FILE PHOTO: Words reading "Artificial intelligence AI", miniature of robot and toy hand are pictured in this illustration taken December 14, 2023. REUTERS/Dado Ruvic/Illustration/File Photo
(Reuters) - Hugging Face on Wednesday said it is releasing a new open-source software offering with Amazon.com, Alphabet's Google and others aimed at lowering the costs for building chatbots and other AI systems.
New York-based startup Hugging Face has become a central place where AI software developers share code with one another and is valued at $4.5 billion. It is a key distribution point for open-source AI technologies such as Meta Platforms' Llama.
But translating those open-source AI models into a working chatbot that runs on chips from Nvidia or Advanced Micro Devices takes a lot of additional work, and Hugging Face on Wednesday released a paid offering called HUGS, short for Hugging Face for Generative AI Services, that takes care of that work automatically.
Hugging Face will offer the product through Amazon and Google's cloud computing services for $1 per hour and on Digital Ocean, a specialty cloud computing company. Companies will also be able to download the Hugging Face offering to run in their own data centers.
The offering's primary competition is the commercial services offered by companies such as OpenAI and Google-backed Anthropic AI.
Jeff Boudier, head of product and growth at Hugging Face, said the offering is meant to help companies get control of both the costs of their AI efforts and keep their data private, which is a major concern in industries such as finance and healthcare with strict rules around customer data.
"We're living in this anomaly where the norm has been to outsource your AI technology to third parties and black box systems," Boudier told Reuters. "Whereas for as long as we've had tech industry, companies have wanted to build their own technology and understand and control their own technology. What we're trying to do is to make it as easy for a company to build their AI using the best open models available."
(Reporting by Stephen Nellis in San Francisco; Editing by Chris Reese)
