How smart is AI these days? Not smart enough, it seems, to spot a Chinese competitor coming apparently from nowhere.
AI assistant DeepSeek was developed in China by hedge fund entrepreneur Liang Wengfeng as a side project, intended, he says, only for research.
Its latest iteration, which appeared on Apple’s app store and Google Play a week ago, has demonstrated capabilities apparently equal to its better known, and apparently far more costly rivals, led by ChatGPT, owned by OpenAI.
In a research paper published last year, DeepSeek said it had been developed using only “limited capability” Nvidia chips – the most advanced technology has been banned from China under export controls since 2022 – and had cost a mere $5.6m (£4.5m) to develop.
Its success appears to pose a fundamental challenge to the received wisdom that the development of AI will require huge investment, vast computing power housed in energy-hungry data centres, and be a race run and won in America.
Last week Donald Trump heralded a £500bn investment project, including $100bn from OpenAI. Google, Meta, Microsoft and the ubiquitous Elon Musk are all in the race, desperate to be first to find the holy grail of artificial general intelligence – a theoretical concept that describes a machine’s ability to learn and understand any intellectual task that a human can.
That now looks like a contest not just between Californian tech bros with macho budgets, but between superpowers.
The implications were clear on US stock markets. The tech-heavy Nasdaq index was down 3%, a loss of a trillion dollars, and Nvidia itself fell almost 17%, shedding $600bn (£482bn) from its market value and losing its status as the world’s biggest company in the process. Google’s parent company fell in value by £100bn and Microsoft, which has a stake in the privately held OpenAI, was down £7bn.
As well as questions about…
