My Take | Chinese start-up DeepSeek looks set to rewrite the rules of the AI game
It could be true that DeepSeek is standing on the shoulders of others, but it has still shown an alternative to the lavish spending on AI
China’s hottest artificial intelligence (AI) start-up has demonstrated an ability to create AI models at a fraction of the cost compared with leading industry players such as OpenAI. Its newly released reasoning model DeepSeek-R1 is “rivalling OpenAI’s o1”, but was developed at a far lower cost.
The breakthrough by DeepSeek has stirred heated debate. For those who believe in DeepSeek, the firm’s success has significant implications for China’s rivalry with the US in AI, and even the future of AI. When Sam Altman’s team unveiled ChatGPT in November 2022, China was caught by surprise. Chinese tech companies scrambled to launch their own LLMs in a chaotic situation known as the “battle of over 100 models”, but the exuberance has been dampened by two dark clouds.
The first potential repercussion is US export restrictions on advanced graphic processing units (GPUs), particularly Nvidia’s much-sought after chips – a restriction that is depriving China of the most powerful weapons in the AI arena.
The second implication is the limited financial resources of Chinese Big Tech firms, compared to their US peers. Under the scaling law, which literally means “the more investment in advanced chips, the better the performance of models”, Chinese AI companies were put in a disadvantageous position.