20VC: DeepMind's Demis Hassabis on Why AGI is Bigger than the Industrial Revolution | Why LLMs Will Not Commoditise & We Have Not Hit Scaling Laws | Bottlenecks in AI & The Energy Crisis Caused By AI | Whether AI Will Do More to Harm or Help Inequality

What It Covers

DeepMind CEO Demis Hassabis outlines his AGI timeline of within five years, explains why scaling laws have not plateaued, identifies continual learning and hierarchical planning as critical missing capabilities, and addresses AI's potential impact on drug discovery, energy, labor displacement, and global inequality.

Key Questions Answered

• •AGI Timeline: Hassabis places a high probability on AGI arriving within five years, framing it as 10 times the magnitude of the Industrial Revolution unfolding at 10 times the speed — compressed into roughly one decade rather than a century. Investors and founders should plan product and hiring strategies around this compressed timeline rather than treating AGI as a distant abstraction.
• •Scaling Laws Still Productive: Returns from scaling large language models remain substantial, though growth rates have slowed from early exponential jumps. The practical implication: labs with the capacity to generate new algorithmic breakthroughs — not just scale existing architectures — will pull ahead over the next two to three years as current ideas approach diminishing returns.
• •Critical Missing Capabilities: Two gaps limit current AI systems — continual learning (models cannot incorporate new knowledge post-training) and long-horizon hierarchical planning. Hassabis draws a parallel to the brain's sleep-based memory consolidation as a potential architectural model. Builders evaluating AI reliability for agentic workflows should treat these gaps as hard constraints, not minor limitations.
• •Drug Discovery Roadmap: Isomorphic Labs targets a complete AI-driven drug design platform within five to ten years, covering compound design, toxicity screening, and genomic patient stratification. The bottleneck then shifts to regulatory trial timelines. Hassabis argues that once a dozen AI-designed drugs complete full trials, regulators will have sufficient backtest data to compress or eliminate certain trial phases.
• •Energy and Inequality Strategy: AI's energy demands can be offset by AI-optimized national grids, which Hassabis estimates could yield roughly 40% efficiency gains, plus breakthroughs in fusion and superconductors. On inequality, he proposes sovereign wealth funds and pension funds taking equity stakes in leading AI companies as a structural mechanism to distribute productivity gains broadly rather than concentrating them among a small number of shareholders.