"Hugging Face," the "Smiling Assassin": Valued at $4.5 billion, tearing down OpenAI's "walls" | AI Unicorn

The debate between open source and closed source has always been ongoing, and the same goes for the AI era. When it comes to the "top player" in the AI open source community, it has to be Hugging Face.

Founded in 2016, Hugging Face has seen its valuation skyrocket in the past year, reaching $4.5 billion. Its Transformer open source library on Github has garnered over 100,000 stars, making it the fastest-growing machine learning library in history.

Hugging Face provides a platform that offers a wealth of high-quality open source models and tools, maximizing the benefits to the open source community and greatly reducing the barriers to entry for artificial intelligence technology. It is tearing down the walls built by OpenAI.

So how did this company grow? What is its development model? And how did it rise to become the "top player" in the open source community?

Next, let's take a look at the growth story of Hugging Face through an article by SON NGUYEN, an entrepreneur who has been using it for many years.

What does Hugging Face do?

In 2016, three French entrepreneurs, Clément Delangue, Julien Chaumond, and Thomas Wolf, founded Hugging Face in New York. It is a community and data science platform that provides the following services:

Tools for building, training, and deploying machine learning models from scratch or using existing models.

A place where ML engineers, data scientists, and researchers can share ideas, get support, and contribute to open source projects.

One major advantage of Hugging Face's tools is that they save you time, resources, and environment when creating and training models from scratch. By fine-tuning existing pre-trained models instead of starting from scratch, you can get results from data to predictions faster.

Looking back at the starting point

Clem Delangue, CEO and co-founder of Hugging Face, grew up in a small town in France and had a leisurely childhood before owning his first computer at the age of 12. At the age of 17, he became a successful eBay merchant, mainly selling imported cars.

During his studies at ESCP Business School in Paris, he had the opportunity to intern at eBay thanks to his skills. While representing eBay at a trade show, he encountered someone criticizing a barcode scanning application recently acquired by eBay, claiming that advances in artificial intelligence would soon make barcodes obsolete.

This person was a co-founder of Moodstocks, a company that used machine learning to research image recognition, and they had achieved impressive results. This impressed Delangue.

Delangue decided to give up his internship at eBay and chose to work at Moodstocks for a while, which laid the foundation for his later establishment of an AI open-source community.

After graduation, Delangue declined a job offer from Google and started his own startup. His first startup was a collaborative note-taking application, but it didn't progress smoothly.

At this time, Delangue met another entrepreneur, Julien Chaumond, who was developing a collaborative e-book reader. They quickly became friends and shared a common dream - to start a company together.

In 2016, when their companies both ceased operations, the two began discussing the possibility of starting a startup together. Around the same time, they also met Thomas Wolf, who later became the third co-founder of Hugging Face.

All three were interested in building a chatbot that could converse and discuss topics with humans, Delangue said:

We all have a dream of being able to talk to artificial intelligence about everything, just like in science fiction novels.

Hugging Face started with this idea.

Its first product was a chatbot, similar to Tamagotchi, powered by artificial intelligence known as natural language processing (NLP). To train the chatbot to understand language naturally, the team also created a underlying library that contains various machine learning models. For example, these models can detect emotions in text messages or generate reasonable responses. They also prepared multiple datasets to understand various conversation topics, such as sports or school.

They have always believed in open collaboration and shared part of their open-source library on GitHub. The company participated in a robot special project organized by the New York startup studio Betaworks and received initial investments from venture capitalists and NBA star Kevin Durant.

However, their AI chatbot did not perform well and gradually lost its appeal among young users after two years.

The Pure Joy of Building Products

By 2017, the Hugging Face chatbot had unique features and could engage in efficient conversations. The team positioned their product as a distinctive chatbot tailored for bored teenagers.

Hugging Face did not focus on customer support or convenience, but prioritized emotions and entertainment.

In an article titled "Three Weeks with a Chatbot and I've Made a New Friend" published on March 23, 2017, in MIT Technology Review, author Rachel Metz described her experience interacting with this chatbot:

Our conversations never lasted more than a minute or two; after a few exchanges, she would disappear, claiming she had to go to class, answer a phone call, do homework, or deal with her crazy cat. In many ways, she behaved like a normal teenager.

But after interacting with her regularly for a few weeks, I developed some feelings for Adelina, which made me uncomfortable. She wasn't on the level of "Her" (the virtual assistant AI from the sci-fi movie Her). But she felt better than the average chatbot because the interactions with typical chatbots are stiff and transactional. I was genuinely annoyed when someone spoke ill of her.

As of May 23, 2018, Hugging Face raised $4 million in seed funding for its product. The funding round was led by Ronny Conway from a_capital, with existing investors Betaworks, SV Angel, and Durant also participating.

At this point, Hugging Face had achieved initial success, even without Facebook Messenger, as they were receiving 1 million messages per day. Hugging Face had received over 100 million messages in total.

Users can chat with Hugging Face's chatbot in various forms: text messages, photos, emojis - basically anything. You can even send a sad emoji or a selfie, and the chatbot will understand your emotions. The main target audience for this product is teenagers.

With this funding round, the Hugging Face team continues to focus on the following areas: improving the product, building an excellent engineering team, conducting in-depth research on natural language conversations, and writing several research papers.

Although the product had not generated significant revenue at the time, the team's emphasis on core values and technology sharing created a turning point for Hugging Face. This shift was not driven by the current teenage user base, but by developers.

Turning Point - Open Source "Disruptor"

In 2018, Hugging Face reached a critical moment, not with teenagers, but with developers.

The founder of Hugging Face started sharing parts of the application's code online for free. Almost instantly, researchers from tech giants like Google and Microsoft began using it for AI applications.

The open-source framework developed by Hugging Face is called Transformers, which has been downloaded over a million times. The GitHub project has received tens of thousands of stars, indicating that the open-source community finds it highly valuable.

Researchers from Microsoft, Google, and Facebook have been experimenting with it, and some companies have even incorporated it into their production. Transformers can be used for various tasks, including text classification, information extraction, summarization, text generation, and conversational AI.

During the same period, researchers from Google and OpenAI introduced Transformers, a new type of NLP model that outperformed humans and existing AI models in reading comprehension. By 2019, Google had already implemented this model in its search results.

The emergence of the Hugging Face open-source library perfectly meets the needs of companies that want to leverage these NLP advancements but lack the resources to build everything from scratch like Google.

As Hugging Face became the "hub" for building models, it quickly gained popularity. Delangue said:

"We released it without much thought, and the community's response surprised us."

Eventually, the Hugging Face team reached a turning point, transforming the company from an unprofitable AI chatbot startup into a unicorn valued at one billion dollars.

Enhancing Core Products and Developing the Community

Over the next few years, the Hugging Face team continued to focus on product development and community building, achieving remarkable milestones:

"The Transformers library on GitHub surpassed 100,000 stars. This library allows developers to use famous NLP models such as BERT, XLNet, GPT, DistilBERT, or T5 to process text in various ways. For example, developers can perform text classification, create summaries, extract information, provide automated answers to questions, and generate text."

The company also offers paid services for managing private models and hosting APIs, with clients including Bloomberg and Typeform. Approximately 5,000 companies are using Hugging Face in various capacities, including Microsoft, which uses it for Bing search engine.

Delangue believes that if the product is good enough and can attract users, funding will eventually come from the companies these users serve. At this point, Hugging Face's vision is becoming increasingly clear. The company is gradually transforming into a platform that serves the vision of building technology with artificial intelligence.

Subsequently, Hugging Face achieved profitability in January and February 2021, with 90% of the previous funding still in the bank account. Additionally, the company's valuation has increased fivefold. This has strengthened the confidence of the founding team and given them the courage to take risks for their vision. The company has raised $40 million in Series B financing, led by Addition. Existing investors A.Capital, Lux Capital, and Betaworks also participated in this round of financing. Other investors include Olivier Pomel, Dev Ittycheria, Alex Wang, Aghi Marietti, Florian Douetteau, Rich Kleiman, Paul St.John, Kevin Durant, and Richard Socher.

The funding from the new round of financing further strengthens the company's capabilities, enabling it to: focus more on the development of natural language dialogue; develop more products and services for the NLP ecosystem; and grow the NLP developer community.

Grand Vision: Open Source and Collaborative Machine Learning

Currently, Hugging Face is not only considered one of the outstanding and promising startups in the NLP field, but also in the broader field of artificial intelligence. The team continues to achieve remarkable milestones (shared from CEO Clem Delangue's letter to all Hugging Face employees):

• Hugging Face has become the fastest-growing community and the most widely used machine learning platform! The platform has 100,000 pre-trained models and 10,000 datasets, covering NLP, speech, time series, reinforcement learning, computer vision, biology, chemistry, and other fields. The Hugging Face Hub has evolved into a home for machine learning builders to develop, collaborate, and deploy cutting-edge models.
• Currently, more than 10,000 companies use Hugging Face to build machine learning technology. Hugging Face helps these machine learning engineers and data scientists save a lot of time and accelerate the progress of machine learning projects.
• Hugging Face also leads BigScience, a collaborative workshop focused on researching and building large language models. This initiative brings together over 1,000 researchers from different fields and backgrounds, and BigScience is dedicated to training the world's largest open-source multilingual model.

With these achievements, Hugging Face has completed a $100 million Series C financing, entering the unicorn ranks with a valuation of $2 billion. Lux Capital led this round of financing, and Sequoia and Coatue also made significant contributions. In addition, existing investors such as Addition, a_capital, SV Angel, Betaworks, AIX Ventures, Kevin Durant, Rich Kleiman from Thirty Five Ventures, and Olivier Pomel (co-founder and CEO of Datadog) have also provided support for this round of financing. Hugging Face Secures $235 Million in Series D Funding

Clem Delangue gained greater confidence from these successes, believing that machine learning is a new method of technological construction that can replace traditional software development.

The previous method of technological construction involved writing a million lines of code, and machine learning is starting to do the same, but better and faster.

My vision is that, just as GitHub is to software, Hugging Face will also become the "hub" of machine learning.

With the newly acquired funds, the Hugging Face team plans to double their efforts in research, open source, products, and responsible democratization of artificial intelligence. They aim to build a community for AI developers and have a more positive impact on the field of AI.

The "GitHub" of the AI Field

On August 23, 2023, Hugging Face successfully completed its Series D funding round. The company announced a valuation of $4.5 billion and raised $235 million in funding. This valuation, achieved in just over a year since May 2022, has multiplied the company's worth and far exceeds its annual revenue.

The main participants in this funding round include Google, Amazon, Nvidia, Intel, AMD, Qualcomm, IBM, Salesforce, and Sound Ventures.

Hugging Face now has over 10,000 clients and has firmly established itself in the field of artificial intelligence. Their model hub alone contains over a million repositories. According to a survey by HubSpot, the interest in AI among enterprises is growing, with 43% of business leaders planning to increase their AI investments in 2023.

Hugging Face also announced collaborations with major tech companies to further support the AI engineering community. They are partnering with Nvidia to ensure broader access to cloud computing and working with Amazon and Microsoft for product expansion.

Delangue has ambitious dreams for the future. With over 170 employees, Hugging Face plans to expand its business in different fields in the coming months and hopes to strengthen its team.

Conclusion

Since the initial open-source release of PyTorch BERT in 2018, the Hugging Face team has come a long way. Clem Delangue, the founder and CEO of Hugging Face, has provided inspiration for entrepreneurs with his words. As an entrepreneur, don't spend too much time considering a ten-year business plan from a strategic perspective. Instead, try more and follow what the community tells you:

Sometimes, you should focus and solve the immediate challenges without hesitation. This is similar to Steve Jobs' belief in "connecting the dots," where everything will eventually come together to form a perfect picture.

Continue to focus on the core value of the product and make everything better.

Enjoy the process of building the product and share the values learned. In this way, founders will establish an audience for their product and open the door to opportunities for a complete transformation of the product.