the big uh device players are going to be interested in that type of technology. >> Demis, we've only got a few minutes left, but I do want to ask a little bit about I was a brand new tech reporter when Google bought DeepMind 2014. I
[41:25] think it was a 400 million pound deal back then. Um so many people didn't know what you what you did. Um and why is Google buying this British company? What's going what's going on here? Um, do you ever look back to that and and think, "Oh, maybe we should have stayed independent at all, or are you happy with how things have turned out?" >> Well, look, I we I knew it's funny. So, so the the head of search at the time, Alan Eustace, he he was sort of in charge with Larry Larry was sponsoring the Larry Page was sponsoring the the deal uh as he was CEO at the time, but Alan Eustace was delegated, the head of
[41:55] search to kind of close the deal. And I did tell Alan that this would be the most important acquisition Google ever made, which is which is quite something given they've, you know, there's YouTube and and and uh Adwords and other things that they they previously acquired. But I kind of knew how important this was going to be. Uh and also how good a fit it was with Google's um uh mission, which is organize the world's information. And AI is a very natural fit to that uh and organizing and understanding information. I mean, what better tool than AI for that? So, I kind of knew
[42:25] that would be a natural fit. And we sort of knew that this, you know, maybe it's now worth, I don't know, 100x thousandx of, you know, what of what we sold it for. But the thing is, I wanted to get back to the science at the time and and and push forward the research, which was still very nent back in 2014. And and you know, fair play to Google is they were one of the few companies in the world, I think, that could recognize uh and specifically Larry at the time, how important this technology was going to be, what it could become, and what we see it for it today. And I don't think we could have done the the great work we
[42:55] did with Alph Go and Alpha Fold and all the science we've done um and uh if we hadn't had their backing and the amount of compute that they could bring uh uh to to play. So I don't have any regrets at all. >> So tech CEOs, AI CEOs, new rock stars of the world. I've seen Jensen Hang here in Europe and the CEO of Nvidia, you know, being followed around by everyone as well. Um Jensen I think said recently that that you and him talk he had great things to say about Nano Banana the new image generation 2 as well. What what what do you guys discuss?
[43:26] >> Oh we disc I mean Jensen's great you know he's incredible pioneer also somebody you know I admire him for sticking to his vision for 20 30 years now. In fact I first started using GPUs in the '9s on for gaming of course for for for writing graphics engines and physics uh engines. So it's funny that it's come full circle to me that that you know my my early gaming days even the hardware that was pushed then is now useful for AI ironically. Um but yeah we talk about he's very interested in science and AI for science and actually
[43:56] you know alpha fold was trained on GPUs. So we and he loves Alpha Fold and the work that we're doing you know uh in drug discovery. So we mostly talk about um AI for science. I I know a lot of the data centers are built in Nvidia systems, but I know Google also has its its tensor processing units, TPU chips. Is there any kind of competitive friendliness there? >> Yeah. Well, look, we we're lucky we have our own we love our TPUs. We we generally use them internally for training our um our best models and actually we found there's a big demand
[44:26] for that from the elite AI teams uh who are trying to build large models or serve uh very large AI models. uh they're specifically built for that. So TPUs are sort of they're a little bit more special case than GPUs. You can think of GPUs as being more general. So you know maybe we would use a GPU when we're trying to um explore some new architecture uh like Alpha Fold was or some new application. Um but then once we're when we're trying to sort of um scale to the maximum things we know then
[44:57] um you know custom silicon can be a lot more efficient. Um so we're lucky we have we have both. We get to use both here at at Google and Deep Mind. >> Great. Damus, just looking to the future, you're obviously so focused on science and the potential for AI to create new drug breakthroughs, do discover new diseases, lots of potential things there. Um, you've also got isomeorphic labs, of course, as well. Where are we on this path to your your vision of of AI unlocking all of these
[45:27] these kind of breakthroughs in the world of science? Well, look, I I I love I always point to AlphaFold as probably the best example so far of AI applied to science. You know, I'm very proud of that project and you know, we solved a 50-year grand challenge in science of protein folding, how the structure of 3D structure of proteins and over 3 million researchers around the world are using it in their critical work. So, I can't imagine a more transformative sort of technology. Um, and what I would love is to see have be able to point to a dozen alpha folds and, you know, each of them
[45:59] revolutionizing their area of science or mathematics. And I think we're well on the way to that. And we're working on half a dozen projects like that in material science, in physics, in in maths, in weather prediction. Um, and uh, and I think that the next 10 years, if if AI goes well and progresses well and we use it in the right way, it could usher in a new golden age of um, scientific discovery. What do you think are going to be the big things in AI in 2026? Any big breakthroughs, any big progresses that you think will happen? >> Agentic systems, systems are able to do
[46:30] things more autonomously are going to start becoming reliable enough to be useful. Um, I think we're going to see some really interesting things in robotics in the next 12 to 18 months. We're working really hard on some very ambitious projects with Gemini robotics. And then finally, maybe um, you know, AI assistants on devices. I think we're going to start seeing them really useful in the in the real world. Uh and then maybe the thing I'm most excited about is advancing world models further, making them more efficient so they can actually be used maybe for planning in our general models.
[47:00] >> Great, Damis. I'm going to take that last answer as a sort of teaser trailer for the next time you and I get to catch up hopefully at some point this year. Thank you so much for joining me, Dis. >> Thank you. Thanks for having me. So Steve, just in that final part of the conversation, I thought what was interesting is the relationship between kind of the deep mind entity and the broader Google business. And there was a part where Deis was saying he speaks to Sundar Pichi, the CEO of Google or Alphabet every day. Um, and and how sort
[47:32] of more integrated they've become. And I think if I'm thinking about that in this AI race, what that signals to me is that Google has clearly figured out how to become speedy at getting AI products to market. But also, you got to think about all these Google products, right? Whether it's Chrome, whether it's uh Gmail, whatever it might be, they are wanting whatever Google AI is being developed to spread all across of all across those products. that gives them an absolutely mammoth user base to kind
[48:02] of almost instantly tap into with some of these products. And I I've always I've said this for for a while now. I think one of Google's biggest strengths really is that when you think about the Android operating system and and you know how large it is, 70% odd market share globally. You know, that is a huge amount of people and devices where Google AI could be effectively installed on and used quickly. So they're in a good position in terms of going to market, I think. And and and clearly um this relationship between DeepMind and
[48:33] the broader Google business is going to be integral for Google to sustain any success over the over the longer run here. >> Yeah. And on the Android front alone, I mean, Samsung, the biggest manufacturer of Android phones, they're already putting Gemini is their main chatbot. Gemini is their main AI. I'm I was a little surprised Samsung didn't try to build their own which like they have in the past but no they've completely gone all in on Gemini. They're partnering with uh Google on those uh the new mixed reality headset that they have. There are some upcoming glasses that they're
[49:03] working on in partnership also uh with companies like Warby Parker to design them. Uh so yeah, Samsung has like really adopted this and that is a huge platform for Gemini. Just just that just the Samsung uh angle of it. Just that huge market share they already have is is great. And then let's talk about Apple. Gemini is actually going to be the engine that powers this new version of Siri we're expecting in just a couple months time. He did talk about his excitement to see Gemini kind of spread
[49:34] on more devices. So, I think it's a really smart move by Apple to kind of realize it can't build this on its own and honestly do what Samsung is doing and say, "Okay, let's just integrate this proven technology. We already have a great relationship with Google." And this is honestly a different kind of Google that I've been see that I've seen for so many years where you had so many different groups kind of working on the same thing. I mean before this big reorg and and Demis got all that control over all of AI, there were multiple groups within Google working on artificial
[50:05] intelligence uh kind of bumping against each other and Sudar Pachai was really smart saying we got to this is a huge moment. We got to reorganize everything. He folded everything under Demis Hbasus and put it into uh DeepMind and that's where we are now and it's it's really paid off in 2025 uh in a big way with Gemini 3. Yeah. And and that consumer space really is uh getting more and more intense when it comes to the the AI side of things particularly as you know you mentioned before when you were talking about some
[50:36] some of the talk about bubbles. These competitors like Open AI you know Google has um big uh balance sheet strong cash flow and it has a huge user base of users and it continues to innovate. And I think this really does given the kind of that reorg and and this kind of speed you're seeing now from Google, I think this is adding going to add a lot of competitive pressure onto OpenAI particularly on the consumer side uh in 2026. So it's all up for grabs. >> Yeah. And we're going to see a lot of different stuff I I anticipate from
[51:07] OpenAI this year. They're going to throw all the spaghetti at the wall they can to see what sticks because they've put enormous pressure on themselves to generate enormous amounts of revenue in order to fulfill all of these promises they made about you know capital expenditure build out of these big data centers with Oracle and all these sorts of things like that. Uh it cannot happen all these committed spending they have unless they productize it better and more effectively. But like to your point we're seeing this with Meta by the way. Meta has a huge opportunity to leverage
[51:38] its user base and it hasn't figured out how to do that in the way Google has. Uh so right now Google feels like they're kind of on top of things. Well, look, part two of this miniseries on DeepMind is going to be out next week and we're speaking to Laya Ibrahim who is the COO over at DeepMind. So catch that. And if you got any comments uh or or thoughts about this episode, please reach out to us. Uh you can reach us pretty much everywhere. I think uh you're you're on uh multiple media. You're a blue sky guy. >> Yeah, we're all over the place.
[52:09] >> No Instagram. I quit Instagram 7 and 1/2 years ago and I do not regret it. >> Wow, that's amazing. >> No more doom scrolling. Love it. >> No more doom scrolling for this guy. >> Oh, thank you all for listening and watching. We'll catch you next time.