Despite Hugging Face's significant funding and valuation, H4 does not directly monetize its tools, instead contributing to Hugging Face’s Expert Acceleration Program. The team collaborates with external groups on joint releases and is currently exploring different alignment techniques. H4 recently released a handbook containing the source code and data sets used to build Zephyr and plans to update it with future AI models. The team emphasizes its intention to empower the open AI community, not compete with other open-source AI initiatives.
Key takeaways:
- Hugging Face's two-person team, H4, is developing tools to enable the AI community to build AI-powered chatbots similar to ChatGPT.
- H4 is behind several open source large language models, including Zephyr-7B-α and a modified version of Falcon-40B, and relies on a dedicated cluster of more than 1,000 Nvidia A100 GPUs for training.
- The team recently released a handbook containing all the source code and data sets used to build Zephyr, and plans to update it with code from future AI models.
- While H4 doesn't directly monetize its tools, they do contribute to Hugging Face’s Expert Acceleration Program, which provides guidance to build custom AI solutions.