AI math breakthroughs keep proving earlier limits wrong, researcher says
herbiebradley · x · 2026-07-24
AI researcher Herbie Bradley argues that most AI math breakthroughs so far have actually been counterexamples to earlier assumptions.
- He says many people hope RSI will come from a 2x–3x jump in sample efficiency, but RL is still not very sample-efficient on long-horizon tasks.
- In his view, current systems still look like transformer-based pattern learners with variations, not something that truly models scientific discovery.
- He adds that models remain noticeably worse at novel discovery, intuition-based R&D, and innovation than human researchers.
- The math breakthroughs we’ve seen are often counterexamples to claims about what models supposedly cannot do.
More from AGI Musings
- A diagram says powerful agents need safety layers from weights to treaties — Paimaamu · 2026-07-24
- A Safety Stack for Powerful Agents Spans Sandboxes, Policy, and Treaties — joshua_saxe · 2026-07-24
- AI May Hit Women’s Jobs Nearly 3× Harder, If Office Work Gets Automated First — RealmLord-Wolf · 2026-07-24
- Research Gets Rebranded as an “Artisanal” Product in the AI Era — _akpiper · 2026-07-24
- 16 Insane AI Scenarios That Will Be Normal in 3 Years — _akpiper · 2026-07-24
- AI-written proofs should be treated as tools, not co-authors, says mathematician — SoloGen · 2026-07-24