Google DeepMind chief warns AI investment looks ‘bubble-like’ | FT Interview

Key Concepts

• Gemini 3: Google’s latest and most powerful AI model, currently topping industry leaderboards.
• Multimodal AI: AI systems capable of processing and understanding multiple types of data (text, image, audio, video) natively.
• AGI (Artificial General Intelligence): Hypothetical AI with human-level cognitive abilities, capable of understanding, learning, and applying knowledge across a wide range of tasks.
• Isomorphic Labs: Google DeepMind’s division focused on AI-driven drug discovery.
• Continual/Online Learning: AI’s ability to learn and adapt continuously from new data and experiences after initial training.
• AI for Science: Utilizing AI to accelerate scientific discovery, exemplified by AlphaFold.
• Talent War: Intense competition among AI companies to attract and retain skilled researchers.

Google’s Position in the AI Race & Competitive Landscape

Google is currently experiencing a period of strong progress in AI, with Gemini 3 achieving top performance across various benchmarks. The company believes its trajectory is the fastest in the industry, building on the competitiveness established with Gemini 2.5. Despite this success, Google acknowledges the “ferocious” and “intense” competition, emphasizing the need for continued innovation and rapid product delivery. Currently, the Gemini app boasts 650 million monthly users, and AI Overview features are used by approximately 2 billion users, making it the most widely used AI product globally. However, Google recognizes they are still “scratching the surface” of their potential.

Regarding competitors, Google acknowledges Anthropic’s advancements in code generation with “claw code” as particularly noteworthy within the developer market. However, Google emphasizes its strength in multimodal AI as a key differentiator. This capability, inherent in Gemini from its inception, allows the system to natively process and understand various data types – image, video, audio, and text – crucial for real-world applications like assistive technology and robotics.

Gemini & Future Product Development

Google DeepMind is positioned as the “engine room” of Google, developing foundational models like Gemini, VO, and Nana Banana. The focus is on integrating these models into existing products (email, Chrome, Search) and exploring new “greenfield” areas, particularly digital assistants like the Gemini app. Google is actively developing new devices, including smart glasses in partnership with WBBY Parker and Gentle Monster. They acknowledge past attempts at smart glasses were premature, lacking a “killer app,” and believe a universal digital assistant could fulfill that role.

Talent Acquisition & Retention

The AI industry is characterized by a significant “talent war,” with researchers receiving offers exceeding $100 million. Google addresses this by offering competitive compensation but emphasizes the importance of mission and purpose. They aim to attract researchers motivated by the potential for positive impact through AI, particularly in areas like scientific discovery (e.g., AlphaFold) and enhancing widely used products like Google Maps and Gmail. The ability to quickly deploy research breakthroughs to a billion users is a key motivator.

Safety, Misuse & Industry Collaboration

Google recognizes societal concerns regarding AI safety and potential misuse, acknowledging the need for responsible development and deployment. They are proactively implementing safeguards like synth ID (watermarking for deepfakes) and usage restrictions for Gemini. They view themselves as role models for responsible AI practices and are doubling down on “AI for science” and “AI for medicine” to demonstrate the unequivocal benefits of the technology.

While some industry groups exist, Google notes a lack of unified action. Different “frontier labs” are pursuing diverse approaches. Google believes a governmental framework may be necessary to establish industry-wide standards.

The AI Bubble & Competition with China

Google acknowledges the possibility of an AI bubble, particularly concerning inflated valuations of early-stage startups. However, they remain optimistic about the long-term transformative potential of AI, citing strong usage and demand for their models. They believe their core business is resilient, even if a bubble were to burst, and their AI-first products like the Gemini app position them for continued growth.

Regarding competition with China, Google suggests a different approach. While the Western focus is often on achieving AGI, China appears more focused on near-term applications and efficiency gains. Google believes Western companies currently maintain a lead in AGI research, though the gap is narrowing, potentially to around 6 months. They acknowledge the rapid progress of Chinese labs in catching up but note they haven’t yet demonstrated innovation beyond the current frontier (e.g., creating a new architecture like Transformers).

AGI & the Future of AI Research

Google DeepMind remains committed to achieving AGI as its “north star.” They believe AGI will unlock unprecedented opportunities. Key breakthroughs needed include continual learning, personalization, and recursive self-improvement – enabling AI models to learn and improve autonomously. They draw parallels to their success with AlphaGo and AlphaZero, but acknowledge the challenges of applying those techniques to the complexities of the real world.

The current timeline for AGI remains consistent at 5-10 years (now refined to 4-9 years), with a roughly 50% probability of achieving it by 2030.

Isomorphic Labs & AI-Driven Drug Discovery

Isomorphic Labs, Google DeepMind’s drug discovery division, is making significant progress in preclinical trials. They have established partnerships with major pharmaceutical companies (J&J, Eli Lilly, Novartis) and have approximately 17 programs underway. Google anticipates sharing more detailed updates on their progress in the first half of the year. They are also investing in a material science lab in the UK to accelerate the discovery of new materials using AI.

Leadership & Future Direction

The speaker expresses contentment in their current role, balancing scientific research with product leadership. While not explicitly ruling out a future CEO position, they emphasize their passion for staying close to the core science and innovation. They are driven by the pursuit of cutting-edge technology and the challenge of solving complex problems.

Quote: “Most top researchers, you know, every they're of course fabulously well paid, but it's then beyond that is like the mission. What are you um trying to do with your skills?” – Speaker, regarding attracting and retaining AI talent.

Quote: “I think we have by far the deepest and uh broadest research bench. I think we have the most talent in the industry.” – Speaker, on Google’s competitive advantage.

Quote: “I think we’re about you know 5 to 10 years away [from AGI]. So maybe it's now like 4 to 9 years.” – Speaker, on the timeline for achieving AGI.