They will clam up when asked about Tiananmen Square but they can land the US’s artificial intelligence empire in DeepSeek, if you know what we mean!
Chinese AI startup DeepSeek’s R1 model is not the only one that will cause headache to US tech giants and test the merit of investing billions of dollars in artificial intelligence projects.
According to a white paper released last year by the China Academy of Information and Communications Technology, the number of AI large language models worldwide has reached 1,328, with 36 per cent originating in China. It makes China the second-largest contributor to AI, behind only the US.
Top names in China’s tech sector, from Alibaba Group Holding and Baidu to Tencent Holdings, have invested significant money into the race to acquire hardware and customers for their AI ventures. For example, Kai-Fu Lee’s 01.AI startup stands out with its open-source approach, like DeepSeek.
This month, MiniMax, an Alibaba- and Tencent-backed startup that is valued at more than $2.5 billion, debuted three new models.
The company claims that its model MiniMax-Text-01 performs better than models such as Google’s recently unveiled Gemini 2.0 Flash on benchmarks like MMLU and SimpleQA or the ability of a model to answer math problems and fact-based questions. MiniMax-VL-01 reportedly rivals Anthropic’s Claude 3.5 Sonnet on evaluations that require multimodal understanding.
In the US, most AI tech companies are promising big investments. For example, Meta last week said it would spend around $65 billion this year on AI development. OpenAI has joined a group of other firms that have pledged to invest $500 billion in building AI infrastructure in the US over a few years.
Chinese AI startups are valued at a fraction of US companies like OpenAI (recently valued at $157 billion) as aggressive competition is leading to a price war among AI model vendors. Beijing-based Zhipu AI, for example, valued at around $3 billion in its latest fundraising round in December, has already showcased its AI agent in late November and released a video-generating model similar to OpenAI’s Sora in July.
DeepSeek’s R1 debut has attracted praise from OpenAI CEO Sam Altman, who has called it an “impressive model, particularly around what they’re able to deliver for the price”.
He said OpenAI would accelerate the release of some upcoming products. US President Donald Trump said the arrival of DeepSeek should serve as a “wake-up call for our industries that we need to be laser-focussed on competing to win”.
Among the major drawbacks of DeepSeek’s R1 model — like with all other Chinese AI models — is self-censorship on topics considered sensitive in China.
The Telegraph asked on the DeepSeek app questions around the 1989 Tiananmen Square protests but we were informed that such queries were “beyond my current scope. Let’s talk about something else.”
DeepSeek’s Cloud infrastructure is being stretched to its limits because of the influx of new users and it’s taking time for new registrations to take place.
The company reportedly spent just $5.6 million training its newest AI model, compared with the hundreds of millions US companies spend on their AI technologies.
Liang Wenfeng, who was born in Guangdong in 1985, founded DeepSeek with 10 million yuan ($1.4 million) in registered capital, according to company database Tianyancha. In an interview with Chinese online media outlet 36Kr in May 2023, Liang said most developers at DeepSeek were either fresh graduates or those early in their AI career. Popular investor Marc Andreessen is calling it “AI’s Sputnik moment”.
The DeepSeek mobile app is ranked number one in iPhone app stores in Australia, Canada, China, Singapore, the US and the UK, according to data from market tracker App Figures.
