Hugging Face introduces HuggingChat: A new open-source alternative to ChatGPT

Hugging Face, a prominent player in open-source AI development, recently introduced HuggingChat, a new alternative to ChatGPT. The user interface enables users to interact with an open-source chat assistant known as Open Assistant, which was designed by LAION, the nonprofit organization responsible for developing the data set that trained Stable Diffusion. Similar to other AI chatbot clients like Poe, HuggingChat will allows users to plug in new chat models.

HuggingChat's creation was based on the belief that open-source alternatives to ChatGPT are essential for more transparency, inclusivity, accountability, and power distribution. Hugging Face CEO, Clem Delangue, strongly advocates for open-source AI development. However, despite its promise, HuggingChat currently has several limitations. Furthermore, licensing issues need to be resolved before it can be used commercially.

According to Hugging Face, currently the hub doesn't save any chat data nor have any user accounts, and it doesn't have access to any Hugging Face account. However, in the future, there may be an option to share chat data with model providers for optimization purposes. The creators of HuggingChat have a vision to build an assistant of the future that can do more than just write emails and cover letters, but also perform meaningful tasks, use APIs, and research information dynamically. They aim to make it personalized and accessible to everyone, while ensuring it runs efficiently on consumer hardware.

Right now, the model is based on Meta's LLaMA which cannot be used for commercial purposes. Nevertheless, Hugging Face is working on the interface and safety mechanisms and plans to support the latest open-source models that are improving rapidly. Hugging Face is committed to fighting for open-source AI, and their long-term plan is to expose all high-quality chat models from the Hub, besides Open Assistant's latest LLaMA-based model. The HuggingChat code is fully accessible on GitHub, allowing anyone to run a copy and customize it, such as changing the model or adding UI elements.