In this series of ai.nl, we feature the most promising AI companies in the world: where do they come from, what have they achieved and what are their plans for the future. In this episode, we are looking at Hugging Face, a New York-based AI company allowing build, train, and deploy art models using the reference open source in machine learning.
Hugging Face calls itself the “AI community building the future” and has become one of the most influential companies in the field of natural language processing and machine learning. The story of Hugging Face is essentially that of a chatbot company transforming the entire AI industry with its focus on open source technology, AI ethics, and offerings like easy to deploy tools.
A $2B machine learning revolution
Named after hugging face emoji, the company was founded in 2016 by Clément Delangue and Julien Chaumond. It formally entered the AI scene with the launch of a chatbot app for iPhone. This chatbot shared selfies of its computer-generated face and cracked jokes. It was an attempt by Delangue and Chaumond to offer an app that could entertain those feeling boredom.
However, the chatbot never made any real money and five years later, Hugging Face is now valued at $2B. According to TechCrunch, Hugging Face closed a $100M Series C round last week that gave the company an implied valuation of $2B and that valuation is also the tale of a company evolving from fun to become most interesting thing in AI.
Hugging Face’s transformation into the most interesting AI company started in 2018 when its founders began sharing some of their chatbot app’s underlying code online for free. Their code was so efficient that even big tech giants like Google and Microsoft began using it for their own AI applications.
This also paved the way for Hugging Face to retire its chatbot app and instead transform itself into a platform that offered ready-to-use machine learning models. If GitHub is the central depot for software then Hugging Face is GitHub for machine learning models.
Hugging Face’s Transformers library on Github
The release of Transformers library by Hugging Face on GitHub completely changed the fortunes of this ML company. It stopped being another messaging application and became the pioneer of AI models. The library on GitHub has 62,000 starts and 14,000 forks, and offers APIs to easily download and train state-of-the-art pretrained models.
On Hugging Face’s website, you can now browse for thousands of pre-trained machine learning models, join the developer community to submit your own model, download datasets, and collaborate with other ML engineers. The popular NLP models available through Transformers include BERT, GPT-2, T5 or DistilBERT.
In the field of natural language processing, Hugging Face offers features such as summarisation, question answering, table question answering, text classification, and fill mask that can be deployed in AI models by NLP practitioners. It also offers features such as sentence similarity, translation, token classification, and feature extraction.
For audio, it offers features such as text to speech, audio to audio classification, voice activity detection, and automatic speech recognition. It also offers image classification, object detection, image segmentation, and text to image features as part of its computer vision suite. All of these features are already being used and deployed by developers across their AI and ML products.
Hugging Face: key members
- Clement Delangue (Co-founder and CEO)
- Julien Chaumond (Co-founder)
- Thomas Wolf (Chief Science Officer)
- Anna Tordjmann (Chief Legal Officer)
- Anthony Moi (Technical Lead)
- Clara Ma (Chief of Staff)
Hugging Face: timeline of major events since its founding
- March 2017: Hugging Face releases its AI chatbot app
- May 2018: Hugging Face raises a $4M seed round
- March 2021: Hugging Face raises a $40M series B round
- December 2021: Hugging Face acquires Gradio
- May 2022: Hugging Face raises a $100M series C round
Hugging Face: what’s next for machine learning pioneer
Hugging Face naturally began its journey as a natural language processing startup but was quick enough to identify the evolving industry trend. Within a year, it became the main repository for all things related to machine learning models. The course of its actions suggest that Hugging Face is building the GitHub of machine learning and it is doing so with a platform that is community-driven and has a ton of repositories to choose from.
Clément Delangue, co-founder and CEO of Hugging Face, goes as far as to say that machine learning is the new way to build technology and it is replacing software. “The old school of building technology was writing a million lines of code. Machine learning is starting to do that, but much better and much faster,” he says in an interview with Forbes.
This is a clear reflection of a founder looking to replicate the success of GitHub, but for machine learning. At the time of its $7.5B sale to Microsoft in 2018, GitHub had recorded revenue of $300M and Forbes says Hugging Face had revenue of less than $10M last year.
Delangue does not reveal the exact revenue figures of his fledgling machine learning startup but sees a future where Hugging Face can make billions in revenue with its AI-minded developers. Like the emoji, there is a lot working in favour of Hugging Face right now and nobody knows that better than Lux Capital, which first invested in the startup in 2019 and led the recent $100M round.
Either by design or virtue, machine learning is difficult and the only way for a number of developers to successfully build good ML models is by collaboration. If managers collaborate with their peers on LinkedIn, then ML developers will collaborate with their peers on Hugging Face. This could result in creation, discovery and collaboration on ML models, datasets, and even ML apps.
In addition to its Inference API, which allows developers to use thousands of models via a programming interface and support for automatically training models, Hugging Face offers a suite of services that are designed to accelerate development of ML models. With 10,000 companies using Hugging Face right now, the future of the company is very much centred around the vision of turning it into a GitHub for machine learning.
The success of Hugging Face will reshape the artificial intelligence industry, accelerate development of ML models, and create a new form of collaborative and open-source infrastructure for developers. This could, in effect, lead to a future where AI-powered technologies are omnipresent and power is not vested in the hands of a few tech companies.
What to read next?
- 🧑🏫 OpenAI wants to be the first to build artificial general intelligence: a look at its history, key members, and major achievements like GPT-3
- 🕶️ Darktrace wants to be the torchbearer for AI-powered cyber defence: a look at its history, key members, and major achievements
- 🧠 DeepMind wants to build a general-purpose AI: a look at its history, key members, and major achievements like AlphaGo