Connectionism vs symbolism: Lampinen cites failed replication in AI symbol debate
jeffrey_bowers · x · 2026-10-07
A debate between jeffreybowers and DeepMind researcher Andrew Lampinen over whether symbol systems add new competencies or merely bypass human cognitive limits. Bowers argues LLMs using external symbols for math fulfills classic PDP connectionism; Lampinen counters that learned, imperfect symbolic skills don't contradict connectionism, and notes a famously cited innate-symbolism study failed to replicate.
More from AGI Musings
- Dostoevsky saw it in 1864: show humans a prediction and they'll break it — billhilf · 2026-10-08
- With Math Delegated to OpenAI, Economics Research Faces a Reshuffle — soumitrashukla9 · 2026-10-08
- Higgsfield cofounder: SWE demand will collapse by end of 2027 — SimplyAnnisa · 2026-10-08
- AI ethics scholar: frontier leaders 'won't slow down until they feel like God' — LuizaJarovsky · 2026-10-08
- Yudkowsky as the Marx of our generation: an 'AI safety welfare state' thought experiment — teortaxesTex · 2026-10-08
- Math's search space is too absurd — autoregressive AI could seed new questions — akbirthko · 2026-10-08