AGI believers need not assume a permanently offense-dominant world, says a reposted thread
sebkrier · x · 2026-07-26
A reposted thread pushes back on a common belief in AGI circles: that if you believe AGI/ASI is possible, you must also believe the long-run equilibrium is heavily offense-dominant.
The quoted post from @emollick adds that many people discussing open-weights models do not deeply buy into the lab-insider vision of AGI/ASI, including fears around grave semi-autonomous biosecurity risks. The broader point is that the relationship between believing in AGI and believing in extreme offensive-risk outcomes is much weaker than many assume.
Related event: Open vs Closed AI Debate Hinges on Belief in AGI Risks(7 posts)→
More from AGI Musings
- Open-model advocates are challenged on why closed models should still exist — beenwrekt · 2026-07-26
- Gavin Baker says AI will stay multi-model, and open source is vital to U.S. competitiveness — GavinSBaker · 2026-07-26
- Ben Horowitz says open-source AI is essential as frontier labs push for control — a16z Podcast · 2026-07-26
- A viral Reddit line frames AI as shifting skill access toward capital — churchkey · 2026-07-26
- Steven Strogatz reposts a case for using AI in math without misleading the field — stevenstrogatz · 2026-07-26
- A joke predicts the first AI Oscar winner will make everyone retroactively supportive — taherdhanera · 2026-07-26