Chinese computer scientist Kai-Fu Lee, famous for his bestseller AI Superpowers, has launched a new venture called 01.AI. The company aims to develop a homegrown large language model (LLM) specifically designed for the Chinese market. Lee’s motivation for starting 01.AI stems from the fact that China does not have access to OpenAI and Google’s LLMs. In response to this need, Lee and his team at 01.AI have released their first model, the open-source Yi-34B.
Key Takeaway
Kai-Fu Lee’s 01.AI has released an open-source large language model, Yi-34B, to cater specifically to the Chinese market. The company aims to address the lack of access to OpenAI and Google’s models in China. With successful fundraising, talent acquisition, and plans for monetization, 01.AI strives to become an ecosystem that fosters application development using their LLMs.
Focus on the Chinese Market
01.AI’s goal is to provide a compelling alternative to the language models available in the rest of the world. By developing their own LLM specifically for the Chinese language, Lee and his team aim to address the unique needs of the Chinese market. The Yi-34B model, a bilingual (English and Chinese) base model trained with 34 billion parameters, has already gained recognition in the field.
Competing with Larger Models
Despite being smaller than other open models, such as Falcon-180B and Meta LlaMa2-70B, Yi-34B has outperformed them in terms of pre-trained LLM models. Kai-Fu Lee expressed his confidence in the future development of their models, stating that upcoming models with 100 billion to 400 billion parameters will be significantly better than the current release.
Successful Fundraising and Talent Acquisition
To support its operations, 01.AI has successfully raised undisclosed funding from prominent investors, including Sinovation Ventures and Alibaba Cloud. The company has already grown to over 100 employees, with more than half of them being LLM experts from major multinational and Chinese tech firms. The team behind Yi-34B includes experienced individuals from Microsoft Research Asia, Google Brain, Huawei, and Alibaba.
Monetization Strategy and U.S. Sanctions
01.AI is aware of the cost implications of building LLMs and has plans for monetization from the start. While the company will continue to open source some of its models, its main objective is to develop a state-of-the-art proprietary model as a foundation for a range of commercial products. To ensure a stable supply of AI processors, the company proactively stockpiled GPUs in anticipation of U.S. sanctions, allowing them to secure the necessary resources for the next 12-18 months.
Building an Ecosystem for Application Development
Looking ahead, 01.AI aims to become an ecosystem where developers can easily build applications using their LLMs. The company is experimenting with concepts in productivity and social directions, and there are hints of an app release within the current calendar year. Kai-Fu Lee believes that China has the potential to excel in building LLM applications due to its strong mobile internet ecosystem.
