Since Chinese AI company Deepseek released an open version of its R1 Reasoning Model earlier this week, many in the tech industry have been making grand claims about what the company has achieved and what it means for the state of AI .
Venture capitalist Marc Andreessen, for example, published that DeepSeek is “one of the most surprising and impressive discoveries I have ever seen.”
R1 apparently matches or beats Openai’s O1 model on some AI benchmarks. And the company says one of its models costs just $5.6 million to train, compared to the hundreds of millions of dollars shoulder companies pay to train theirs.
It also appears to have achieved that in the face of US sanctions that ban the sale of advanced chips to Chinese companies. MIT Technology Review writes that the company’s success illustrates how sanctions “drive startups like DeepSeek to innovate in ways that prioritize efficiency, resource pooling, and collaboration.” (On the other hand, the Wall Street Journal reports that Deepseek’s Liang Wenfeng recently told China’s premier that American export restrictions still represent a bottleneck.)
Curai CEO Neal Khosla offered a simpler explanation, claiming that the company is a “CCP state psyop” that is “misrepresenting how low the cost was to justify setting the price and hoping that everyone changes (for ) harm the competitiveness of artificial intelligence in the United States.” (A community note was attached to his post pointing out that Khosla offers no evidence for this and that his father Vinod is an Openi investor.)
Meanwhile, journalist Holger Zschaepitz suggested that DeepSeek could “pose the biggest threat to US stock markets”—if a Chinese company can build a cutting-edge model at low cost, without access to advanced chips, it will call into question.” the utility of Hundreds of billions of Capex have been poured into this sector.”
In response, Y Combinator CEO Garry Tan argued that Deepseek’s success would actually be good for its American competitors: “If training models become cheaper and easier, the demand for inference (real world use of AI) will grow and accelerate even faster Ensures that the computing provision will be used. “
And Meta’s chief AI scientist Yann Lecun discussed looking at Deepseek’s announcement through the lens of China versus the United States. Instead, he suggested that the real lesson is that “open source models are outperforming proprietary ones.”
“DeepSeek has profited from open, open source research (e.g. Pytorch and Llama by Meta),” Lecun wrote. “They invented new ideas and built them on top of the work of others. Since their work is published and open source, anyone can profit from it. “
All the debate seems to be pushing consumers to try the product — as of Sunday afternoon, Deepseek’s AI assistant is the top app in Apple’s App Store, just ahead of Chatgpt.
