AI model on par with OpenAI’s o1 proves US’ chip restrictions fruitless: Expert
A new player in the global AI scene, DeepSeek, a Chinese tech start-up, is making waves in the US, presenting a low-cost AI model that rivals the performance of OpenAI’s o1. Despite facing strict semiconductor restrictions from the US government, DeepSeek’s breakthrough has raised alarm among US tech giants, prompting media outlets and industry experts to reconsider the effectiveness of the US’s chip export controls.
The breakthrough came on January 20, when DeepSeek released its highly anticipated large language model, DeepSeek-R1. According to reports from Nature, the new model has shown exceptional performance in several tasks, including chemistry, mathematics, and coding, equalling the capabilities of OpenAI’s o1. This development is generating both excitement and unease in Silicon Valley, as DeepSeek has managed to build an AI model that competes with those of US companies, but at a fraction of the cost and using less powerful chips.
A CNBC article expressed the growing concern in the US over DeepSeek’s success, noting that a previously unknown AI lab from China has ignited panic within the tech industry. The report highlighted that DeepSeek’s AI models are outperforming those developed by US companies, despite being built with lower-cost infrastructure. Similarly, a report by The New York Times pointed out that DeepSeek has managed to create a competitive AI chatbot with fewer high-end computer chips than those used by industry giants like Google and OpenAI. This development has cast doubt on the efficacy of the US’s chip export restrictions.
DeepSeek’s rise to prominence began in December 2024 when it released its large-scale AI model, DeepSeek-V3. This model quickly garnered attention from AI enthusiasts worldwide, generating discussions across social media and tech forums. The company’s success in leveraging open-source technology has further fueled its growth, creating a buzz that transcends borders.
However, DeepSeek has remained tight-lipped about the specifics of its technology. When approached for comment, the company informed the Global Times that it could not share additional details at the moment, urging media outlets to refer to its technical report for more information. DeepSeek has also stated that it does not engage in external project collaborations or offer privatized deployment services, instead focusing on research and development to produce even more advanced models.
DeepSeek’s open-source approach has been described as a game-changer for the AI industry. According to Li Baiyang, an associate professor at Nanjing University, DeepSeek’s ability to challenge US dominance in AI technology demonstrates the inefficacy of US chip restrictions. He noted that these restrictions, which target various aspects of AI development such as chips, talent, computing power, algorithms, and data, have failed to stifle China’s progress in the field. In fact, the restrictions have inadvertently hindered collaboration opportunities between Chinese and US companies, preventing mutual innovation.
Tian Feng, former dean of the Intelligence Industry Research Institute at SenseTime, pointed out that DeepSeek’s AI model is not only cheaper to develop but also performs on par with OpenAI’s o1 model. The cost to train DeepSeek’s V3 model is reportedly around $6 million, a mere fraction of the $60 million Meta spends on comparable models. This efficiency in both cost and performance has set a new standard for the AI industry, with some experts predicting that it could redefine the development rules for AI technology.
More Read
The implications of DeepSeek’s success are far-reaching. Jim Fan, Senior Research Manager at NVIDIA, noted on X that the company represents a new frontier in AI research, one that stays true to the original mission of OpenAI: open, frontier research that empowers everyone. Meanwhile, insiders at Meta have expressed concern, with reports indicating that engineers are scrambling to analyze and replicate DeepSeek’s success.
As the US struggles to maintain its AI leadership, DeepSeek’s model has raised questions about the future of AI innovation. The company’s approach has already attracted a global following, drawing users and developers from around the world. This could undermine the market share and influence of US companies, further intensifying the competitive pressure on American tech giants.
The competition between the US and China in AI is intensifying, with both countries vying for dominance in this transformative technology. While the US has invested heavily in AI infrastructure, including a $500 billion initiative announced by President Biden, China’s focus on innovation and open-source technology may position it as a central player in global AI development. As both nations continue to innovate, there is significant potential for collaboration, particularly in AI governance.
In conclusion, DeepSeek’s low-cost, high-performance AI model marks a significant milestone in the AI race, challenging US dominance in the field. Its success underscores the importance of open-source technology and cost-efficient development, offering a new pathway for AI innovation in the global landscape.
