DeepSeek The release of its newest open-source models has marked China’s “ChatGPT moment” by sparking a craze among enthusiasts. artificial intelligence (AI) Adopting this approach raises doubts regarding the feasibility of creating pre-trained models internally, as stated by Lee Kai-fu, the founder and CEO of the Chinese startup 01.AI.
During an interview with the South China Morning Post on Friday, Lee, who previously led Google China, mentioned that 01.AI is using DeepSeek’s widespread appeal to offer AI solutions to business clients, notably within the finance, video gaming, and legal industries.
Lee stated that it was crucial for them to adopt [DeepSeek] as their main choice. He noted this became evident following the significant increase in demand from Chinese CEOs for the company’s models post-late January, comparing it to something obvious written on a wall.
Are you looking for insights into the most significant issues and developments globally? Find your answers here. SCMP Knowledge Our latest platform features handpicked content including explainers, FAQs, analyses, and infographics, all provided by our prestigious team.
Established in mid-2023 , 01.AI stands as one of China’s leading unicorn companies focused on developing large language models (LLMs). The company started shifting from being an LLM developer to becoming a solutions provider last year; however, this transformation gained significant momentum due to what Lee referred to as the "godsend" opportunity offered by DeepSeek. Since then, numerous Chinese hardware and software providers have aligned their offerings with those based on DeepSeek's models.
Lee mentioned that AI's expertise in large language model (LLM) development sets it apart, considering that fewer than ten Chinese firms have comparable skills. The company’s focus on DeepSeek further differentiates Lee’s initiative from rivals who prefer their own proprietary models.
Many hardware providers easily provide on-site plug-and-play options equipped with extensive business capabilities; however, 01.AI distinguishes itself through its capability to customize and optimize DeepSeek models specifically for each client, as mentioned by Lee.
"He mentioned, 'How do you train it, how do you optimize it, how do you implement reinforcement learning, and how do you achieve rapid inference? The final aspect can only be accomplished by firms with large language model expertise,'" he stated.
Although AI, with approximately 200 staff members, has started manufacturing its own servers, Lee mentioned that the company is not entering the server marketplace. "all-in-one" boxes , integrating both AI-specific hardware and software. This technological approach is advocated by prominent Chinese corporations such as Lenovo Group, Huawei Technologies, and Inspur Group.
Lee, who is 63 years old, anticipated that 01.AI would generate 100 million yuan (approximately US$13 million) in revenue for the first quarter, which is equivalent to its total revenue for the entire year of 2024. Nevertheless, as he mentioned, the firm was still not profitable.
The widespread adoption of DeepSeek's open-source models has significantly influenced the AI sector globally, inspiring numerous Chinese AI startups to take action. reassess their strategies Lee mentioned that it was difficult for startups to "validate" the necessity of expensive pre-trained models considering open-source alternatives like DeepSeek .
"How will you rationalize your investments in graphics processing units, and what is the trajectory of your business revenue growth? Additionally, are there any prospects for achieving profitability at some point? These are the issues that the 'six tigers' are facing," Lee commented, alluding to them. China's top generative AI startups Zhipu AI, Moonshot AI, MiniMax, Baichuan, Stepfun, and 01.AI.
Lee stated that a pre-trained model can only be validated with hundreds of millions of users. Therefore, companies like Alibaba, Google, DeepSeek, and ByteDance can justify their use, whereas the majority of others cannot. It should be noted that Alibaba owns the Post.
Lee, an author with multiple publications on artificial intelligence, raised doubts about the "benefit" offered by OpenAI amidst the rise of DeepSeek. He mentioned, "[OpenAI CEO] Sam Altman’s worst fear must be having a rival that offers services at no cost." Lee also noted that he has encountered numerous individuals who have discontinued their ChatGPT memberships due to the availability of the free alternative from DeepSeek.
Recently, OpenAI advised the U.S. government to take action. ban models from DeepSeek , referring to the Chinese startup as "state-controlled." The U.S. company's efforts to lobby against DeepSeek indicated paranoia, according to Lee.
He mentioned that they are observing their carefully constructed house of cards beginning to crumble since someone else created an equivalent structure at no cost.
More Articles from SCMP
Missiles ought not to damage enhanced relations
Trials will guarantee that the low-altitude economy can reach great heights in Hong Kong.
Who is US Vice-President J.D. Vance’s relative, Nate? The ex-US Marine, who battled in Ukraine, gained attention when he condemned the vice-president for how he handled Volodymyr Zelensky during his time at the White House.
China’s sixth-generation stealth aircraft; ultranationalist blogger penalized: SCMP news roundup
The article initially appeared on the South China Morning Post (www.scmp.com), which is the premier source for news coverage of China and Asia.
Copyright © 2025. South China Morning Post Publishers Ltd. All rights reserved.
No comments:
Post a Comment