The Open Source Initiative (OSI) is working to define open-source AI, with the aim of releasing a 1.0 version by the end of October 2024. The article also discusses the importance of open data sets and the challenges associated with them, such as the prevalence of personal information, toxic language, and low-quality sentences. It highlights the work of the OpenLLM France consortium, which is building open-source AI models and technologies for the French language, and its plans to expand to other European languages.
Key takeaways:
- The Open Source Initiative (OSI) is working to define open-source AI, with a 1.0 release expected by the end of October 2024. The definition will be based on the four basic freedoms applicable to open-source software: to use, study, modify, and share.
- There is a trend of imposing ethical restrictions on AI models through licensing, which can create barriers against the use and reuse of the models. These restrictions are not interoperable with free-software licenses.
- Open-source language models, particularly for non-English languages, are being developed by various organizations. However, many of these models are not truly open-source due to limitations on their use.
- OpenLLM France is working on a 100% open-source language model, Lucie, which will be trained with open-source data sets of French, English, German, Spanish, and Italian texts. The consortium is renaming itself to OpenLLM Europe with a mission to develop an open-source large language model for each European language.