A Near-Autonomous AI Chemist Improves a Challenging Reaction
This is a concrete example of AI moving from literature and code into a closed-loop experimental setting. OpenAI describes the workflow as near-autonomous because human chemists still supplied high-level steering, judgment, lab preparation, and replication.
Results to watch
The system identified TEMPO as a useful additive for the studied primary sulfonamide Chan-Lam coupling. Across two cycles, Maria ran 10,080 reactions, large enough to test whether the improvement transferred across a broad substrate set.
ASI relevance
The important pattern is not “AI replaced chemists.” It is that models can propose research moves, operate with lab automation, and learn from large experimental batches. That is one bridge from software-only AI R&D automation to physical scientific discovery.