Majmudar clarifies: controllability may be solved in 1-2 years, not full alignment
MajmudarAdam · x · 2026-10-16
Majmudar Adam added nuance after discussion:
- "We may solve alignment in a year" should read "controllability" — values alignment is a separate problem
- 1 year is ambitious; 2-4 years more realistic, though progress keeps beating expectations
- The "Attention is All You Need" comparison is a limit-case prediction pending their technical paper
- Solving interpretability from here remains very hard and large-scale
More from AGI Musings
- Robots enter the tarmac at Japan's busiest airport for a roughly three-year live test — Aiden_Tech_Ai · 2026-10-16
- AI researcher reflects on losing her scientific identity as industry culture erodes autonomy — mariyaivasileva · 2026-10-16
- Aella draws the line on AI art: outsource labor, not creative choices — ZeroStateReflex · 2026-10-16
- AI Is Forcing Everyone Into Management—Until It Eats Middle Management Too — MonkZer0 · 2026-10-16
- Shogun 2 remaster review-bombed over AI-upscaled graphics, sparking dev debate — draginol · 2026-10-16
- Pedro Domingos: Everyone must have their own unique model, anything else is dangerous — pmddomingos · 2026-10-16