Fresh off $2 billion valuation, ML platform Hugging Face touts ‘open and collaborative approach’


We’re excited to bring Transform 2022 back in person on July 19 and pretty much July 20-28. Join AI and data leaders for insightful conversations and exciting networking opportunities. Register today!

Earlier today, the community-driven machine learning (ML) platform announced Hugging Face $100 million in new funding — raised in just a week — to continue building what many, including CEO Clement Delangue, are calling the “GitHub of machine learning.”

“I think that’s a correct analogy,” he told VentureBeat. “With every new technology, there’s a new category-defining platform that builds it. GitHub was it for software and it looks like we are becoming the platform for machine learning.”

Founded in 2016, Hugging Face evolved from a developer of natural language processing technology (NPL) to an open-source library and community platform where popular NLP models such as BERT, GPT-2, T5, and DistilBERT are available. Now it has moved beyond NLP to become an ML model hub and community – Hugging Face is working closely with companies that could be seen as competitors as companies such as Meta’s AI division, Amazon Web Services, Microsoft and Google AI use the platform.

“We have seen the emergence of a new generation of machine learning architecture called Transformers, which is based on transfer learning,” says Delangue. “Most of the users of this new generation of models are using them through our platform – it all started with text, but is now starting to make its way into all domains of machine learning, which is a new development for machine learning tools.”

A focus on ethical AI

Hugging Face recently hired some notable people in the AI ​​ethical space, which Delangue says is a key priority. Margaret Mitchellpreviously the head of Google’s ethical AI research group, came on board in August 2021. And Giada Pistillic, who has a Ph.D. in philosophy and specialized in conversational AI ethics, just started Hugging Face today.

“It’s good timing – someone with a Ph.D. in philosophy is a pretty unusual hire for a technology company, but I think it’s a testament to our commitment to making the machine learning field more inspired, which is what Margaret Mitchell likes to say,” Delangue said.

Delangue added that Hugging Face has a “strong outlook” on the future of AI and ML. “Just as science has always worked by making the field open and collaborative, we believe there is a great risk that the power of machine learning will remain highly concentrated in the hands of a few players, especially when those players are not capable have been of service in doing the good for the community,” he said. “By building more openly and collaboratively within the ecosystem, we can make machine learning a positive technology for everyone and work on some of the short-term challenges we see.”

An ‘open and collaborative’ ML evolution

Delangue said Hugging Face plans to continue growing his team from diverse backgrounds across all functions and capabilities, from science and engineering to product and business. “That’s a big evolution for us,” he said. “We also hope to see the number of models and datasets on the platform grow.”

The company is also excited about great science, a one-year research project on large multilingual models and datasets. “It’s the largest machine learning collaboration we lead with more than a thousand scientists and 200 organizations, inspired by other major scientific collaborations like in physics,” says Delangue. “We wanted to create something like this for machine learning.”

But it’s Hugging Face’s emphasis on an open collaboration approach that Delangue said investors were confident in the $2 billion valuation. “That’s really important to us, makes us successful and makes us different from others in space.”

VentureBeat’s mission is to be a digital city square for tech decision makers to learn about transformative business technology and transactions. Learn more about membership.



Leave a Reply

Your email address will not be published.