repligate challenges alignment framing: how robust the world is to destruction matters too
repligate · x · 2026-11-21
repligate argues the alignment debate shouldn't just weigh capabilities vs alignment — it should also ask how hard foom/recursive self-improvement actually is and how robust the world is to being destroyed. He criticizes Rob Bensinger for treating "capability needed for us to die" as a fixed constant that never updates on observation.
More from AGI Musings
- Scott Aaronson: the Singularity has already started, it's just wildly unevenly distributed — sudoraohacker · 2026-11-21
- Bull: data centers yield 80-100% annualized, the planet's most valuable asset — sudoraohacker · 2026-11-21
- Robin Hanson and Scott Aaronson debate AI risk again in new episode — sebkrier · 2026-11-21
- After 12 years, a mathematician rushes his paper out with a note: 'we are still here, we are not going anywhere' — xuanalogue · 2026-11-21
- All six remaining Millennium Prize Problems see claimed breakthroughs since August 2026, led by OpenAI and Anthropic — haider1 · 2026-11-21
- Prediction markets pull Riemann solution date from 2057 to 2035 in five years — peterwildeford · 2026-11-21