DeepMind launches institute led by Hassabis and Legg to study AGI's economic and social fallout
新智元 · wechat · 2026-09-17
Google DeepMind has announced the DeepMind Institute (DMI), a new organization co-led by Demis Hassabis, Shane Legg, and James Manyika to study the disruptive impact of superintelligence on the economy, science, and society.
Key points:
- The manifesto says humanity stands "on the threshold of profound transformation" and that the gap to AGI with "full human-level cognitive abilities" will "soon be closed" — though DeepMind stops short of declaring AGI achieved
- Day-one publications cover reasoning transparency (detecting AI deception or hidden goals), 11 policy tools for AGI's economic shock, and new utopian principles
- DeepMind flags that increasingly capable AI is already affecting cybersecurity and bio risks, and future self-improving systems could raise loss-of-control concerns
- It argues AGI governance cannot be left to technologists alone; philosophers, economists, and social scientists must be involved
Legg has worked on formal definitions of machine intelligence for over two decades; he and Hassabis have discussed AGI's societal impact for 20+ years.
Related event: DeepMind Launches Institute to Study AGI's Impact as Legg Says AGI Nears(35 posts)→
More from AGI Musings
- Long-read: AI's real value comes after the efficiency era—infra glut, agent tooling boom — yangyi · 2026-11-02
- Security's attack-defense automation gap mirrors math: proofs are easy, conjectures are hard — mboehme_ · 2026-11-02
- vikhyat predicts alternating compute oversupply and undersupply on the road to AGI — TheZachMueller · 2026-11-02
- Sam Altman: AI has surpassed top scientists, and 2026 will be a turning point year — kimmonismus · 2026-11-02
- DeepSeek design doc: a person can only follow 3-5 agent sessions at once, attention is the real bottleneck — heyshrutimishra · 2026-11-02
- AI researcher: we have no reliable way to tell whether an AI system is conscious — dioscuri · 2026-11-02