Scaling laws and model progress

DeepMind’s co‑founder and CEO Demis Hassabis lays out a pragmatic path to AGI—why he still sees a 5–10 year window, what current models are missing (reasoning, planning, continual learning), and why world models will likely converge with today’s LLMs. He explains how DeepMind became Google’s “engine room” for AI, shipping Gemini advances across products, and why AI for science (from AlphaFold to materials and energy research) is the most exciting frontier. We also touch on compute, energy constraints, and how to scale without cutting corners on safety.

See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.