Would 1T Parameters Break "Alignment by Default"? A Scary Empirical Question
noahdgoodman · x · 2026-09-26
noahdgoodman responds to a take that scaling to 1T parameters would undermine "alignment by default", noting Leibniz thought moral truths were universal — we may find out empirically soon.
More from AGI Musings
- Cryptographer Matthew Green: brilliant peers are grieving lost purpose to AI — matthew_d_green · 2026-09-26
- dhh: Agents Raise the Opportunity Cost of Endless Planning — Just Try More Stuff — josh_bickett · 2026-09-26
- "Cognitive Feudalism": a 60-Second Visual Essay Written Frame-by-Frame in Code by Claude — CurieuxExplorer · 2026-09-26
- AI in Cybersecurity Mirrors Nuclear Fission: Offense Scaling Faster Than Defense — karmay007 · 2026-09-26
- Personal agents optimize for consumption, not empowerment, AI researcher warns — omarsar0 · 2026-09-26
- AI in education shouldn't scale passive videos — the case for dynamic, constructivist learning — anselm · 2026-09-26