This content is adapted from an interview with Demis Hassabis on the Sequoia Capital channel, publicly published on April 29, 2026. Original reference: https://www.youtube.com/watch?v=AFpeWo1GTeg
Summary: Demis Hassabis Interviewed at Sequoia Capital's AI Ascent 2026
The Deep Ties Between AI and Gaming: Games are an excellent testing ground for AI. By making AI a core gameplay mechanic, it not only effectively validates algorithmic concepts but also provides early computational support for tech development. The "Timing Theory" of Entrepreneurship: You should aim to be "five years ahead of your time, not fifty." Entrepreneurs need to keenly capture the sweet spot between technological breakthroughs and real-world demand; being too far ahead often leads to failure. The AGI Roadmap: DeepMind's mission is crystal clear and unwavering—Step 1: Build Artificial General Intelligence (AGI); Step 2: Use AGI to solve all other complex problems, including those in science and medicine. The Core Value of "AI for Science": AI is the perfect language for describing biology and complex natural systems. With AI-powered simulations, the drug discovery cycle could shrink from years to weeks, potentially even enabling truly personalized medicine. The Birth of a New Scientific Discipline: The very complexity of AI systems will give rise to a new engineering science, such as "mechanistic interpretability." Simultaneously, AI-driven simulation technologies will allow humanity to conduct controlled experiments on complex social systems like economics, thereby opening entirely new branches of science. Information as the Essence of the Universe: Matter, energy, and information can be interconverted. The universe itself might fundamentally be a massive information-processing system, lending profound significance to AI in deciphering the universe's underlying operating principles. The Computational Boundaries of the Turing Machine: Modern AI systems like neural networks have shown that classical Turing machines are sufficient to simulate problems once thought solvable only by quantum computing (like protein folding). The human brain is likely a highly approximate Turing machine. Philosophical Musings on Consciousness: Consciousness might be composed of components like self-awareness and temporal continuity. On the march toward AGI, we should first view it as a powerful tool, and then use that tool to help us explore the grand philosophical question of "consciousness."
Content Description
Demis Hassabis, co-founder and CEO of Google DeepMind and 2024 Nobel Prize laureate in Chemistry for AlphaFold, engaged in a wide-ranging and profound dialogue with Sequoia Capital partner Konstantine Buhler at the AI Ascent 2026 summit, exploring the path to AGI and the future beyond it.
During their discussion, he explained why he believes AGI is achievable by 2030, why the lengthy cycle of drug development might collapse from a decade to just a few days, and why we should view "information," rather than matter or energy, as the core, foundational essence of the universe. He also explored what Einstein might critique about today's AI models were he still alive, and why the next year or two will be a critical juncture for humanity's destiny.
Full Interview Transcript
Host: Demis, thank you so much for being here.
Demis Hassabis: It's great to be here. Thanks to all of you for coming. It's fantastic to chat with everyone.
Host: We are incredibly honored to welcome you to our chocolate factory.
Demis Hassabis: I just heard about that. I'm very much looking forward to trying the chocolate later.
Host: Awesome. Demis, let's dive right in. Today we have a true OG (Original Gangster) of the industry: he wears many hats—original thinker, founder, visionary—and is a pioneer in all things AI. Demis is a true believer and a true scientist.
Demis's Original Vision and Inner Thread
Our conversation today will start with the early stories of DeepMind's founding, then delve deep into science and technology, and finally open up to audience questions. Let's get right into it.
Demis, you have been a chess prodigy, a game company founder, and a neuroscientist. You are the founder of DeepMind and now lead a massive, hugely influential enterprise. These identities seem completely disparate, but you've mentioned there is a single unifying thread running through them. Can you share that with us?
Demis Hassabis: Indeed, there is a thread, though there might be a bit of post hoc reasoning involved. But my desire to dive into AI has been a long-standing one. I identified very early on that this was the most important and interesting endeavor I could dedicate my life to. From the age of 15 or 16, every field of study I chose, everything I did, was aimed at one day building a company like DeepMind.
Games: The Training Ground for Artificial Intelligence
I took a "roundabout" path into the games industry because, back in the 90s, that's where the most cutting-edge technology was being incubated. Not just AI, but also graphics rendering and hardware technology. The GPUs we all use today were initially designed for graphics engines, and I was already using the earliest GPUs in the late 90s. All the games I worked on, whether for Bullfrog or my own company, Elixir Studios, used AI as a core gameplay mechanic.
My best-known work is probably "Theme Park," which I developed around age 17. It's an amusement park simulation where thousands of little people flood the park, ride various attractions, and decide what to buy in shops. Underneath the game's surface, a complete economic AI model was running. Like "SimCity," it was a pioneering title in its genre. When I saw it sell over 10 million copies and witnessed firsthand how much players enjoyed interacting with the AI, it solidified my determination to dedicate my life to the field of AI.
Later, I pivoted to neuroscience, hoping to draw inspiration from the brain's workings to derive different algorithmic ideas. When the optimal moment finally arrived to found DeepMind, blending all these accumulated experiences felt like a natural convergence. Naturally, we subsequently also used games as an early training ground to validate our AI concepts.
The Entrepreneurial Gauntlet of Elixir Studios
Host: The room is packed with entrepreneurs today, and you must resonate deeply with this, having founded not just one company but being a two-time startup veteran. Let's rewind to your first venture, Elixir Studios. What was that experience like? Even though it's not the company you're best known for, you achieved tremendous success there too. How did you lead that company? What did that experience teach you about "how to build a company"?
Demis Hassabis: So, I founded Elixir Studios right after graduating from university. I was fortunate to have previously worked at Bullfrog Productions. Those who know gaming are aware that it was a legendary studio in the early days of the industry, probably the top-tier game studio in the UK, if not Europe, at the time.
I wanted to do things that pushed the boundaries of AI. Back then, developing games was my roundabout way to fund AI R&D, constantly challenging the technological frontiers and merging that with ultimate creativity. I believe this philosophy is still applicable to the blue-sky research we conduct today.
Perhaps the most profound lesson I learned is this: You need to be 5 years ahead of your time, not 50. At Elixir, we were trying to develop a game called "Republic," which aimed to simulate an entire nation. The premise was that players could overthrow a ruling dictator in various ways, and we simulated vivid, breathing cities inside the game.
Mind you, this was the late 90s, with computers running on Pentium processors. We had to run all the graphics rendering and AI logic for a million inhabitants on the home computers of that era. It was just too ambitious—even a bit grandiose—and that sparked a cascade of problems.
I took that lesson very much to heart: You need to be ahead of the curve, but if you're 50 years ahead, you will almost certainly fail disastrously. Of course, when an idea becomes obvious to everyone, it's too late to jump in. So, the key is finding that delicate balance.
Founding DeepMind in 2009
Host: Okay, speaking of not being too far ahead of your time, we jump to 2009. You were convinced that Artificial General Intelligence (AGI) would be achieved. That time, perhaps you were only 10 years ahead, which is better than 50. Talk to the entrepreneurs in the room about 2009. How did you persuade that initial cohort of top-tier geniuses? Because you genuinely recruited incredibly high-caliber staff and early team members. Back then, AGI sounded like complete science fiction. How did you get them to buy into it all?
Demis Hassabis: At the time, we had picked up on some intriguing clues. We thought we were only 5 years ahead, but it was probably more like 10. Deep Learning had just been invented by Jeff Hinton and his academic peers, but almost no one had grasped its significance. We had deep expertise in Reinforcement Learning, and we felt that combining these two techniques would achieve breakthroughs. Prior to this, they had almost never been used together—if at all, only on academic "toy problems." In the field of AI, they were two entirely separate islands.
Furthermore, we foresaw the potential of compute; GPUs at the time were poised to shine. Of course, we use TPUs now, but back then, the accelerated computing industry was going to be a massive tailwind. Simultaneously, towards the end of my PhD and postdoc, as some of the people I assembled were computational neuroscientists, we had extracted enough valuable ideas and principles from brain mechanisms. One core belief among them was that reinforcement learning could ultimately lead to AGI through scale.
We felt we had gathered the core ingredients. We even felt like guardians of some monumental secret, because absolutely no one in either academia or industry believed AI could make any significant breakthroughs. In fact, when we proposed focusing on developing AGI—or what was sometimes called "strong AI"—many academic figures would openly roll their eyes at us. To them, it was clearly a dead end; people had tried and hit a wall back in the 90s.
I was a postdoc at MIT, a bastion of Expert Systems and first-order logic language systems. Looking back, it seems unbelievable, but even then, I found that approach too rigid and outdated. Yet, whether in Cambridge, UK, or here at MIT, the traditional strongholds of AI research, people were still using those old methods. This, paradoxically, only strengthened my conviction that we were on the right track. At the very least, if we were destined to fail, we would fail in a completely novel way, not repeating the failures of those pursuing AGI in the 90s. That made it feel worth attempting no matter what; even if it was just an uncertain research project, even if it ultimately failed, at least we would have failed with considerable originality.
DeepMind's Mission and the Bet on AGI
Host: Did your early beliefs face any widespread resistance? To get those early followers on board, did you have to prove something to yourself or to them?
Demis Hassabis: Regardless of the circumstances, I would dedicate my life to artificial intelligence. As it turned out, the progress has far exceeded our rosiest expectations. However, it is still within the bounds of our 2010 forecasts—back then, we envisioned this as a 20-year journey.
I believe that as players in the field, our progress is perfectly on track, and we have clearly played our part in it.
Stepping back, even if things hadn't evolved this way and AI remained a niche, academic discipline, I would still have stuck to this path because it is, in my mind, the most important technology ever conceived. My goal has always been very clear; DeepMind's original mission statement was: Step 1: Solve intelligence, i.e., build Artificial General Intelligence (AGI). Step 2: Use it to solve everything else. I've always held that this is the most important, and also the most fascinating, technology humanity could ever invent.
It is both a tool for scientific exploration and a fascinating artifact in its own right, and one of the best avenues for understanding our own minds—such as the nature of consciousness, dreams, and creativity. As a neuroscientist, I often felt the lack of an analytical tool like AI when I was pondering these questions before. It provides a comparative mechanism, allowing us to study and contrast two different systems in depth, much like conducting a controlled experiment.
The Culture of "AI for Science"
Host: Contrasting different systems. Let's talk about "AI for Science." You entered this field early; you are a staunch believer and a pure idealist. This is the core mission that drives you. How did the model and culture you established when founding DeepMind keep it at the absolute forefront of "AI for Science"?
Demis Hassabis: That is indeed our ultimate goal. For me personally, the most fundamental driver is building AI to advance science, medicine, and our understanding of the world. This is how I live my mission—through a "meta-way": first build the ultimate tool, and once it matures, use it to achieve scientific breakthroughs. We've already had achievements like AlphaFold, and I believe many more will emerge.
DeepMind has always prioritized this goal. In fact, we have an "AI for Science" division led by Pushmeet Kohli, established almost a decade ago. We officially kicked off this work almost immediately after returning from the AlphaGo match in Seoul, which was exactly ten years ago, give or take.
Before that, I had been lying in wait, biding my time until the algorithms became powerful enough and the ideas sufficiently general. For me, conquering Go was that historic tipping point; in that moment we realized the time had come to apply these ideas to real-world, important problems, starting with these grand scientific challenges.
We have always believed that this is the most beneficial application of AI. What could be more wonderful than using it to cure diseases, extend human healthspan, and assist healthcare? Hot on its heels must come critical areas like materials science, environment, and energy. I am confident AI will shine brightly in these areas over the next few years.
Breakthroughs in Biology and Isomorphic Labs
Host: How is AI achieving breakthroughs in the field of biology? You've been deeply involved in the work at Isomorphic Labs, an area you're passionate about. From the beginning, you've been unwavering in your belief in AI's potential to cure disease. In biology, when can we expect a "Kodak moment" similar to what we've seen in language and programming?
Demis Hassabis: I believe the advent of AlphaFold already gave us biology's "Kodak moment." Protein folding and their three-dimensional structures represented a 50-year grand challenge in science. Solving this is critical if you want to design drugs or decipher the fundamental code of biology. Of course, it's just one part of the drug discovery process, extremely crucial, but just one part.
Our newly spun-out company, Isomorphic Labs (which I'm immensely enjoying running), is dedicated to building the core technologies in biochemistry and chemistry. These technologies can autonomously design chemical compounds that perfectly fit into specific pockets of a protein. Now that we know the protein's shape and surface features, we have defined the target. Next, we need to manufacture a corresponding compound that binds potently to that target, ideally while avoiding any off-target effects that could cause toxic side effects.
Our ultimate dream is this: shift the exploration process, which currently consumes 99% of the workload and time in R&D, entirely into computer simulation (in silico), leaving only the final validation for actual physical "wet lab" experiments. If we can do this—and I firmly believe it will be achievable within the next few years—we can collapse the average 10-year drug discovery cycle down to months, weeks, and perhaps eventually just days.
I believe that once we cross this inflection point, tackling all diseases will become accessible. Concepts like personalized medicine—for instance, drug variants tailored to an individual patient—will become reality. I see the entire landscape of medicine and drug R&D being completely reshaped in the coming years.
New Sciences Born from Simulators
Host: That's brilliant. You've mentioned "AI for Science" multiple times. Do you think at some juncture, AI will give birth to entirely new systems of science? Much like the Industrial Revolution spawned thermodynamics. Will fundamentally new disciplines appear in our education systems? If so, what might they look like?
Demis Hassabis: On this point, I think a few things will happen.
First, understanding and dissecting AI systems themselves will evolve into a full-fledged discipline—an Engineering Science. The artifacts we are building are incredibly fascinating and extremely complex. Eventually, their complexity will rival that of the human mind and brain. Therefore, we must study them intensively to thoroughly understand how these systems work, which is far beyond our current level of comprehension. I believe a whole new field will inevitably rise; mechanistic interpretability is just the tip of the iceberg, and there's vast territory to explore in analyzing these systems.
Secondly, I also believe AI itself will unlock new scientific frontiers. One area I'm most excited about is "AI for Simulations." I am obsessed with simulations; all the games I've ever written not only contained AI but were essentially simulators. I believe simulators are our ultimate path to cracking problems in social sciences like economics and other humanities.
The tricky part about these disciplines is that, like biology, they are emergent systems, making reproducible, controlled experiments extremely difficult. Suppose you want to raise interest rates by 0.5%; you just have to do it in the real world and observe the consequences. You can have all sorts of theories, but you can't run this experiment hundreds of thousands of times. However, if we can accurately simulate these complex systems, then conducting rigorous sampling-based deductions using highly accurate simulators might just establish a brand new science. I believe this will give us the ability to make better decisions in areas currently fraught with high uncertainty.
Host: What conditions are necessary to achieve these incredibly precise simulations? Things like World Models—what kind of scientific and engineering breakthroughs are required to get there?
Demis Hassabis: I've been thinking deeply about this. In our work, we heavily use learning simulators. These are simulators applied in areas where we either lack sufficient mathematical understanding, or the systems are just too complex. We can't simply write a direct simulation program tailored to a specific situation because that wouldn't be accurate enough and couldn't encompass all the variables.
We've already put this into practice with weather forecasting. We have the world's most accurate weather simulator, "WeatherNext," which runs far faster than the tools currently used by meteorologists. I'm not sure if we can know everything, nor if that's a good idea, but the first step is better understanding these complex systems.
Even within biology, we are working on something called a "Virtual Cell"—an incredibly dynamic emergent system. Just as mathematics is the perfect descriptive language for physics, machine learning will be the perfect descriptive language for biology. In biology and many natural systems, there's an abundance of weak signals, weak correlations, and vast amounts of data, which far exceed the analytical capacity of the human brain. Yet, within these massive datasets, there do exist inherent connections, correlations, and thought-provoking causal relationships.
Machine learning is the perfect tool for describing such systems. To this day, mathematics has been unable to do this, either because the systems are too complex for even top mathematicians to handle, or because the expressive power of mathematics is insufficient to grasp these highly emergent, dynamic systems—partly due to their messy, stochastic nature.
Eventually, once you have mastered these simulators, a new branch of science might potentially derive from it. You could try to extract explicit equations from these implicit or intuitive simulators. Since you can sample the simulator countless times at will, perhaps one day, you could discover fundamental scientific laws on par with Maxwell's equations.
Maybe. I don't know if such laws exist for emergent systems, but if they do, I see no reason why we couldn't discover them via this approach.
Host: That would be phenomenal. You've also spoken about a theory, on a more theoretical level, that the fundamental building block of everything in the universe might be akin to information. How do you see this? And what does it mean for the traditional, classical Turing machine?
Demis Hassabis: Of course, you can cite the famous E=mc² and all of Einstein's work, showing that energy and matter are essentially equivalent. But I actually believe that information also possesses a kind of equivalence. You can view the organization of matter and structures—especially systems like living organisms that resist entropy—as fundamentally being information-processing systems. Hence, I think one can interconvert these three.
However, I have a feeling that information is the most fundamental layer. This is the exact opposite view of classical physicists in the 1920s, when they regarded energy and matter as primary. I actually think it's a better way to understand the world by viewing the universe as being built of information first and foremost.
If this holds true—and I think there's mounting evidence supporting it—then artificial intelligence has an even more profound significance than we imagined. It is already deeply significant because its core is organizing information, understanding information, and constructing informational objects.
In my view, AI is fundamentally about information processing. If you take information processing as the primary lens for understanding the world, you'll find extremely deep interconnections between these disparate fields.
Host: So, do you think the classical Turing machine can compute everything?
Demis Hassabis: Sometimes I reflect on our work and see myself as a "defender of Turing," because Alan Turing is one of the scientific heroes I admire most in my life. I believe his work laid the foundation not only for computers and computer science but also for Artificial Intelligence. The theory of the Turing machine is one of the most profound results in history: anything that is computable can be computed by a machine that is relatively simple to describe. Therefore, I think our brains are very likely approximate Turing machines.
It's fascinating to think about the connection between Turing machines and quantum systems. However, what we've demonstrated through systems like AlphaGo and, especially, AlphaFold, is that classical Turing machines, dressed in the garb of modern neural networks, can model problems previously thought to require quantum mechanics to solve. For example, protein folding is, in a sense, a quantum system involving minuscule particles. One might assume you'd have to account for all the quantum effects of hydrogen bonds and other complex interactions.
As it turns out, a near-optimal solution can be achieved using a classical system. Thus, we may discover that many things we thought we must rely on quantum systems to simulate or run, can actually be modeled on classical systems if approached correctly.
Consciousness Philosophy
Host: You've always viewed AI as a tool, much like the telescope, microscope, or astrolabe of past centuries. But when you're faced with a machine that can simulate almost everything—as you mentioned, even quantum systems—at what point does it transcend the category of being just a tool? Will that day ever truly arrive?
Demis Hassabis: I feel very strongly that on the mission and march toward building AGI, those of us on this journey—including many in this room—believe the best approach is to first build a tool: an extraordinarily intelligent, useful, and precise tool, and then cross the next threshold. That in itself is profound enough. Of course, that tool might become increasingly autonomous, exhibiting more agent-like characteristics, which is what we're currently witnessing. We are riding a wave right now in this Agent Era.
However, there are further questions: Does it possess agency? Is it conscious? These are questions we will inevitably have to confront. But I recommend we treat this as Step 2, perhaps using the tool built in Step 1 to help us explore these profound questions.
Ideally, through this process, we can also better understand our own brains and minds, and be able to define concepts like "consciousness" with more precision than we can today.
Host: Do you have any rough predictions for the future definition of consciousness?
Demis Hassabis: No, I don't have much to add beyond what's been discussed in philosophy for millennia. But it seems clear to me that certain components are obviously necessary. They are likely necessary but not sufficient conditions. Things like self-awareness, a concept of self versus other, and some form of temporal continuity are obviously required for any entity that appears to be conscious.
However, what the complete definition is remains an open question. I've discussed this with many great philosophers. A few years ago, I had deep conversations on this topic with Daniel Dennett, who sadly passed away not long ago. One of the core issues is the system's behavior: does it behave as if it's a conscious system? You could argue that as some AI systems get closer to AGI, they might eventually do just that.
But then the question follows: why do we assume each other is conscious? One reason is our mode of behavior; we act like conscious beings. But another factor is that we all run on the same underlying substrate.
So I think, if both these points are true, it is the most logically parsimonious assumption that your experience is the same as mine, which is why we don't normally argue about whether the other is conscious. Obviously, however, we can never achieve the same substrate equivalence with an artificial system. So I think it will be very difficult to bridge that gap entirely. You could look at it behaviorally, but what about experientially? After achieving AGI, there might be ways to address this, but that might be beyond the scope of today's discussion, even within a talk on "AI and Science."
Host: Fantastic. We'll open up to audience questions very soon, so please have your questions ready. You just mentioned philosophers, specifically Kant and Spinoza, as two of your favorites. Kant is a quintessentially deontological philosopher, heavily emphasizing the concept of duty; Spinoza, on the other hand, held an almost deterministic view of the cosmos. How do you reconcile these two vastly different philosophies? And what is your own fundamental understanding of how the world operates?
Demis Hassabis: The reason I'm drawn to and impressed by these two philosophers is that Kant proposed an idea—which I felt deeply during my neuroscience PhD—that "the mind creates reality." I think this is largely correct. It provides another fantastic reason for studying how the mind and brain work. Since what I ultimately seek is the nature of reality, we must first understand how the mind interprets that reality. That's the insight I took from Kant.
As for Spinoza, it's more about the spiritual dimension. If you try to use science as a tool to understand the universe, you begin to brush up against the deep mysteries behind how the cosmos operates.
That's exactly how I feel about our current endeavor. When I engage in scientific research, delve into AI, and build these tools, I feel as though we are, in some way, reading the language of the universe.
Host: Beautiful. That's the most beautiful description of your daily work: Demis, you are a scientist, an orator, and a philosopher rolled into one. Before we wrap up, let's do a few rapid-fire questions. He hasn't seen any of these beforehand. Prediction for the year we achieve AGI—earlier or later than expected? Or you can choose not to answer.
Demis Hassabis: I'll go with 2030. I've been pretty steadfast in that prediction.
Host: Okay, 2030. So, when we achieve AGI, what book, poem, or paper would you recommend reading?
Demis Hassabis: For the world post-AGI, my favorite book is David Deutsch's "The Fabric of Reality." I believe its ideas are still highly applicable. I hope to use AGI to answer the profound questions posed in that book, which will form the main body of my follow-up work in the AGI era.
Host: Fantastic. What has been your proudest moment to date at DeepMind?
Demis Hassabis: We've been lucky to experience many peak moments. I suppose the proudest would be the creation of AlphaFold.
Host: Okay, a couple of final game-related questions. If you were in a high-stakes, turn-based strategy game—like Civilization or Polytopia, these hardcore games—and you could pick one scientist from history as your teammate, say Einstein, Turing, or Newton, who would you choose to join your squad?
Demis Hassabis: I think I'd pick von Neumann. You need a Game Theory expert in these situations, and I believe he is the ultimate.
Host: That is a god-tier teammate right there. Demis, you are truly a Renaissance man. Thank you so much for being on the show today. Everyone, please join me in a round of applause for Demis's brilliant sharing. Thank you very much.
Reference: https://www.youtube.com/watch?v=AFpeWo1GTeg, publicly published on April 29, 2026