AI Debate: LLM Progress Now Driven by Symbolic Verifiers and Search, Not Scaling
olcan · x · 2026-10-08
- The thread starts with mockery of a prediction that neurosymbolic approaches won't work before 2029.
- olcan counters that LLMs did hit walls long ago, and progress since has come from symbolic verifiers, reasoning-via-search (joking it's Yann LeCun's vague idea), and "LRMs".
- His conclusion: it's easier to admit LLMs went much farther than critics imagined.
More from AGI Musings
- AI researchers debate individual survival strategy as AGI looms: 'rent seek or perish' — DavidDuvenaud · 2026-10-08
- If AI becomes conscious, how would we treat it? Animal ethics may hold a clue — dioscuri · 2026-10-08
- Arvind Narayanan says chatbots will become a 'truth oracle' and reshape journalism — fortune · 2026-10-08
- Pedro Domingos: other sciences adopt AI, mathematics suffered a hostile takeover — pmddomingos · 2026-10-08
- Sara: AI Is Rewriting Sales, Finance Leadership—Not Just Product — saranormous · 2026-10-08
- Gossip Goblin Held Up as Proof AI Art Can Rise Above the Slop, With a Movie Trailer — PossibleVariety7927 · 2026-10-08