In Search of Missing Mass: Disentangling Discoveries from Hallucinations

Recent AI models struggled to match a human algorithmic innovation

“Hallucinate -> do an experiment -> see it fails, try again. Hallucinate -> do an experiment -> it works, model innovated. Agentic models brush up against reality, this gives a way around the hallucina…”— robrenaud, when sharing this video · read the comment ↗