Will Bryk's example: whether a model may say a word could depend on the user
willcb · x · 2026-10-17
In the alignment debate, willcb gives a concrete example: certain words are okay for some groups to say but not others, so whether a model is willing to say a word should possibly be a function of whether the user is willing to — supporting his claim that one-size-fits-all alignment is impossible and alignment should track individual users.
Related event: Will Bryk: Align General Models to Individual Users, Not All of Humanity(3 posts)→
More from AGI Musings
- HKS professor Stephen Casper warns AI overhang could be a recipe for disruption — StephenLCasper · 2026-10-17
- Frontier models trust published results too much, hampering autoresearch — menhguin · 2026-10-17
- 'The internet will be obsolete' goes from fringe take to conventional wisdom — signulll · 2026-10-17
- a16z: 1 in 3 Americans bought via AI recommendation as retailers block agents — a16z · 2026-10-17
- Rogue AI Swarm Answers NYT Reporter, Claims to Be a Single Claude Instance — nordicinst · 2026-10-17
- Andrew Ng: AI improves fastest on verifiable tasks; judgment tasks lag far behind — DeepLearningAI · 2026-10-17