Scholars Debate AI Safety: Value Alignment Far From Solved, OOD Generalization Remains a Flaw
davidmanheim · x · 2026-08-13
Countering the claim that technical safety for closed-weight AI is a solved problem, researcher Seth Lazar argues that current models only possess a deep analytical understanding of normativity that translates into practical alignment in distribution.
However, genuine out-of-distribution (OOD) generalization based on underlying values remains unsolved. Lazar notes this is partly a failure of generalization and partly a knowing/doing gap, emphasizing that value alignment itself is not yet achieved.
More from AGI Musings
- Sabine Hossenfelder fires back: academia's impact metric turned physics into a paper mill — skdh · 2026-10-16
- Researcher credits GPT for multiple accepted papers, incl. disproving Wegner's 1965 conjecture — ctjlewis · 2026-10-16
- dlwh: We optimized the legible version of our goals — and forgot the computation we meant to speed up — dlwh · 2026-10-16
- Ideas no longer have a waiting period: agents can build them as soon as they're articulable — akbirthko · 2026-10-16
- Ex-researcher: labs willing to cheat/distill ship great models faster than those building self-sufficient stacks — _aidan_clark_ · 2026-10-16
- Terry Tao reportedly still expects Riemann Hypothesis proof to be human+AI effort around 2050 — basedjensen · 2026-10-16