DeepMind Researcher: AI's Future Is Millions of Small Models, Not One Giant

DeepMind researcher Andrew Trask argues AI's endgame is an internet of millions of small integrated models rather than a few giant frontier models, and proposes building AI-proof sandboxes via adversarial training that keeps attacking until no vulnerabilities remain.

2026-10-08 ~ 2026-10-08 · 3 related posts

• DeepMind's Andrew Trask: adversarially train models to break sandboxes until no holes remain — iamtrask · 2026-10-08
• DeepMind's Andrew Trask: AI heads for a mainframe-to-PC moment of millions of small models — iamtrask · 2026-10-08
• DeepMind's Andrew Trask: AI's endgame is millions of ensembled models, not one giant — iamtrask · 2026-10-08