07-02-2025
China's ability to launch DeepSeek's popular chatbot draws US government panel's scrutiny
China's ability to launch DeepSeek's popular chatbot came under scrutiny before a US government advisory panel on Thursday, with one witness stressing the role that American technology played and another cautioning that the country's ability to "iterate" other breakthroughs in the industry could overcome this factor.
Testimony before the US-China Economic and Security Review Commission (USCC) also came with a warning about China's ability to dominate the "potentially world-changing technology of" artificial general intelligence, or AGI.
Hanna Dohmen of Georgetown University's Centre for Security and Emerging Technology, said China's "most advanced [AI] models, including DeepSeek R1, are largely relying on US-made semiconductors, including chips that companies stockpiled before the controls took effect.
Do you have questions about the biggest topics and trends from around the world? Get the answers with SCMP Knowledge, our new platform of curated content with explainers, FAQs, analyses and infographics brought to you by our award-winning team.
"As these stockpiles deplete over the next couple of years, the controls have the potential to create a growing gap between the AI chip quantity and quality inside and outside of China," said Dohmen.
"But that gap will only hold if US export controls on chip manufacturing tools and other measures aimed at slowing China's semiconductor fabrication capacities are effective.
"Similar to the stockpiles of AI chips, Chinese equipment firms also stocked up on foreign equipment before controls were implemented," she added. "This, once again, imposes a lag between the control when the controls were implemented and when the controls will bite."
The hearing, titled "Made in China 2025: Who Is Winning?", referred to a plan Beijing unveiled in 2015 to establish the country's dominance in industries ranging from AI, robotics, aerospace and new materials to new energy vehicles.
It came amid mounting concern among US policymakers about the unexpected success of DeepSeek-R1, the company's open-source reasoning model released on January 20.
The AI model has shown capabilities comparable to those of more advanced models from OpenAI, Anthropic and Google, but with significantly lower training costs.
While Dohmen asserted that export controls could slow China's AI advances in the medium term, she cautioned that they create more incentive to "innovate around the controls".
"We must also consider China's own innovation capabilities and response," she said. "By limiting China's access to chips and equipment, export controls are creating an incentive to innovate around controls."
"Chinese companies are also pursuing technical strategies such as chiplet packaging and focusing on compute and algorithmic efficiencies to overcome restrictions," Dohmen added.
Chiplet manufacturing integrates smaller chips, each with a dedicated function such as data processing or storage, that are connected to become one system instead of putting all components on a single piece of silicon.
Hannah Dohmen of Georgetown University in her testimony in Washington on Thursday says China's most advanced AI models are largely relying on US-made semiconductors. Photo: SCMP alt=Hannah Dohmen of Georgetown University in her testimony in Washington on Thursday says China's most advanced AI models are largely relying on US-made semiconductors. Photo: SCMP>
MIT Technology Review chose chiplets as one of its "10 breakthrough technologies of 2024".
The USCC asked another witness - David Lin of Special Competitive Studies Project, a think tank founded by former Google CEO Eric Schmidt - why it should be concerned about China's AI advances if DeepSeek relied on a "distillation" of OpenAI's ChatGPT.
Distillation is when developers create smaller models that learn from larger ones by copying their behaviour and decision-making patterns. DeepSeek's performance has sparked a wave of accusations that it has reverse-engineered the capabilities of leading US technology, such as the AI-powering ChatGPT.
Lin replied that China should be seen as an "iterator" as opposed to an "imitator".
China "hasn't replicated the US success in introducing true scientific breakthroughs that have revolutionised the world, going back to mRNA vaccines, going back to ChatGPT, going back to the iPhone", he said.
However, Lin continued, the country has "been able to iterate on these things to make [them] better, faster, cheaper, and we're seeing that unfold before our eyes with DeepSeek's R1".
The former US diplomat said he wanted to "put on the record that DeepSeek's stated company mission is to work towards AGI, which is a testament to China's AI ambitions".
"This is a matter about competition toward being a first mover advantage, the first person to AGI will bring with [them] a whole downstream effect of economic, political and energy supply-chain influence around the world," Lin added.
This article originally appeared in the South China Morning Post (SCMP), the most authoritative voice reporting on China and Asia for more than a century. For more SCMP stories, please explore the SCMP app or visit the SCMP's Facebook and Twitter pages. Copyright © 2025 South China Morning Post Publishers Ltd. All rights reserved.
Copyright (c) 2025. South China Morning Post Publishers Ltd. All rights reserved.
