deep seek
Chinese tech startup DeepSeek is shaking up the U.S. AI sector with its open-source approach and low-cost model, showing how the Chinese company has developed competitive AI systems to match those developed The question of how it was created has attracted curiosity from the US media and AI forums, and has been widely reported. Major U.S. high-tech companies are developing products one after another while overcoming the strict semiconductor regulations imposed on China by the U.S. government. The Global Times on Saturday spoke to the company and several AI industry observers to explain the phenomenon behind it.
“China’s cheap open AI model DeepSeek has scientists excited,” Nature wrote three days after a large-scale language model called DeepSeek-R1 was released by the Chinese company on January 20. Its performance on certain tasks in chemistry, mathematics, and coding is on par with that of DeepSeek-R1, according to the Nature report, which is OpenAI’s o1.
A CNBC report on Friday highlighted concerns about “How China’s new AI model DeepSeek threatens U.S. dominance” and the impact of DeepSeek on U.S. dominance in AI. The paper said, “After a little-known AI research institute in China announced an AI model that outperformed the best in the United States, despite using cheaper and less powerful chips, the entire Silicon Valley It caused panic,” he wrote.
The New York Times report, in a report entitled “How China’s AI Startups Are Tackling,” said, “The company has fewer high-end computer chips than U.S. giants like Google and OpenAI. and developed a cheap and competitive chatbot. This shows the limits of chip export controls.” DeepSeek competes with Silicon Valley giants. ”
The Hangzhou-based startup gained attention after announcing its large-scale AI model DeepSeek-V3 in December last year. This new model quickly gained traction among AI enthusiasts, sparking cross-border discussions and spreading to international social media platforms and technical forums.
At the time of writing, there had been no official response from the team, who were widely seen as a “mystery team” who rarely spoke to the media and never overtly promoted themselves.
When asked on Saturday for comment on the external attention surrounding DeepSeek’s development of free, open source, large-scale language models, DeepSeek told the Global Times that it could not provide additional information at this time. They recommended viewing the technical report for a deeper understanding of the model.
A notice published by DeepSeek in an open communications group said the company “does not collaborate on external projects and does not provide privatization deployments or related support services.” Furthermore, DeepSeek emphasized that it will focus on research and development to develop more advanced models, and urged people to “look forward to it.”
“Deep Seek’s technological approach challenges the US dominance and monopoly in AI technology and proves that US chip regulations have been ineffective,” said Lee, associate professor of intelligence research at Nanjing University. – Bayan told the Global Times on Saturday that Deep Seek explained the reason. It attracted a lot of attention not only from China but also from Silicon Valley in the United States.
The United States is currently focused on protecting its advantage in computational and algorithmic capabilities, and is seeking to maintain its dominance in artificial intelligence through various policies and administrative regulations, experts said.
Tian Feng, former director of the Intelligence Industry Research Institute at Chinese AI software giant SenseTime, praised DeepSeek’s ability to achieve impressive production results at lower training costs, in addition to its fully open-source approach to the technology and underlying model. “There is a possibility that the following movement will occur.” Redefine AI development rules. ”
This represents similar benefits to OpenAI’s o1 model, with the V3 model costing only about $6 million to train, about one-tenth of what Meta would invest in a comparable AI model. Tian told the Global Times on Saturday.
NVIDIA Senior Research Manager Jim Huang told “I live within the line.”
Media outlet The Decoder reported that Meta’s AI division is feeling the pressure, according to anonymous posts on Teamblind, a forum for verified Big Tech employees. In the same post, Meta wrote, “We’re in a panic. Engineers are working frantically to dissect DeepSeek and copy everything they can from it.”
Meta’s Mark Zuckerberg recently outlined his company’s response on Facebook, saying that by 2025, Meta aims to develop an AI assistant that can serve more than 1 billion people. “This will be a defining year for AI,” Zuckerberg wrote, according to a report in The Decoder.
The approach taken by Chinese technology startups is putting unprecedented pressure on American AI companies. This could draw more users and developers around the world to participate in high-end AI development, potentially weakening U.S. AI market share and influence. According to Tian,
The US’s strict measures are wary of China in the AI field, covering various aspects such as chips, human resources, computing power, algorithms, data, etc., and it can be said that “no stone will be spared in sanctions”. Li Baiyang said. .
However, the state of China’s AI technology innovation has proven that its impact is minimal, and if anything, such limitations have prevented many US companies from collaborating and developing opportunities with China’s industrial technology innovation. Li Baiyang said he was missing out.
The New York Times, as mentioned above, left the question of China-US competition in AI open-ended. “They argue that if the best open source technologies come from China, U.S. developers will build systems based on those technologies. In the long term, China will emerge as the world’s center of AI innovation. It is possible,” he wrote.
Li Baiyang said that DeepSeek’s open source technology root application embodies China’s inclusive and accessible mindset, which is of great importance in strengthening China’s AI industry ecosystem. .
“As long as China continues to innovate, it can make great progress by pursuing its own path,” Tian said.
US President Donald Trump recently announced $500 billion in AI infrastructure investment in the US. Three top technology companies will form a new company called Stargate to grow artificial intelligence infrastructure in the United States, according to CNN.
Healthy competition between China and the United States in the AI field depends on America’s attitude. Li Baiyang added that as the two largest countries in the world’s AI field, China and the United States compete in the AI industry, but there is also great scope for cooperation, especially in AI governance.
