Statistics is More Like Auto-Formalization

roydanroy · x · 2026-07-16

The author argues that the field of statistics has not historically been organized around 'formalized open problems'. One reason is that the hardest part of many difficult problems isn't the proof itself, but clearly defining the problem in the first place.

Drawing an analogy to AI, he suggests that much of the truly impactful work in statistics is closer to auto-formalization than to theorem proving. In other words, the core challenge often lies in compressing ambiguous reality into a tractable problem representation.

Original post →

More from AGI Musings

AGI Musings channel →