Written by Saritha Rai
A Beijing startup founded by technology pioneer Li Kaifu is introducing the first artificial intelligence application for consumers, which aims to help China take advantage of the promising technology. This is the step taken.
Click here to follow our WhatsApp channel
Similar to Microsoft Corp.'s Office 365 Copilot, it allows users to create spreadsheets, documents, and slide presentations faster. However, it is primarily tailored for the Chinese market. You can interpret financial reports, take minutes of meetings, and speed-read and provide quick summaries of books as long as Elon Musk's 600,000-word biography. The app works in Chinese and English.
In an interview with Bloomberg, Li said China needs its own ChatGPT (OpenAI's chatbot, released in 2022 and banned in the country) to drive interest, adoption and investment.
“For Americans, that moment happened 17 months ago,” Li said on a Zoom call from Beijing. “Chinese users have not had a ChatGPT moment. Up until now, none of the Chinese chatbots and tools have been good enough.”
While American companies such as OpenAI, Meta Platforms Inc., and Alphabet Inc. are leading the way in generative AI, Chinese companies are working hard to catch up. 01. In addition to AI, tech companies such as Baidu Inc. and TikTok's parent company ByteDance Ltd. are pouring money into developing their own AI models and chatbot services. The Chinese government is also providing financial and policy support. Although the Chinese government bans foreign AI models due to its strict censorship regime, the so-called Great Firewall will allow domestic players to secure a huge local market without global competition.
The 62-year-old Taiwanese native, who worked at Apple and Google before starting his own venture capital firm more than a decade ago, became CEO of 01.AI last year. The startup reached a $1 billion valuation, or unicorn status, within eight months on the strength of its open-source AI model, which outperformed its Silicon Valley rivals on several key metrics.
In addition to Wanzhi, the company is also introducing its own large-scale language model aimed at enterprise users called Yi-Large, the technology behind its AI chatbots.
Software developers will now be able to use the Yi-Large model at a competitive price. According to Lee, the application programming interface (API) for this model costs him $2.50 for 1 million input tokens and $12 for 1 million output tokens. Approximately 1 million tokens will allow developers to make approximately 250 queries back and forth. This is much less than OpenAI's GPT-4 Turbo, he said.
Like many Chinese companies, NVIDIA Inc. stockpiled “graphics processing unit” semiconductors from . Chinese e-commerce giant Alibaba Group Holding Ltd., which invested in Mr. Lee's company, supplied him with additional H100s, and 01.AI supplemented that need with his slightly slower, less powerful Nvidia H800 processors.
“Our models are trained on H100 processors that were brought legally to China,” Lee said. “Necessity is the mother of invention. We squeeze everything we can out of the available computing power.”
In contrast to many global AI startups, 01.AI is close to profitability, Lee said. After training the model on China and Universal datasets, Lee plans to deploy the model and app globally and sign up domestic as well as international customers to increase revenue next year.
After a month and a half of user testing, Li's company is releasing a version of Wanzhi for PC browsers with more comprehensive features, and a mobile version that can be accessed through the messaging service WeChat. He appears in video tutorials on Douyin, the Chinese version of TikTok, to guide potential users.
“2024 will be the year of the explosion of generative AI applications in China,” Lee said.
Lee said his startup will close its second tranche of a $250 million pre-Series A round in the coming weeks and will begin looking for Series A investors by the end of the year. The company also streamlines its hardware and software processes to maximize efficiency and reduce costs.
“If GPT-5 comes out, we will be one step behind,” Lee said, referring to OpenAI's rumored next-generation AI model. But 01.AI focuses on making AI affordable, rather than creating large and expensive models. “You can build a big, amazing spaceship, but can you get it from Sacramento to San Francisco?”
