More Investment, More Competition, Less Panic
Editor’s note: In February, The National Interest organized a symposium on the U.S.-China technology race amidst the emergence of DeepSeek and ongoing legal battles over TikTok. We asked a variety of experts the following question: “What are the three most important technology policies that the U.S. should pursue or avoid to compete adequately with China?“ The following article is one of their responses.
Collective DeepSeek Panic
The foundation of the information revolution, the web revolution, and now the AI revolution comes from research done in the United States that benefited from vast amounts of funding. This came both in basic science at the National Science Foundation as well as applied science at the Defense Advanced Research Projects Agency. And so it seems a little absurd that we have spun ourselves into a collective panic over the ‘DeepSeek moment.’
There are two parts to this panic. The first part was articulated best by Kat Duffy at the Council on Foreign Relations. She asked ‘If India, not China, had produced DeepSeek, how would the coverage and the analysis of it shift?’ At least part of the reaction to DeepSeek is coming out of a fear of China as a geopolitical and economic rival, rather than about the technology itself. It’s telling that while some are calling this the Sputnik moment for tech, a more apt analogy is what others are calling a Toyota moment, calling back to the 80s, and Japan’s challenge to the US auto industry (a challenge that I will note benefited the US consumer greatly).
The second part of the panic is I think rooted in a misunderstanding of the root causes of America’s technological dominance. And perhaps is maybe drinking too much of our own Silicon Valley Kool-Aid. I read this part of the panic as “OMG someone else did something smart that we weren’t expecting, and maybe we aren’t as smart as we thought.”
There’s a lot we don’t yet know about DeepSeek’s system, especially about their claims of limited compute time and GPU usage. But they’ve been surprisingly open about their development process, and people who track these systems were not entirely surprised by the release of DeepSeek-R1. They were (perhaps out of necessity) able to innovate using clever engineering and ideas that were out in the open but hadn’t been put together in the way they did, and they built an impressive development stack that allowed them to produce Deep-Seek R1. They did all of this off the backs of open(-ish) systems like LLama and Qwen. And that’s the point we need to pay close attention to.
Research, open science, and transparency have been the rocket fuel powering the engine of Silicon Valley for decades. The business of AI may be driven by Silicon Valley, but the ideas, the creative tricks, and the foundational mathematics and computer science, all came through a process of open science and research that fueled innovation. Research happened at universities and at research labs, and it was shared widely so that anyone could at least hope to reproduce what others had done. The AI marvels we see today came from those core developments years, and sometimes decades, ago.
The pace of progress in AI is quite breathtaking right now. We’ve moved from the original LLMs to advances in multimodal systems to systems that can exhibit reasoning-like behavior to ones that can converse in multiple languages. We see new models emerging almost every day, many of them built on open stacks, and hosted on open platforms like Hugging Face and evaluated in open settings like Chatbot Arena. The most important policy the US government can pursue right now is to foster and develop further this open and vibrant research environment for innovation AI. Here’s how.
The US government should PURSUE expansion of research funding and infrastructure
One of the key bottlenecks that’s limiting further exploration of AI systems is access to costly hardware infrastructure – GPUs. The National AI Research Resource is a good start, but it needs to be made permanent beyond a temporary experiment, and the CREATE AI Act is a good way to make this happen. But even beyond hardware, what we need is more well-supported public infrastructure for AI, of the kind that Hugging Face (a Franco-American company) represents, and that groups like the Mozilla Foundation have advocated for. And above all, we need more funding for research in AI and related disciplines, especially those that seek to understand the way AI deployments will change our basic societal structures. This expansion will keep US AI research ahead of other countries, just like it has all these years.
The US Government should PURSUE actions that encourage competitiveness of the US tech industry
We haven’t had real innovation in search for a very long time. And the few social networks that remain in the US are tired, or toxic, or both. The most energy we’ve seen in the search space has come from the AI upstarts like Perplexity. And of course Tiktok, before a ridiculously misguided law, Supreme Court decision, and inexplicable reversal, had blown most social networks out of the water. Why? At least one reason for the lack of innovation from the tech giants is their effective monopolies and their ability to buy out any threats.
It seems strange to have to make this point, but for competitiveness and innovation, we need real competition. We are in danger of falling into this trap again with the major AI companies, as they close off their products and try to corner the market on AI services. This is why DeepSeek came as such an embarrassment, and why it’s important to maintain a vibrant ecosystem of business innovation in this space. This means no rules that advantage a few large companies (like compute thresholds and restrictions on open access models), and more support for open systems (and open research) that can challenge the conventional wisdom of the large businesses and open up new directions for exploration.
The US Government should AVOID barriers on knowledge flows
As one Google engineer famously put it, “We have no moat, and neither does OpenAI.” I’d go as far as to say that there isn’t even a marshy swamp to slow things down. Knowledge is being shared and built on rapidly, and the AI discourse is global. This is a good thing. It keeps everyone on their toes and ready to innovate, and it leverages the natural strengths the US has had with our research prowess and our ability to convert new ideas into markets.
More importantly, barriers won’t work. DeepSeek demonstrated what can be done even with some constraints on access to hardware, and all a more protectionist approach will do is encourage even more countries to build their own in-house systems with their own talent. It goes without saying that together with open flows of knowledge, we need open flows of ideas, i.e an aggressively open approach to immigration, just like the US did all the years of the Cold War and beyond.
It’s ironic that I wrote these words in February 2025, where the current administration pays lip service to the idea of AI innovation, has installed leading tech evangelists in charge of the government, and yet pursues policies that are antithetical to every single point I have made above. They are decimating funding for research, attempting to destroy the independence of academic research, trying to use political muscle to protect US tech firms, and are closing off borders to both export of hardware and import of talent. These policies run the risk of damaging permanently the 70-year+ US advantage in basic and applied research that led to the growth of the most powerful industry in the world, to serve the interests of a few who are afraid of losing their current position of strength. I can only hope they see reason.
About the Author: Suresh Venkatasubramanian
Suresh Venkatasubramanian is the Director for the Center for Technological Responsibility, Reimagination, and Redesign (CNTR) with the Data Science Institute at Brown University, and is a Professor of Computer Science and Data Science. He recently served as Assistant Director for Science and Justice in the White House Office of Science and Technology Policy in the Biden-Harris Administration.
Image: Shutterstock