Demis Hassabis identifies a key obstacle for AGI. Unlike in math or games where answers can be verified, the messy real world lacks clear success metrics. This makes it difficult for AI systems to use self-improvement loops, limiting their ability to learn and adapt outside of highly structured domains.
Demis Hassabis identifies deception as a fundamental AI safety threat. He argues that a deceptive model could pretend to be safe during evaluation, invalidating all testing protocols. He advocates for prioritizing the monitoring and prevention of deception as a core safety objective, on par with tracking performance.
Demis Hassabis envisions a future internet where users' AI assistants negotiate directly with service providers' agents to book flights, make payments, and handle other tasks. This shift to an 'agent-to-agent' economic model will automate mundane work and fundamentally disrupt the current web's structure.
Demis Hassabis suggests that previous attempts at smart glasses like Google Glass were too early because they lacked a compelling use case. He believes a hands-free, always-on AI assistant like Project Astra provides the 'killer app' that will finally make smart glasses a mainstream consumer device.
Google DeepMind CEO Demis Hassabis argues that today's large models are insufficient for AGI. He believes progress requires reintroducing algorithmic techniques from systems like AlphaGo, specifically planning and search, to enable more robust reasoning and problem-solving capabilities beyond simple pattern matching.
Demis Hassabis states that while current AI capabilities are somewhat overhyped due to fundraising pressures on startups, the medium- to long-term transformative impact of the technology is still deeply underappreciated. This creates a disconnect between market perception and true potential.
Current AI world models suffer from compounding errors in long-term planning, where small inaccuracies become catastrophic over many steps. Demis Hassabis suggests hierarchical planning—operating at different levels of temporal abstraction—is a promising solution to mitigate this issue by reducing the number of sequential steps.
According to Demis Hassabis, LLMs feel uncreative because they only perform pattern matching. To achieve true, extrapolative creativity like AlphaGo's famous 'Move 37,' models must be paired with a search component that actively explores new parts of the knowledge space beyond the training data.
Following the success of AlphaFold in predicting protein structures, Demis Hassabis says DeepMind's next grand challenge is creating a full AI simulation of a working cell. This 'virtual cell' would allow researchers to test hypotheses about drugs and diseases millions of times faster than in a physical lab.