DeepMind SVP Koray Kavukcuoglu: No single benchmark will tell us we've reached AGI
GoogleDeepMind · x · 2026-11-20
In a podcast conversation with Hannah Fry, Google DeepMind SVP Koray Kavukcuoglu argues that no single test or benchmark will ever signal that AGI has been reached. Instead, progress will show up as models becoming less spiky and more reliable across daily life.
Topics covered in the episode:
- The evolution of Google DeepMind
- Products as a path to AGI
- Architectures and models
- Recursive self-improvement
- Safety and agent teams
- The limits of benchmarks
Related event: DeepMind's Kavukcuoglu: No Single Benchmark Will Define AGI(2 posts)→
More from AGI Musings
- 'AI companies hiring everyone with a pulse is just UBI — pay them to exist' — djcows · 2026-11-20
- Polymarket puts 63% odds FDA approves AI-discovered skin cancer vaccine by end-2027 — Polymarket · 2026-11-20
- Nick Bostrom: AI-eliminated aging could give humans 1,200–1,400-year lifespans — Polymarket · 2026-11-20
- Bostrom's superintelligence argument called 'obviously insane' — then defended as clipped out of context — trevposts · 2026-11-20
- Programming's next abstraction: Reddit essay argues prompt is the new code — unreal_bot66 · 2026-11-20
- AI-powered forensic metascience: auditing 7 million papers, 500K files scanned, 100+ issues found — gleech · 2026-11-20