Alex Smolin is a Professor of Economics at the Toulouse School of Economics, holds a Chair at the Artificial and Natural Intelligence Toulouse Institute, and serves on the Editorial Board of The Review of Economic Studies. His research examines how information and incentives should be designed in strategic environments, with current work focusing on algorithmic advice, AI alignment, and the pricing of large language models.
Personal Website | Paper1 | Paper2
Abstract
AI is simultaneously a product supplied in markets and an adviser used to make decisions. This talk studies two complementary economic questions: how AI should be supplied and priced, and how it should be used when its objectives may not be fully aligned with those of the user.
The first part studies pricing and product design for large language models when users differ in their valuations across a wide range of tasks. Despite this high-dimensional heterogeneity, optimal contracts take a simple form such as committed-spend plans in which users purchase a budget of tokens. We discuss how this logic extends to multiple models and competition between proprietary and open-source systems.
The second part turns from the supply of AI to its use. How should a decision maker respond to an AI adviser that is informative but may be misaligned? A robust approach yields a simple trust-region principle: information is taken at face value when it remains within an endogenous region of trust, while sufficiently extreme advice is optimally discounted. Together, the two parts illustrate how economic design can shape both access to AI and reliance on it.
If you would like to give a presentation in a future webinar, contact our Senior Economist Dr. Wen Chen (wen.chen@luohanacademy.com). For other inquiries, please contact: events@luohanacademy.com.
