Webinar by Christian Peukert, University of Lausanne: "Exposure Compensation: Strategic Behavior and Artificial Intelligence Training Data".
TECHNIS is pleased to invite you to a free webinar. TECHNIS webinars focus on IP and innovation examining recent legal, economic, managerial, ethical and policy issues related to technological innovation. Our approach is interdisciplinary and presentations are given by experts in different fields such as economics, law, management, STS, sociology, anthropology and philosophy. Webinar presentations last for 20min and are followed by a 40min discussion.
Please join us for a webinar on Tuesday the 27th of February 2024 at 11:00 London time i.e. 12:00 Brussels time, 13:00 Athens time. The speaker is Christian Peukert, HEC Lausanne (Faculty of Business and Economics at University of Lausanne). The title of the talk is “Exposure Compensation:
Strategic Behavior and Artificial Intelligence Training Data".
This webinar is free and open to all. The moderator is Dr. Andreas Panagopoulos.
Meeting ID: 964 7954 1729
NOTE: To participate please contact Andreas Panagopoulos at least an hour prior to the webinar.
Abstract: Human-created works represent critical data inputs to artificial intelligence (AI). Strategic behavior can play a major role for AI training datasets, be it in limiting access to existing works or in deciding which types of new works to create or whether to create new works at all. We examine creators’ behavioral change when their works become training data for AI. Specifically, we focus on contributors on Unsplash, a popular stock image platform with about 4 million high-quality photos and illustrations. In June 2020, Unsplash launched a research program by releasing a dataset of 25,000 randomly selected images for commercial use. We study contributors’ reactions, comparing contributors whose works were included in this dataset to contributors whose works were not included. Our results suggest that treated contributors deleted works from the platform at a higher-than-usual rate and substantially slowed down the rate of new uploads. The effect is long-lasting and we do not find evidence for heterogeneity across professional and amateur photographers. We also show that such strategic behavior affects the variety and novelty of contributions on the platform, with long-run implications for the stock of works potentially available for AI training. We discuss implications for copyright and AI policy.