The Biden administration this week introduced new export restrictions designed to control AI’s progress globally and ultimately prevent the most advanced AI from falling into China’s hands. The rule is just the latest in a string of measures put in place by Donald Trump and Joe Biden to keep Chinese AI in check.
With prominent AI figures including OpenAI’s Sam Altman and Anthropic’s Dario Amodei warning of the need to “beat China” in AI, the Trump administration may well escalate things further.
Paul Triolo is a partner at DGA Group, a global consulting firm, a member of the council of foreign relations, and a senior adviser to the University of Pennsylvania’s Penn Project on the Future of US-China Relations. Alvin Graylin is an entrepreneur who previously ran China operations for the Taiwanese electronics firm HPC. Together they have been tracking China’s AI industry and the impact of US sanctions. In an email exchange, Triolo and Graylin discussed the latest sanctions, Silicon Valley rhetoric, and the dangers of seeing global AI as a zero sum game.
This interview has been edited for clarity and brevity.
What do you make of the new AI diffusion rule from the US government this week, which aims to curb China’s access to AI?
Paul Triolo: Generally, it focuses on clusters of high-performance computing. The rule also puts controls on proprietary model weights for the most advanced “frontier” models but it is unclear how performance levels will be determined, and most open-weight [freely shared] AI models are tuned and improved by users, including major AI companies in China.
The complex rule and unclear compliance conditions inject considerable uncertainty into the long-term plans of both medium and major US and western hyperscalers.
For hyperscalers like Google, Microsoft, AWS, and Oracle, the rule introduces critical issues, including slowed or more complex international expansion, new compliance and legal costs, impact on global R&D, and uncertain enforcement requirements.
How have previous measures, including the sanctions introduced by the first Trump administration, affected the AI industry there?
Paul Triolo: US export controls have slowed China, but at a high level the sanctions have unified the will and efforts of the Chinese government to become more self-reliant. It has plowed tens of billions into helping local players catch up technologically or scale capacity in core areas, resulting in significant changes within the semiconductor industry and its ability to support the advanced hardware for developing frontier AI models.
Chinese AI developers have gotten very good at leveraging legacy AI hardware from western firms and gradually integrating domestic alternatives into their development process. Chinese firms will continue to innovate across the AI hardware and software stack, if not at the pace of their western counterparts.
Why do you think so many in Silicon Valley are now talking about the need to “beat China” in AI?
Paul Triolo: There is a growing link between conservative venture capitalists, mostly located in Silicon Valley, and technology companies whose business models depend on hyping the China threat. This is a troubling combination that conflates the China threat, personal gain, and push back against regulation of advanced AI. It also portrays US China competition around AI as zero sum, which is particularly dangerous.
The Trump administration will be heavily influenced by this zero sum narrative, given the prominent role that will be played by technology leaders and venture capitalists. Already key players in the AI space with close ties to the US government such as OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei have written op-eds or made comments or approved company statements that align with the China AI threat framing and zero sum competition.
It is likely that Elon Musk will push back on this narrative. He has already called for international cooperation and responsible AI governance. And he has voiced concerns about AI safety, which he believes must be addressed on a global basis, including through engagement with China.
Shouldn’t the US want to beat China?
Alvin Graylin: This type of framing suggests that any collaboration between the US and China, even in areas such as medicine and energy development, should be avoided in the name of ensuring the “dominance” of the US and allies in the development of AI. In truth it is impossible for either country to maintain a sustainable lead over the other and “win” this race.
The dominant paradigm driving US policy on China and AI is the assumption of future conflict, and this comes with the still fuzzy notion that AI will be decisive in both economic and military domains. Officials have, however, struggled to identify the exact “significant military applications” that are of concern.
On the other hand, collaborative research has proven critical to fundamental progress for AI in recent decades. Almost half of all top AI researchers globally (47 percent) were born or educated in China, according to industry studies. Breaking this virtuous cycle appears risky and counterproductive, but the trend lines are in this direction. Another major risk is that China pulls out of multilateral efforts to develop an AI governance framework, making it impossible to have a workable global arrangement for AI safety.
How much of a lead does the US actually have?
Alvin Graylin: The Silicon Valley consensus on AI previously held that the US had a one or two year lead over China in AI development.
Some popular generative video and photo models from China now appear to be as good as leading US offerings. The recent release of the surprisingly competent DeepSeek r1 and v3 models, which outperform the most advanced OpenAI o1 reasoning and GPT4o models respectively, seems to indicate that gap is actually closing. A pivot to inference compute [used when running models rather than training them], which is less reliant on chips fabricated with the most advanced semiconductor nodes, is nullifying the sanctions. What’s most impressive is that DeepSeek was only founded in 2023 and has less than 100 staff and much more limited computing resources than the western frontier labs.
The technology controls the US government has been imposing on China is actually forcing Chinese developers to become more creative and efficient with resources than their western counterparts. China may also have advantages in accessing the private data that will be needed to train frontier models. With today’s national policies limiting access to data from each country to others, we are heading down a very dark path where we can potentially have 193 ultra-intelligent “sovereign AI” models each limited to a certain culture and values. This is great for the Nvidia stock price, but a dystopian nightmare for humanity.
What advice would you give to the incoming president and his staff?
Paul Triolo: The escalating AI competition between the US and China poses significant threats not only to both nations but also to the entire world. They could lead to outcomes that threaten global peace, economic stability, and technological progress.
The US and China nations need to recalibrate their approach to AI development, and move away from viewing AI primarily as a military asset. They should also establish a robust dialogue for the development of common AI governance standards and support the development of a global AI safety coalition. Both nations should also agree on shared standards for the responsible use of AI and collaborate on tools that can monitor and counteract misuse globally. Governments worldwide should provide incentives for academic and industry collaborations across borders. A global effort akin to the CERN for AI will bring much more value to the world and a peaceful end than a Manhattan Project for AI, which is being promoted by many on the Hill today.
Our choice is stark but simple: proceed down a path of confrontation that will almost certainly lead to mutual harm, or to pivot towards collaboration, which offers the potential for a prosperous and stable future for all.
Do you think the US should stop trying to slow China’s AI advancements? Share your thoughts in the comments below.
Source : Wired
