After DeepSeek released the open source version of its R1 inference model earlier this week, many American tech leaders praised the startup’s achievements. On January 27, DeepSeek’s app surpassed ChatGPT on the App Store’s top-rated free app chart.

deepseek shutterstock
DeepSeek is a small startup from Hangzhou, China. Photo: Shutterstock

Venture capitalist Marc Andreessen called DeepSeek “one of the most impressive and surprising breakthroughs I've ever seen.”

R1 appears to keep up with or beat OpenAI’s o1 model on certain AI metrics. The $5.6 million training cost is also much smaller than the hundreds of millions of dollars spent by companies like OpenAI.

DeepSeek's success illustrates how bans are "pushing startups like DeepSeek to innovate in ways that prioritize efficiency, pooling resources, and collaboration," MIT Technology Review notes.

Meanwhile, Curai CEO Neal Khosla said that DeepSeek is faking training costs to “justify low prices” and hope people switch to using it, thereby affecting competition in the AI market.

Journalist Holger Zschaepitz suggests DeepSeek “represents the biggest risk to the US stock market.” If a Chinese company can build advanced models at low cost without access to advanced chips, it raises questions about why hundreds of billions of dollars are invested in the industry.

In early January, at the Asian Finance Conference, computer scientist Kai-Fu Lee commented that the US has an advantage in research and innovation, but China is better at engineering. “In this era, if you have limited computing resources and money, you learn how to make things efficiently.”

In response, Y Combinator CEO Garry Tan argued that DeepSeek’s success could be good for its US rivals. “If models become cheaper and easier to use, the demand for inference (using AI in real-world applications) will grow and accelerate even more.”

Yann LeCun, chief AI scientist at Meta, said DeepSeek’s announcement shouldn’t be viewed through the lens of US-China competition. Instead, he suggested the takeaway is that “open source models are surpassing proprietary models.”

“DeepSeek benefits from open research and open source (e.g. PyTorch and Llama from Meta). They come up with new ideas and build on top of others’ work. Since their work is public and open source, everyone can benefit from it,” he wrote on LinkedIn.

(According to TechCrunch, Bloomberg)