Measuring the Openness of AI Foundation Models: Competition and Policy Implications

Sciences Po Digital, Governance and Sovereignty Chair, Working Paper

35 Pages Posted:

See all articles by Thibault Schrepel

Thibault Schrepel

Vrije Universiteit Amsterdam; Stanford University's Codex Center; University Paris 1 Panthéon-Sorbonne; Institut d'Etudes Politiques de Paris (Sciences Po)

Jason Potts

Royal Melbourne Institute of Technolog (RMIT University)

Date Written: May 14, 2024

Abstract

This paper presents the first comprehensive evaluation of AI foundation model licenses as drivers of innovation commons. It introduces a novel methodology for assessing the openness of AI foundation models and applies this approach across prominent models such as OpenAI’s GPT-4, Meta’s Llama 3, Google’s Gemini, Mistral’s 8x7B, and MidJourney’s V6. The results yield practical policy recommendations and focal points for competition agencies.

Keywords: Foundation models, large language models, LLMs, generative AI, competition, antitrust, open source, innovation commons, public policy, ChatGPT

JEL Classification: K21, K20, L17, L86, E14, B52, K24, 030, 031, 032, 033, 034, 038

Suggested Citation

Schrepel, Thibault and Potts, Jason, Measuring the Openness of AI Foundation Models: Competition and Policy Implications (May 14, 2024). Sciences Po Digital, Governance and Sovereignty Chair, Working Paper, Available at SSRN: https://ssrn.com/abstract=

Thibault Schrepel (Contact Author)

Vrije Universiteit Amsterdam ( email )

De Boelelaan 1105
1081 HV Amsterdam
Netherlands

Stanford University's Codex Center ( email )

559 Nathan Abbott Way
Stanford, CA 94305-8610
United States

University Paris 1 Panthéon-Sorbonne ( email )

France

Institut d'Etudes Politiques de Paris (Sciences Po) ( email )

27 rue Saint-Guillaume
Paris Cedex 07, 75337
France

Jason Potts

Royal Melbourne Institute of Technolog (RMIT University) ( email )

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
581
Abstract Views
1,267
PlumX Metrics