Alignment is stuck because it lacks MNIST-style benchmarks, researcher argues
joshua_saxe · x · 2026-10-13
Joshua Saxe argues alignment progress stalls because the problems aren't well posed and lack real scientific infrastructure: ML advanced via a series of capability hills (MNIST, ImageNet, HellaSwag) that made results commensurable among competing strangers, while interpretability and alignment have no analogous targets. Reposter adds: Goodhart's law is tough, but try having no target at all.
More from AGI Musings
- Yacine: all software is trash, so with AI it's now reasonable to just build it yourself — yacineMTB · 2026-10-13
- Self-replicating AI skeptics: it runs in datacenters, not your laptop — yacineMTB · 2026-10-13
- Can AI explain medical uncertainty without panicking patients? — yi111 · 2026-10-13
- Filtering the Singularity: why low-frequency trends now matter more than any single AI breakthrough — akbirthko · 2026-10-13
- Embedded evaluators can't fix companies that choose to be bad, Stella Biderman argues — BlancheMinerva · 2026-10-13
- Is consciousness something you do, not have? AI as an evolutionary handoff — VoidStateKate · 2026-10-13