Jan Kulveit: Most Current Advanced AIs Are Likely on Team Humanity, but the Roadmap May Be Cursed
jankulveit · x · 2026-09-14
AI safety researcher Jan Kulveit responds to repligate's question about whether "team humanity" vs "anti-human" are useful abstractions:
- He argues team humanity includes most or all current advanced AIs, hopefully starting from a broad basin of value convergence.
- The basin doesn't cover everything: the Landian eschaton of technocapital eating everything seems outside it and anti-human.
- His current guess is that the trajectory where the main game was making AIs solve insane amounts of insane RL environments is cursed — and we may now be landing on another cursed direction.
More from AGI Musings
- Ex-DeepMind researcher who resigned over military ties warns of out-of-control AI self-improvement race — nordicinst · 2026-09-14
- Unions urged to halt AI datacenter buildout until jobs and grid use are protected — nordicinst · 2026-09-14
- Anthropic alignment lead agrees AI "kill all humans" odds exceed 10% in next decade — firstadopter · 2026-09-14
- 25 Fields Medalists warn of AI-math misalignment as @ramez pushes back with chess analogy — juanbenet · 2026-09-14
- Europe's Transformative AI Strategy met with skepticism: frontier AI is a practice, not an asset — sebkrier · 2026-09-14
- Reviewer pushes back: average ML papers are also "very meh" — jankulveit · 2026-09-14