AI needs science's search history: decision paths beyond papers are key to research taste

nathanbenaich · x · 2026-08-13

Nathan Benaich of Air Street Press argues that research papers only capture the happy path of scientific discovery. To give AI authentic scientific taste, it must learn from the branches that were considered, rejected, and never written down. The article cites a Cambridge lab that logs the winding path of discovery, turning ideas into a living graph. The author calls this a trillion-dollar dilemma, suggesting centuries of scientific records may be insufficient.

Related event: Experts Urge AI to Learn from Failed Scientific Experiments(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →