Photo: VCG

Editor's Note:
DeepSeek, once a little-known Chinese startup, has made headlines with the release of its powerful AI model, showcasing China's growing capabilities in frontier technology. What does this breakthrough reveal about China's AI innovation? How might it shape the future development of AI models? And what does it mean for the evolving technological dynamic between China and the US? Global Times (GT) reporter Wang Wenwen interviewed Karman Lucero (Lucero), an associate research scholar at Yale Law School who researches the governance of AI and AI policy in the US and China, to talk about these issues.


GT: The recent rise of DeepSeek has become a heated topic. What technological revolution will it bring? 

Lucero: One of the nice things about DeepSeek being open source is that they have published multiple papers, particularly one related to their recent advancements that articulate exactly what they did. It is about efficiency gains and the idea of achieving more with less. And that's a very important advancement in the AI space and one that other actors can utilize, whether they are open-source or closed-source developers. 

The second is more specific to the open-source community and highlights the importance of open-source models. It's about the ecosystem of the technology and the different institutions involved. It has definitely changed the conversation around open source in a good way. Going forward, national governments should not view open source solely in the context of national security. They must also consider the benefits to consumers both domestically and internationally. 

To be clear, national competitions over AI development will continue. However, it cannot be solely focused on national security. Focusing solely on national and other forms of security leads to a race to the bottom. In contrast, advancements in open source constitute more of a "race to the top" - where everyone can benefit not only from the technology but also from the competition. 


GT: Some believe that DeepSeek demonstrates the strength of China's homegrown innovation. In your opinion, what drives China's tech innovation?

Lucero: A couple of things. There is a pretty vibrant ecosystem of academics and experts who are able to engage in different ideas and try them out. This kind of debate and experimentation is important. Another reason is you have a lot of excitement and talent in China around AI development and using it for many different reasons. Third, you do have government investment in infrastructure and other means of supporting AI companies both directly and indirectly. For now, the balance of the regulatory environment allows for AI development, particularly for smaller companies.

Karman Lucero Photo: som.yale.edu

GT: How do you analyze the reaction of the US toward DeepSeek?

Lucero: I think the reaction is pretty varied. Some people in Silicon Valley and those who follow technology development, especially AI development, were already aware of DeepSeek and other open-source and closed-source models emerging from China. Some were less surprised, and some were even happy that there is a strong open-source developer sharing insights with the world. In Washington, reactions may have been a bit more surprised, but even within the policy world, there are different responses. The most extreme voices are calling for a ban, while other calls are for the US to invest more in open-source development and essentially compete better and smarter, rather than taking for granted that the US will always be ahead.

In China, recent developments in AI have reinvigorated the excitement and energy people devote to developing AI and imagining how AI can shape our future. 


GT: Some say that the focus of China-US competition in the future will be technology. You once said that the AI development of China and the US is like two galaxies in the universe. Is a clash inevitable or can it be avoided?

Lucero: It depends on the approach that both countries take. If they support open-source development, for example, then there will be a lot more integration and knowledge-sharing across the two ecosystems. If both countries allow for dialogue to continue on AI governance, there will still be a sense of competition, but it will be the sort that benefits both sides. Again, this is the difference between a "race to the top" sort of competition that benefits everyone and a "race to the bottom" sort of competition that decreases opportunities for everyone.

Ultimately, it depends on both countries having more confidence in their own approaches, focusing on improving their own approaches rather than prioritizing attacks on each other and supporting risk taking.

There are potentially many areas where the US and China can cooperate in the tech sector, but the issue is the political will. It's up to both sides to be politically creative and willing to work together. The main point of the galaxy analogy was to demonstrate that AI in the US and China is not just one simple thing but rather two complex ecosystems with many dynamics. I am not an astrophysicist, but my understanding is that two galaxies interacting don't need to cause wanton destruction. Instead, they can combine to make the night sky brighter for everyone.  


GT: You attended the AI Summit in Paris recently. What is your takeaway from this summit regarding AI development?

Lucero: There were essentially two big conversations going on at the summit. One focused on a more traditional view of AI safety, trying to convince developers and countries to pay more attention to the associated risks, even if that means slowing down or stopping the development of AI. 

The other conversation was led by those who believe that recent developments in AI including advancements by OpenAI and DeepSeek are positive. While discussing risks and regulations are important, there are also reasons to celebrate and support the development and success of AI. 

Whereas previous summits in the UK and South Korea focused more specifically on safety and risk, this summit saw a set of conversations focused on reducing regulation. While cutting back on regulation and focusing on safety in AI are not necessarily inherently contradictory (more effective regulation can achieve both), the two conversations are currently talking past each other. 


GT: Is AI governance an area where China and the US could collaborate?

Lucero: There's a growing number of challenges for cooperation, but communication is very important. There should be communication channels at three different levels. The first is an official channel between Washington and Beijing. While it is clear that both countries are serious about developing AI, they should demonstrate that they are both serious about supporting dialogue focused on the future of AI as well. The second is people-to-people and academic exchanges. The third is for other kinds of expert exchanges and discussions. AI is such a big topic, it's too big for just government actors to decide everything. You need interactions between the two societies as well. It would be valuable to continue to allow for those kinds of exchanges. 

The present day is of historic importance in multiple ways. One is the advent of AI and other transformative technologies. Another is the extent to which individuals and societies in different parts of the world can interact with each other, in ways that were, if not impossible, at least very difficult until recently.