Featured Speaker Webinar with Lee Branstetter:Quantifying the Impact of AI Invention on Productivity

Why It Matters

By improving identification of AI invention and tying patents to firms, the research provides a scalable way to measure AI’s contribution to productivity growth and informs policymakers and firms about where impactful AI innovation is occurring. This helps resolve divergent findings from surveys, hiring data, and RCTs and guides targeted industrial and innovation policy.

Summary

Professor Lee Branstetter presented new empirical work that uses large language models to classify AI-related patents across the U.S. patent corpus and link those inventions to firms using confidential U.S. Census microdata. The project recreates an expanded, human-validated AI patent database and shows exponential growth in AI patenting, with accelerations around 2012 (deep learning) and 2018 (transformers). Branstetter contrasts this invention-focused measure with surveys of AI adoption and hiring studies—both of which show modest or mixed productivity effects—while noting randomized trials often find large, context-specific gains. His firm-level linkage aims to quantify how AI invention, distinct from adoption or hiring, contributes to productivity at scale.

Original Description

Professor Lee Branstetter presents recent research on AI innovation and productivity using large-scale patent and firm-level data. He examines how AI inventions and top AI talent are reshaping productivity and employment, and highlight the growing role of Japanese firms and Asia in the evolving AI landscape.