The paper introduces the Foundation Model Transparency Index, a system that assesses the transparency of major AI foundation models. This index uses 100 indicators to evaluate transparency, considering factors such as the resources used to build the model, the model’s details, and its usage. It is used to score major developers like OpenAI, Google, and Meta, based on their transparency practices for their flagship foundation models. The aim is to increase transparency in the AI industry to improve public accountability, scientific innovation, and governance.
Publication date: 19 Oct 2023
Project Page: https://arxiv.org/abs/2310.12941v1
Paper: https://arxiv.org/pdf/2310.12941
