Since Chinese AI companies Deepseek released the open version of the reasoning model R1 earlier this week, many people in the high -tech industry have achieved what they have achieved and what they mean to the current status of AI. We have made a big presentation.
For example, Venture Capitalist Mark and Leisen posted DeepSeek as “one of the most surprising and impressive progress I have ever seen.”
R1 seems to be in line with O1 model of OPENAI in a specific AI benchmark. The company claims that one of its models is only $ 5.6 million in training, compared to a major US company paying hundreds of millions of dollars to model training.
It seems to have achieved it despite the US sanctions that prohibit the sale of advanced chips to Chinese companies. Mit Technology Review states that the success of the company indicates how sanctions are “encouraging a startup company like DeepSeek to innovate in terms of efficiency, resource pools, and collaboration.” 。 (Meanwhile, Wall Street Journal reported that Deep Shik of Deep Seek recently told China that US export restrictions were still a bottleneck.)
Curai CEO Neal Khosla gives a simpler explanation, and the company is “falsifying that the cost was low to justify the price, and we expect everyone to transfer to it.” He claimed that it was a psychological operation. [to] AI in the United States may impair the competitiveness. (The community notes are attached to his post, Khosla has not shown evidence for this, and pointed out that his father, Vinod, is an investor in Openai.)
Meanwhile, journalist Holger Shahpitz suggested that Deep Seek could be the biggest threat to the US stock market. If Chinese companies can build cutting -edge models at low costs without using advanced chips, their usefulness will be questionable. Hundreds of billions of dollar capital investment has been poured into the industry. “
GARRY TAN of Y Combinator CEO argued that DeepSeek's success would be really good for US competitors. Guarantee that computing supply is used. “
Meta's chief AI scientist, Yann Lecun, opposed the Deepseek's presentation through a lens, China and the United States. Rather, he suggested that the real lesson was that “open source models are being exceeded the prolipric model.”
“DeepSeek has gained profits from open research and open source (eg, Meta's Pytorch and LLAMA),” Lecun wrote. “They came up with a new idea and built it on other people's works. Their works are open and open source, so everyone can make a profit.”
All of these discussions seem to encourage consumers to try this product. As of Sunday afternoon, DeepSeek's AI Assistant has become the top free app with Chatgpt suppressing Chatgpt in the Apple App Store.
