Gary Marcus on the neurosymbolic debate: symbolic verification plus neural candidate generation
GaryMarcus · x · 2026-10-07
Debate over whether a recent AI math system counts as "neurosymbolic." Gary Marcus defines it: a system that uses a symbolic verifier plus a neural network producing many candidates, some of which survive, qualifies. Critics push back, arguing he has no basis for assuming Lean was used that way — on the Erdős forum Lean is reportedly only ever used for verification because writing proofs in it is painful.
More from Research
- Two Years After First Reasoning Model, AI Has Produced '20 Fields Medals' of New Math — __nmca__ · 2026-10-07
- NUS releases SafeActBench: 656 cases reveal where tool-using agents break the evidence-to-action chain — NationalUniversityofSingapore · 2026-10-07
- MEND: RL for flow models via proximal velocity matching beats Flow-GRPO in 100 vs ~4k updates — UTEXAS · 2026-10-07
- JLD: perceptual distance from a frozen encoder's Jacobian, fitted in 35s from 100 images, beats LPIPS and DISTS — Shreshth Saini · 2026-10-07
- HKUST surveys in-parameter memory: storing post-training knowledge in LLM weights instead of context — HKUST · 2026-10-07
- Joke thread pokes at claimed sub-n log n results: it's actually n(log n)^0.5477… — burny_tech · 2026-10-07