Ben Recht: what compressed sensing teaches us about AI's 'conquest of mathematics'

beenwrekt · x · 2026-09-18

UC Berkeley professor Ben Recht's new arg min post, "Applied Pure Mathematics," is his first attempt at grappling with AI companies' conquest of mathematics. Using the compressed sensing gold rush from his early career as a case study — deep math from geometric functional analysis and convex geometry that yielded faster MRI and better recommendation systems — he argues the theorems were never prescriptive: they assumed unverifiable facts about reality or demanded impractical measurement systems. Pure math's real value was as a narrative frame for design principles, new algorithms, and noise-robust schemes. This is the opening of a planned series on our new mathematical condition.

Original post →

More from AGI Musings

AGI Musings channel →