As global competition in artificial intelligence (AI) intensifies, the words and actions of Demis Hassabis, CEO of Google DeepMind, are drawing attention in the IT industry. At the World Economic Forum last month, he said, "There is an increasing bubble in AI investment," adding, "It's not sustainable for brand-new startups with no products or technology to pull in billions of dollars in seed funding." At the same time, he expressed confidence, saying, "Even if the bubble bursts, Google will be fine."
On the 11th (local time), Fortune said his decision to sell DeepMind, which he co-founded in Jan. 2014, to Google is seen as an event that had a significant impact on the subsequent structure of the AI industry.
At the time, Hassabis rejected Meta's offer, which proposed a higher acquisition price, and chose Google. The explanation is that he judged Google's vast computing resources and research infrastructure to be advantageous for long-term AI development. This transaction became one of the triggers for Tesla CEO Elon Musk to worry about the possibility of AI monopolies, and there is also analysis that later led him, along with Sam Altman and others, to establish OpenAI as a nonprofit. The sale of DeepMind is cited as one of the starting points that ultimately formed the current competitive landscape with Google and OpenAI as the two pillars.
Hassabis founded DeepMind in 2010, setting the development of artificial general intelligence (AGI) as the goal. He was a chess master in his youth and had a deep interest in astronomy and cosmology. It is known that his motivation for founding was to study AI as a way to understand the structure of human thought. At the time, AI did not receive the kind of industrial attention it does now, but he continued his research with the long-term impact on science and industry in mind.
DeepMind drew global attention in 2016 when AlphaGo defeated 9-dan Lee Sedol. Evaluations said a Reinforcement Learning-based system proved it can surpass the highest human level in complex strategy games. It later expanded its research into biology by unveiling the protein 3D structure prediction AI "AlphaFold." AlphaFold, a technology that predicts protein structures from amino acid sequences alone, is seen as a foundation that can improve the efficiency of new drug development and disease research. This achievement led to a Nobel Prize.
Hassabis now oversees Google's AI research and product strategy. He unified research capabilities by integrating DeepMind and the Google Brain organization, and is working to apply AI to core services such as Search, YouTube and Chrome around the Gemini model. He has emphasized a structure that quickly reflects research results in products. His position is that securing a top-tier model is the starting point for competitiveness.
Hassabis also mentions a balance between research and commercialization. He founded the new drug development company Isomorphic based on AlphaFold to apply AI to the pharmaceutical field. Traditional new drug development takes an average of more than 10 years and has a low success rate. He laid out a strategy to shorten the discovery process for candidate substances through simulation-based search and design. Programs are underway in cancer, cardiovascular and immune diseases, and some aim to enter the clinical stage.
Attention is also drawn to his organizational management style. He adopted a collaborative structure that groups biologists, chemists and physicists with Machine Learning researchers and engineers into one team. The judgment is that lowering the boundaries between fields improves research efficiency. His personal schedule management is also unique. He said he concentrates meetings during the day and maintains a work pattern of allocating nighttime hours to research and strategy planning.
He offers a relatively optimistic outlook on the future direction of AI. While he expects continued progress in autonomous agents, smart glasses and robotics, he also emphasizes the need for safety and responsible development. His position is that technology must succeed commercially to sustain research, but at the same time must secure reliability and stability.
Hassabis believes AI can bring structural changes to medicine, energy and new materials within the next 10 to 15 years. However, he said the premise is to advance the technology within a controllable range. The choices and strategies he has pursued since the sale of DeepMind are influencing not only the direction of Google AI but also the global AI competitive landscape.