The launch of China’s DeepSeek, a cost-effective and efficient artificial intelligence (AI) model, was a “Sputnik moment” for the United States, akin to the country’s shock when the Soviet Union launched the first-ever artificial Earth satellite in 1957. The DeepSeek news triggered a plunge in the respective stock prices of Nvidia and Constellation Energy.

DeepSeek’s model raised questions about US dominance in AI development. In response, OpenAI alleged that DeepSeek used its technology to train the latter’s own budget model. However, it would be more constructive for the US AI industry to learn from DeepSeek to develop AI applications that are affordable, accessible, and sustainable.

With a budget of less than $6 million, DeepSeek managed to develop an AI model V3, rivaling OpenAI’s GPT-4 and Meta’s Llama 3.1, which invested significantly more resources. This feat was achieved through frugal innovation, leveraging older Nvidia H800 chips and optimizing software to reduce costs and energy consumption.

The success of DeepSeek’s V3 model demonstrated that it is possible to develop high-performing AI models without the need for exorbitant budgets or cutting-edge hardware. This development prompted OpenAI to suspect that DeepSeek might have used some of its data, leading to an independent evaluation of DeepSeek’s performance by Microsoft.

While it is difficult to prove whether DeepSeek used a “distillation” of OpenAI’s GPT models to train the open-source V3 models at a fraction of the cost of what Western tech giants are spending on their own models, DeepSeek’s successful development is good for the whole world for several reasons.

First, it shows that AI models can be developed in a way that is much cheaper, more accessible, and more sustainable. This democratizes AI development, making it possible for more countries and companies to participate in and benefit from AI advancements.

Second, DeepSeek’s success demonstrates that developing countries with limited resources can learn from its approach to develop their own innovative AI models. This can help bridge the gap between developed and developing countries in terms of technological capabilities.

Third, DeepSeek, along with other Chinese AI developments such as Alibaba’s Qwen and Huawei’s AI models, are now seen as major competitors to US AI developments. This competition is healthy for the AI industry as it reduces concerns over a monopoly and forces US companies to consider adopting open-source models and fine-tuning their learning models to become more efficient. The presence of multiple strong players in the AI field ensures that innovation continues to thrive and that the benefits of AI are more widely distributed.

It is healthy to have multiple winners in the AI development race as it ensures that more countries and people can participate in and benefit from AI advancements. Embracing AI models that are affordable, accessible, and sustainable is crucial for ensuring that AI benefits the world and not just a few elites. However, the world needs to come together to develop safeguards to ensure that AI does more good than harm

The Biden administration responded to the rapid advancements in Chinese AI development by imposing export bans on advanced chips and equipment to impede AI development in China. By restricting access to advanced chips like Nvidia’s H100, the US aimed to slow down China’s AI progress. However, this approach had a limited effect. Instead of halting progress, it prompted Chinese companies to develop workarounds using available resources. For example, DeepSeek reportedly used “distillation” and other techniques to train their AI models with less computational power. This has led to the development of smarter software that can achieve high performance even with less advanced hardware.

The export control ban inadvertently encouraged China to innovate even faster by finding more efficient ways to develop AI. This has positioned China as a formidable competitor in the AI race, with companies like DeepSeek leading the way in frugal innovation. As a result, the US needs to rethink its development strategy by understanding and learning from China’s approach.

Instead of relying solely on wealth and talent, the US should consider how to leverage the strengths of others to enhance its own capabilities.

The AI development race is not a sprint but a marathon, much like the development of the automobile industry. Different countries have led the way in different aspects of automobile development over the years. Germany was known for producing the first automobile in 1885, the US for mass production in the 1950s, Japan for superior quality in the 1980s, and now China for its dominance in electric vehicles.

Similarly, the AI race will have different winners in different rounds. The US was ahead after the debut of OpenAI’s ChatGPT, but now, with DeepSeek’s V3 model showing comparable performance, China is catching up quickly in the second round. As countries learn from each other’s advancements, new winners will naturally emerge in future rounds.

It is healthy to have multiple winners in the AI development race as it ensures that more countries and people can participate in and benefit from AI advancements. Embracing AI models that are affordable, accessible, and sustainable is crucial for ensuring that AI benefits the world and not just a few elites. However, the world needs to come together to develop safeguards to ensure that AI does more good than harm.

Instead of accusing each other of wrongdoing, it is important to acknowledge what each country has done right and learn from their successes.

The AI development race is akin to judo, where an opponent’s weight, power, and momentum are used against them to gain an advantage. Competition in developing AI should be viewed as a multiround judo contest, in which countries leverage each other’s strengths to build a better world.

By working together and learning from each other, we can ensure that AI development benefits everyone and leads to a more equitable and prosperous future.

The author is a distinguished professor at the UCLA Anderson School of Management.

The views do not necessarily reflect those of China Daily.