Keynote: Model Openness Framework

Ofer Hermoni presents the Generative AI Commons initiative from the Linux Foundation AI & Data, focusing on defining Open-Source AI. The talk introduces the Model Openness Framework, which categorizes AI model openness into three classes (Open Model, Open Tooling, Open Science) based on components like code, data, and documentation, and their associated licenses. Hermoni discusses the complexities of defining open-source AI, drawing parallels and distinctions with open-source software, and highlights the ongoing efforts by the Open Source Initiative (OSI) to establish a formal definition. The presentation touches upon the importance of global input in shaping these definitions and concludes with a discussion on the debate surrounding whether AI should be open-sourced at all.