Harvard Physicist Argues for AI-Science 'Impedance Mismatch' and Builds Toolkit for Exact Calculations

Key Info

Harvard physicist Matthew Schwartz argues in a Science Blog guest post that AI and science suffer an "impedance mismatch" — using LLMs like human collaborators isn't the best way to draw out their scientific strengths — and he built a toolkit for exact calculations in quantitative science.

Highlights

• Schwartz reframes the problem as an impedance mismatch: two capable systems poorly matched in how they're used together.
• He created a toolkit for exact calculations in quantitative science rather than relying on freeform collaboration.
• Because similar calculations recur across disciplines, Claude surfaced links to ecology, population genetics, and a dozen other fields.
• Schwartz worked with domain experts to validate and extend the connections the toolkit surfaced.