Chinese start-up DeepSeek challenges US’ leading position in artificial intelligence
The AI race has a new contender: the recently launched DeepSeek V3. The Chinese start-up’s model is making waves with its open source language model and impressive performance, already beating OpenAI’s GPT-4o and Meta’s Llama on various benchmarks.
What sets the Chinese start-up apart is that it developed its model with a fraction of the computing power available to the world’s leading AI companies. Trained with a modest budget of $6 million using Nvidia H800 GPUs, DeepSeek V3 is designed for commercial use and adheres to Chinese regulations, though it carries noticeable political biases.
DeepSeek V3 can code, translate, and write essays and e-mails, starting from a descriptive prompt. The start-up also recently unveiled DeepSeek-R1, a reasoning model with remarkable performance in specific tasks, further positioned as a cost-effective alternative for enterprises. Despite China’s AI chip restrictions, the start-up’s advancements highlight significant progress in AI, inviting businesses to rethink their AI strategies.
DeepSeek’s emergence didn’t go unnoticed, given its ability to create a powerful model with a small budget. According to AI researcher Andrej Karpathy, this level of capability typically demands clusters of around 16,000 GPUs, while the Chinese company declared to have used 2,000 of such chips.
Despite the export restrictions, DeepSeek shows China’s ability to create a serious contender in the AI race by leveraging open-source development. The New York Times underlined how this model’s performance “raises questions about the unintended consequences of the American government’s trade restrictions.” Chinese researchers simply circumvented restrictions using available tools.
Ion Stoica, a professor of computer science at the University of California, Berkeley, told the NYT how the progressive movment of the open source community toward China could represent a huge risk for the US. This way, the Asian superpower can speed up the development of new technologies.
In his last days as president, Joe Biden issued new rules to prevent China from acquiring the most advanced AI chips from other countries. It is unclear whether newly elected President Donald Trump will maintain these rules.
Despite the export restrictions, DeepSeek shows China’s ability to create a serious contender in the AI race by leveraging open source development.
Last week, the Trump Administration announced the Stargate Project a $500 billion partnership between OpenAI, Oracle, Japan’s Softbank, and Emirati sovereign wealth fund MGX to build AI infrastructure and lead to the creation of 100,000 jobs.
The emergence of DeepSeek as a formidable competitor in the AI landscape signals a shifting dynamic in global AI development. While the model demonstrates impressive technical capabilities, including self-checking outputs that enhance reliability in physics, science, and math, enterprises must carefully evaluate the geopolitical implications of adopting these technologies. The model’s MIT license allows for unrestricted commercial use, but as industry experts suggest, organisations need to carefully assess both the technical benefits and potential risks associated with implementing Chinese AI technology in global operations.
Innovation Origins
Subscribers 0
Fans 0
Followers 0
Followers