Sandberg: Governments Are Far Worse at Frontier Models Than Companies
anderssandberg · x · 2026-10-28
Oxford researcher Anders Sandberg argues that governments have so far been far worse than companies at building frontier models, and that recognizing the best models are uncontrollable tends to slow things down in general — though stupidity and orders from on high can keep building more risk.
The thread raises the question of whether such incidents ultimately slow AI progress by exposing dangers, or accelerate it as governments see offensive potential.
More from AGI Musings
- Johns Hopkins lands up to $52M from ARPA-H to build AI that decodes protein misfolding — JosephJacks_ · 2026-10-28
- Google AI's Pilar Manchón on AI for cancer detection, crystal discovery and dolphin communication — pilarmanchon · 2026-10-28
- Echoes of 1999: fiber glut left 97% unused, but hoarded $30k AI chips won't flood the market — aakashgupta · 2026-10-28
- Gartner: 47% of supply chain firms cut 10.5% of staff due to AI, entry-level hit hardest — krishnan · 2026-10-28
- OpenAI engineer: frontier token demand keeps growing even if model intelligence plateaus — iaindunning · 2026-10-28
- CS professor: if you never learned to write well yourself, how can you judge LLM-assisted papers? — CSProfKGD · 2026-10-28