DeepSeek, a startup founded just two years ago is being credited with a ‘Sputnik moment’ in the global AI development race by training an AI model on a par with OpenAI’s ChatGPT with less than $6m.
Founded in 2023 the startup is the passion project of millennial hedge fund entrepreneur Liang Wenfeng, who decided to focus on his obsession with AI and became successful by spotting patterns in stock prices using the technology.
Initially designed as a purely research-focused organisation with no profit drive, DeepSeek is bankrolled by Liang’s hedge fund, which reportedly has assets under management surpassing £11bn.
DeepSeek’s success in developing a powerful model with relatively modest financial resources may prompt further tightening of US controls, or undermine the idea that they can work effectively to halt China’s AI advancements.
The company’s preference for hiring local talent from Hangzhou, China, could also be helping to drive its success.
Some investors and analysts have expressed scepticism about DeepSeek’s claims, however, while some have suggested that the company has stockpiled advanced chips that it has not disclosed publicly because of US sanctions.
Regardless of doubts regarding DeepSeek’s claims, the fact that a startup has achieved this level of AI development competition with companies such as Microsoft, which plans to invest around $80bn in AI infrastructure this year, will surely cause some concern.
DeepSeek’s success may, in part, result from the company innovating because of, rather than in spite of, constraints.
The company is also known to pay well for top talent, poaching developers with job offers from bigger companies such as Nvidia.
There is much speculation right now over what could help China’s AI development progress in the face of US sanctions.