Google DeepMind Pushes ‘World Model’ Theory with New Research Agent
Demis Hassabis outlines the path to AGI as Google releases an autonomous research tool.
Beyond Next-Token Prediction
On Dec. 16, Google DeepMind CEO Demis Hassabis articulated a roadmap for Artificial General Intelligence (AGI) centered on "World Models." In a widely circulated podcast, Hassabis argued that true AGI requires an internal physics engine capable of understanding causality, space, and time, rather than merely predicting text tokens. He revealed that DeepMind’s video generation research suggests AI systems are learning intuitive physics through passive observation, challenging the scientific view that physical interaction is required for perception.
Deep Research Agent Deployed
Translating this research into product, Google released the Deep Research Agent via the Gemini API …
Archive Access
This article is older than 24 hours. Create a free account to access our 7-day archive.