In under a year, Kai-Fu Lee’s new venture, 01.AI, has soared to unicorn status with its groundbreaking open-source AI model, surpassing Silicon Valley benchmarks. Spearheaded by Lee, the head of Sinovation Ventures, 01.AI achieved a $1 billion valuation, propelled by investment from Alibaba’s cloud division.
01.AI’s major achievement is Yi-34B, a superior large language model (LLM) available in Chinese and English for developers globally. These models, which simulate human-like text, images, and code, have placed Yi-34B ahead of competitors like Meta’s Llama 2 on key performance metrics. The introduction of ChatGPT by OpenAI catalyzed a generative AI race involving major players like Alphabet, Microsoft, and Meta, alongside Musk’s Grok.
Gold Standard
“Llama 2 has been the gold standard and a big contribution to the open-source community,” said Lee in an interview with Bloomberg over Zoom. “We want to provide a superior alternative not just for China but for the global market.”
“Our proprietary model will be benchmarked with GPT-4,” said Lee, about OpenAI’s LLM while also noting that providing a system that supports both English and Chinese languages will benefit international firms such as financial institutions, insurance companies, and trading firms. The startup intends to incorporate additional languages in the future.
In China, the AI landscape is vibrant yet segregated from the U.S. due to technology accessibility issues and political friction that has led to American restrictions on AI chip sales to China.
Lee described the circumstances as unfortunate, yet he confirmed that 01.AI has accumulated an adequate supply of chips for the upcoming period. The company proactively gathered these semiconductors earlier in the year, even taking out loans from Sinovation Ventures to finance the acquisitions.
“We basically bet the farm and overspent our original bank account,” he said. “We felt we had to do this.”
Lee — a veteran of Google, Microsoft, and Apple — is powered by a diverse team exceeding 100 people. The firm is expanding beyond its open-source origins, crafting specialized models, like an upcoming 100-billion parameter version, and planning multi-lingual expansions. The move aligns with Lee’s lifelong AI ambition, dating back to his 1982 graduate school application, and complements his authoritative books on AI’s future.
“It’s the biggest breakthrough for humanity,” said Lee. “It’s also the final step to understanding ourselves.”
Opportunity
Lee expressed his past skepticism about whether he would witness the fulfillment of technology’s potential within his lifetime. However, the advent of the latest LLMs changed his perspective, compelling him to shift from venture capitalism to leading a startup as its CEO. He saw this move as an opportunity to integrate his expertise as a computer scientist, author, and executive, harnessing the technology’s potential.
Regarding the newly launched AI system with 34 billion parameters, Lee mentioned that this size was strategically selected to ensure affordability, allowing the system to operate on less costly computers. The company is also aiming to reach a wider range of developers by offering a 6B model, which Lee believes is a deliberate move to serve real-world needs instead of just adding to the existing pool of models.
When questioned about managing his responsibilities as the CEO of both his venture firm and 01.AI, Lee explained his time allocation strategy. He justified that even with a 40-hour commitment to Sinovation Ventures each week, he has sufficient time left in the week to dedicate to 01.AI without neglecting his duties at Sinovation. He indicated that, by reserving six hours for sleep and other personal activities daily, he can maintain a balance between his dual roles.
Featured image: Bloomberg
