Extent of Automatable Scientific Work

Determine the extent to which large parts of scientists’ daily research work can be automated.

Background

Beyond the feasibility of LLMs assisting in research reasoning, the authors identify uncertainty over how much of scientists’ routine tasks can be automated. Their paper examines a structured analytic workflow (Hartree-Fock derivations) to understand which components are amenable to automation.

The paper’s results provide data points on automating specific analytic steps, but the broader quantitative extent of automatable work in scientific practice remains unspecified and is flagged as unclear.