The newsletter also discusses the Foundation Model Transparency Index, a new initiative from Stanford's CRFM lab that aims to define openness standards for open models. The index scores models and labs on 100 points of transparency, allowing them to choose their level of transparency. The newsletter also mentions the first AI Engineer Summit and the various talks and workshops that took place, as well as the latest developments in AI from companies like Anthropic, Modal Labs, and Anyscale.
Key takeaways:
- Mistral 7B, an open-source AI model, has been released and is rumored to be raising $400m at a $2.5b valuation from a16z. It is smaller but better than Llama 2 and has significantly impacted the open models landscape.
- Stanford's Foundation Model Transparency Index is a new initiative that defines openness standards for AI models, scoring models and labs on 100 points of transparency.
- There is a growing interest in custom finetuning of AI models, with companies like Brev and WandB leading the way. However, there are concerns about the openness of models, particularly when their datasets are not open.
- October saw the first AI Engineer Summit, a significant event where many developers met in person for the first time. The event sparked debates and discussions on various AI topics, including GitHub's Copilot product journey and its $100m ARR milestone.