North Mini Code. Cohere's first model for developers.

A widely cited 2023 paper estimating that 80% of U.S. workers have tasks exposed to large language models is being misapplied in major policy decisions. The concept of an "exposure score," which estimates the tasks an AI could perform in a given occupation, has significant limitations that are not being properly considered. These scores are often based on a single model and a U.S.-centric job taxonomy, which compounds their constraints and leads to a mismatch with real-world application. The analysis argues for more dynamic, representative evidence and for policymakers and researchers to partner with workers to close the gap between available data and necessary policy decisions.