Chinese startup 01.AI achieves unicorn status with open-source AI model Yi-34b.

In the tech sphere, 01.AI, a Chinese startup led by computer scientist Kai-Fu Lee, has gained attention for achieving unicorn status within a relatively short time frame, attaining a valuation surpassing $1 billion in just under eight months. At the core of this sudden rise is their innovative open-source LLM (large language model) known as the Yi-34B. On specific metrics, the Yi34B has outperformed some of Silicon Valley’s finest, like Meta Platforms Inc’s well-regarded Llama 2, according to Hugging Face’s evaluation. The asset of Yi-34B is that it’s not limited to a particular region or language; contrarily, it is available to developers worldwide in Chinese and English.

The company’s journey to unicorn status was bolstered by a recent funding round that included participation from Alibaba Group Holding Ltd’s cloud unit, which underlines the startup’s potential perceived by one of China’s tech giants and the Chinese broader interest in advancing AI technology. Kai-Fu Lee, the CEO of the prominent venture firm Sinovation Ventures, serves a role as the CEO of 01.AI as well. The significance of this event lies in its role in stimulating competition within the AI realm, where we are witnessing the rise of Chinese AI models.

The rapid growth of AI has been challenging since the escalating political tensions between the United States and China have complicated the landscape of AI development. Restrictions imposed by the American government on selling advanced AI semiconductors to China have been implemented, which could have far-reaching implications. To mitigate potential supply constraints, 01.AI proactively stockpiled the AI chips required for their operations. This approach demonstrates their commitment to ensuring the continuity of their AI development.

01.AI has assembled a team of over 100 individuals, comprising not only AI experts but also experienced business professionals. This combination of expertise enables the startup to tackle various aspects of AI development, from research to business strategy, since the startup intends to work with customers on developing proprietary AI models tailored for specific industries or competitive scenarios. The ability to offer AI systems in both English and Chinese provides a competitive edge for global companies, particularly in sectors such as banking, insurance, and trading. The startup also has plans to expand its language support in the future.

Kai-Fu Lee’s dedication to AI extends over many decades. He envisions AI as a transformative breakthrough for humanity and a means to better understand ourselves. His commitment to realizing this vision led him to step into the role of a startup CEO and to create an open-source model strategically aimed at making AI more accessible to developers. Just as open-source software allows for modification and enhancement, open-source AI systems provide a collaborative platform for innovation. In the dynamic landscape of AI development, 01. AI’s rapid ascent is a pledge to the potential and ambition within the global AI community. With its open-source approach the company has not only reached unicorn status but also laid the foundation for a promising future in AI.

What brings Yi-34B, and why does it matter?

These large language models (LLMs) are computer algorithms trained on extensive data to understand and generate human-like text, images, and code. Yi-34B has demonstrated superior performance in critical metrics compared to leading open-source models, including Meta Platforms Inc.’s Llama 2, as confirmed by evaluations from Hugging Face. Kai-Fu Lee, 01. AI’s CEO emphasized the goal of providing a superior alternative not only for China but also for the global market.

The AI model Yi-34B is named after its 34 billion parameters used in training, but 01.AI is already developing an even larger model with over 100 billion parameters. The startup made its model open-source to cater to a broader audience of AI developers who may not require or afford the largest, costliest models. The 34 billion parameter model was designed to run on more affordable computers, and the company plans to release a 6 billion parameter model to serve a broader range of developers. Kai-Fu Lee emphasized that their decisions are calculated and driven to offer valuable models to the AI community, rather than arbitrary choices. This open-source approach is akin to open-source software, allowing the modification and enhancement of the source code, which practically means that anyone can look under the bonnet of its algorithm and address its vulnerabilities.

Meanwhile, AI giants like OpenAI, Alphabet, Microsoft, and Meta are investing heavily in generative AI and chatbot development, as demonstrated by the recent unveiling of Elon Musk’s chatbot, Grok.