AI safety will follow engineering tradition: formal proofs for simple cases, evals for complex

burny_tech · x · 2026-09-23

Replying to safety researcher David Krueger, AndyXAndersen argues AI safety will work like other engineering disciplines: formal proofs for simple cases and empirical evaluation for complex ones — treating AI as industrial machinery we already have experience managing.

Original post →

More from AGI Musings

AGI Musings channel →