Image: © local_doctor /Stock.adobe.com
DeepSeek’s success is a win for open source, says Meta VP and chief AI scientist Yann LeCun.
DeepSeek, a Chinese AI start-up, released its latest reasoning model last week, and now, the company’s AI chat assistant app has taken the top spots in the Apple App stores in both the UK and the US, overthrowing ChatGPT.
Released on 20 January, DeepSeek’s large language model R1 left Silicon Valley leaders in a flurry, especially as the start-up claimed that its model is leagues cheaper than its US competitors – taking only $5.6m to train – while performing on par with industry heavyweights like OpenAI’s GPT-4 and Anthropic’s Claude 3.5 Sonnet models.
However, in order to build its models, DeepSeek, which was founded in 2023 by Liang Wenfeng – who is also the founder of one of China’s top hedge funds, High-Flyer – needed to strategically adapt to the increasing constraints imposed by the US on its AI chip exports.
Earlier this month, the outgoing US administration capped the number of AI chips that could be exported from the US to most countries, while maintaining a block on exports to countries including China and Russia.
Although seen as a measure to ensure the US its leadership in AI innovation, the regulations have seemingly allowed China to reduce its reliance on American-made technology. Zihan Wang, a former DeepSeek employee, told MIT Technology Review that in order to create R1, DeepSeek had to rework its training process to reduce strain on the GPUs it uses – a variety specifically released by Nvidia for the Chinese market that caps its performance at half the speed of its top products.
Moreover, rather than weakening China’s AI capabilities, US sanctions on the country seem to be pushing start-ups like DeepSeek to innovate in newer ways that prioritise efficiency, resource-pooling, and collaboration, the publication further reported.
“DeepSeek’s surprising rise to the top of the Apple download charts in the United States, even under the weight of sanctions, poses an interesting question around the prevailing narrative of US dominance in artificial intelligence,” said John Clancy, the founder and CEO of Galvia AI.
“Could this be an indicator of over investment in the sector, and could the market be overestimating the long-term demand for chips? If so, valuations of companies like Nvidia might need a reality check.”
World-leading chipmaker Nvidia invested $1bn in AI companies in 2024, becoming a crucial backer of start-ups in the sector, including OpenAI, as more and more tech giants employ and offer an increasing number of AI products and services. The company, that has heavily invested in AI over recent years, reported a “record” revenue of $35.1bn for the latest financial quarter.
However, following R1’s release, Nvidia stocks have plummeted, falling down by more than 11pc today.
Although, Yann LeCun, Meta’s VP and chief AI scientist, said that DeepSeek’s capabilities should be seen as a win for open-source models, and not as a competition between US and China.
“DeepSeek has profited from open research and open source (eg PyTorch and Llama from Meta). They came up with new ideas and built them on top of other people’s work. Because their work is published and open source, everyone can profit from it. That is the power of open research and open source.”
Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.
