Valued at $1 billion, Kai-Fu Lee’s LLM startup introduces open source model

Kai-Fu Lee, the computer scientist known in the West for his bestseller AI Superpowers and in China for his bets on artificial intelligence unicorns, has a new company – and big ambitions.

In late March, Lee founded a company called 01.AI with the vision of building a homegrown large-scale language model for the Chinese market. The company puts him in competition with other prominent Chinese tech leaders, including Sogou’s founder Wang Xiaochuan, who have quickly gathered talent and venture capital to establish China’s equivalent of OpenAI.

“I think necessity is the mother of innovation, and there is clearly a great need in China,” Lee, who is 61 and leads 01.AI as CEO, said in an interview with TechCrunch, explaining the motive for founding it the company. “Unlike the rest of the world, China does not have access to OpenAI and Google because these two companies have not made their products available in China. So I think a lot of LLM graduates are trying to do their part in developing a solution for a market. “It really needs that.”

The growth of 01.AI appropriately reflects the rapid development in the field of generative AI. Seven months after its founding, the startup released its first model, the open source Yi-34B. The decision to launch an open LLM as a debut product is a way to “give something back” to society, Lee said. For people who felt that LLaMA was a “godsend,” “we have provided a compelling alternative,” he added.

At the time of writing, Yi-34B, a bilingual (English and Chinese) base model trained with 34 billion parameters and significantly smaller than other open models such as Falcon-180B and Meta LlaMa2-70B, ranked first among pre-trained LLMs models, according to a ranking by Hugging Face.

“We still believe that larger models, when trained well on a large amount of high-quality data, will always substantially outperform smaller models of comparable quality and comparable technology, I think.” [Yi-34B] Outperforming much larger models is something we don’t normally see,” Lee said. “We’re pretty confident that we’ll be bringing 100 to 400 billion models to market over the next year and a half. These models will be significantly better than the current model we announced.”

The startup’s ability to quickly begin model training is undoubtedly the result of its seamless fundraising, which is critical to securing top-notch talent and AI processors. While Lee wouldn’t disclose how much 01.AI has raised, he said the company is valued at $1 billion after receiving funding from Sinovation Ventures, Alibaba Cloud and other unnamed investors.

01.AI has already grown to more than 100 employees, more than half of whom are LLM experts from large multinational and Chinese technology companies. For example, its vice president of technology is an early member of Google’s Bard, and its chief architect was a founding member of TensorFlow and worked with renowned researchers such as Jeff Dean and Samy Bengio at Google Brain. The key figures behind Yi-34B are Wenhao Huang, a veteran of Microsoft Research Asia, and Ethan Dai, who held senior AI positions at Huawei and Alibaba.

With the backing of over ten unicorns and the founding of seven companies through Sinovation Ventures, Lee is perhaps one of the most well-connected investors and entrepreneurs in China.

“It’s been more than 25 years since I founded Microsoft Research Asia, and everything I’ve done has been about attracting great talent,” said Lee, of Microsoft Research Asia, the US giant’s largest research center abroad. before leading Google China. Over the years, Microsoft Research Asia has earned a reputation as the “West Point” for nurturing China’s AI entrepreneurs.

“Of course you want to pay people fairly, and you have to be competitive with pay, but I really think it’s also about people believing that they can make a difference and that the company can be successful,” Lee said during an appearance on a video call at 9:30 p.m. Beijing time. His employees were equally self-motivating. One of the startup’s infrastructure experts worked until the wee hours of the morning that day and still texted Lee at 2:15 a.m. to express his excitement about being part of 01.AI’s mission.

It’s no secret that pursuing LLMs is an expensive endeavor. To maintain its cash-intensive operations, 01.AI has monetization plans in place since the beginning. While the company will continue to open source some of its models, its goal is to develop a cutting-edge proprietary model that will serve as the foundation for a diverse range of commercial products.

“We can’t open source everything,” Lee said. “We were aware of the fact that these large language models require a lot of computing power and are therefore very expensive. If we raise a lot of money, most of it will be spent on the GPU. With that in mind, our first step was to get as much GPU as possible, which we did.”

Like other LLM players in China, 01.AI has proactively stockpiled GPUs in anticipation of US sanctions; It borrowed money to buy processors even before it received financing. Last year, the Biden administration tightened restrictions on China’s access to high-quality AI chips, prompting Chinese companies to pay inflated prices for chips. The foresight was rewarded – 01.AI now has a supply that will last for at least the next 12-18 months.

Aside from causing headaches for Chinese companies, the US sanctions have also been a catalyst for innovation by encouraging them to optimize the use of computing power. “With a very high-quality infrastructure team, we can potentially get 2,000 GPUs worth of workload out of every 1,000 GPUs,” Lee said.

01.AI’s path to monetization largely depends on its ability to find a product-market fit for its expensive AI models. While there is a shortage of top LLM scientists, there is no shortage of product talent in China.

“China is not ahead of the US in LLM, but there is no doubt that China can develop better applications than American developers, largely due to the phenomenal mobile Internet ecosystem that has been built over the last 12 years or so,” Lee argued.

While the founder didn’t provide any details about the services in the pipeline, he hinted that the company is experimenting with concepts in productivity and social and he would be “disappointed” if 01.AI doesn’t release an app this calendar year.

The startup’s ultimate goal, according to Lee, is to become an ecosystem where external developers can easily build applications. “The duty is not only to advance good research models, but more importantly to simplify application development so that there can be compelling applications,” he said. “At the end of the day. It’s an ecosystem game.” Time will tell whether Lee’s AI efforts will pay off.