Demis Hassabis doesn’t just want to build AI. He wants to use it to understand the universe.
In a sweeping two-and-a-half-hour conversation on the Lex Fridman Podcast, the Google DeepMind CEO and co-founder shared his philosophy, predictions, and personal motivations for building advanced AI.
What emerged was not only a technical masterclass, but also a profoundly human story about one of the most important figures shaping our future.
What do you need to know about what one of AI’s top minds thinks about the near-future of AI?
I got the scoop from Marketing AI Institute founder and CEO Paul Roetzer on Episode 162 of The Artificial Intelligence Show.
One of the most striking takeaways from the conversation?
Hassabis believes we have a 50% chance of achieving artificial general intelligence (AGI) by 2030. And he sets an unusually high bar for AGI, defining it as a system that demonstrates consistent, cross-domain brilliance in reasoning, creativity, planning, and problem-solving. not just dominance in narrow tasks.
It’s a vision that reaches far beyond chatbots. Hassabis imagines AI systems inventing new scientific theories, designing elegant games from scratch, or modeling complex physics simply by watching the world, just as DeepMind’s Veo 3 video model learned intuitive physics from YouTube clips.
Yet, for all his optimism, he remains cautious. He notes that, while models are scaling well, it's unclear if scaling alone will get us to this point, acknowledging that new scientific breakthroughs may still be required for AI to become this powerful.
Unlike some of the other high-profile leaders in AI. many of whom are entrepreneurs, capitalists, or social media moguls, Hassabis is a scientist at heart. A child chess prodigy turned neuroscientist, he founded DeepMind with the goal of solving intelligence, and then using that to solve everything else.
And it shows, says Roetzer.
"Listening to him speak about AI in the future versus all the other AI lab leaders, it's somewhat jarring actually how start the contrast is."
Jarring in a good way. Hassabis' commentary is often steeped in scientific curiosity. He doesn’t speak in product pitches or valuation metrics. Instead, he talks about using AI to model biological cells, simulate the origins of life, and crack questions like P vs NP or the nature of consciousness.
"When I listen to Demis, it gives me hope for humanity," says Roetzer.
"I feel like his intentions are actually pure and science-based. It's almost like you could go back and listen to Von Neumann or Einstein or Tesla, and hear their dreams and aspirations and visions in real time as they were reinventing the future."
Throughout the interview, Hassabis circles back to a singular theme: using AI to understand the mysteries of the universe.
He speaks passionately about modeling cells (starting with yeast), simulating the emergence of life from a "primordial soup," and developing AI systems that can propose scientific conjectures worthy of Einstein. He doesn’t just want to build tools—he wants to build partners in discovery.
To get there, he’s exploring hybrid systems that combine neural networks with search algorithms and evolutionary techniques.
The goal? AI systems that don't just mimic intelligence, but extend it, uncovering new knowledge and exploring the unknown.
And, despite painting an extraordinary vision of the future, Hassabis is perhaps the least hype-driven voice in AI. He stresses caution, transparency, and the need for global cooperation to safely develop AGI. He worries about misuse by bad actors and emphasizes the importance of understanding the values and motivations behind every model being built.
That might be Hassabis’ most powerful trait: the ability to dream big while staying grounded in science and responsibility. Ultimately, it's also why he's worth listening to, says Roetzer.
"You can see why there's some optimism of a future of abundance if the world Demis envisions becomes possible."