Harvard physicist Matthew Schwartz writes on the Science Blog that there is an impedance mismatch between AI and science: large language models perform well, but are less effective as human collaborators. Schwartz therefore developed a toolkit for precise calculations, which he tested with Claude and domain experts, and which revealed connections across dozens of fields, including ecology and population genetics.
AI-generated text
Impedance mismatch: Matthew Schwartz's toolkit for precise calculations to address gaps between science and large language models
Harvard physicist Matthew Schwartz writes on the Science Blog that there is an impedance mismatch between AI and science: large language models perform well, but are less effective as human collaborators.
