Researcher: 'not quite there yet' on LLMs often means superhuman within a year
proteinrosh · x · 2026-11-04
A short exchange in the AI-for-science circle: the author argues that when people say "LLMs are good at X but not quite there on Z," it usually means superhuman performance on Z within 6 months to a year. He admits his sense of small-molecule SOTA is mostly vibes, but notes he's seen solid work where LLMs genuinely excel at analyzing small molecules, referencing an interesting post by Rowan and recent work from another researcher.
Related event: LLMs near SOTA on small-molecule optimization, but slow(3 posts)→
More from AGI Musings
- Scoble's Unaligned Newsletter: As AI Optimizes Everything, Human Messiness May Win — Scobleizer · 2026-11-04
- Denis Villeneuve sits down with Nina Beguš to discuss Arrival and AI's future — begusgasper · 2026-11-04
- PhD candidate researches solo founders as the post-GenAI default startup model — julianweisser · 2026-11-04
- Tolan's "character-driven computing": 250k monthly users chat with an AI that isn't a tool — ajaymehta · 2026-11-04
- What happened to science fiction? It tracks the speed of technological change — NerdyWeightLifter · 2026-11-04
- Biofx job candidates all use Copilot or web Claude, none know Claude Code — iskander · 2026-11-04