Demis Hassabis: Google DeepMind’s CEO on AI’s Future

$Alphabet(GOOGL.US) $Alphabet - C(GOOG.US) Recently, I read some interviews with Demis Hassabis, the CEO and co-founder of Google DeepMind, and his perspective on DeepMind’s role is pretty fascinating. He basically calls DeepMind the “AI engine room” driving Google’s future. Models like Gemini, Genie, and Veo are all being deeply integrated into Google’s existing products. What really caught my eye is Genie 3—the world model that’s trying to “reverse engineer” the laws of physics. This could seriously revolutionize robotics in the next few years. Google is even building robot models based on Gemini, aiming for something like a universal “Android system for robots.”

As for the hot topic of AGI (Artificial General Intelligence), Hassabis is pretty level-headed. He straight-up says that claims from some competitors about having “PhD-level intelligence” are basically nonsense. He didn’t name names, but it’s pretty obvious he’s referring to OpenAI’s Sam Altman and his controversial statement about GPT-5 reaching PhD-level capabilities.

Hassabis thinks the biggest flaw in current AI is that it can’t generate truly creative ideas. To achieve AGI, we still need one or two major technological breakthroughs. He also mentioned his AI drug discovery company, Isomorphic, which is making rapid progress—he expects it to enter preclinical studies next year, with the goal of shrinking drug development timelines from years to just weeks.
 

Hassabis has actually been working in AI since the 1990s and is especially passionate about “AI for Science.” He believes AGI could become the ultimate tool for scientific exploration, potentially ushering in a new golden age for science, maybe even another Renaissance.