Scientists in China are potentially on track to build next-level artificial intelligence that is infused with Chinese Communist Party values and which could propel China ahead of the US in the race for human-like, “artificial general intelligence”, a new report says.
The testbed is the central city of Wuhan, notorious for being the place from which COVID-19 emerged, possibly from a laboratory, but a city which is also a major center for other scientific and technological research — including AI.
Aided by massive state support, two leading AI institutes that are headquartered in Beijing have set up branches in Wuhan to cooperate on sophisticated alternatives to the large generative AI models – LLMs – that occupy nearly all of western AI developers’ and policymakers’ attention, a team at Georgetown University’s Center for Security and Emerging Technology (CSET) said in the report published on Monday and made available exclusively in advance to Newsweek.
China’s multifaceted and innovative approach to AI meant the United States risked being left behind – and it might already be too late, lead author William C. Hannas told Newsweek.
“We need to work quickly and smartly. Pouring billions more into data centers isn’t
enough. Competing approaches are needed,” Hannas said.
“The two advantages the U.S. has, chips and algorithms, are being eroded by indigenous Chinese workarounds. Worse, the two sides are not playing the same game. U.S. companies are fixated on large statistical models, whereas China covers its bets by funding multiple AGI paths,” said Hannas, CSET’s lead analyst and formerly the CIA‘s senior expert for China open-source analysis.
AI competition between China and the U.S. is intensifying, with China surprising the world in January by launching DeepSeek, a successful generative AI model in an area where the U.S. was believed to hold an uncontested lead with offerings such as OpenAI‘s ChatGPT.
Yet Chinese scientists and the state were going further by “embodying” AI into real-world environments, Hannas and two other CSET researchers, Huey-Meei Chang and Daniel H. Chou, wrote in “Focus on Wuhan’s AI Development: China’s Alternative Springboard to Artificial General Intelligence (AGI)”.
“China’s top state-funded AI institutes are exploring alternative approaches to AGI that involve embodying AI algorithms in real environments. Imbued with the Chinese Communist Party’s pre-defined values, the AI interacts with its natural surroundings, learning as it proceeds,” they wrote. A key component of China’s pathway is the integration of neuroscience approaches.
“The Wuhan implementation is seen by its state-backed entities as a stepping stone to deployment throughout China, raising questions about the type of technosociety with which the United States needs to compete,” the researchers wrote.
The broad scientific consensus is that whoever first develops successful models of artificial general intelligence would have massive repercussions on global power.
“While focusing on AI safety and the dangers of weaponization, we should also keep in mind the real possibility of being out-competed by a country that moves more quickly and decisively to realize the promises that AI offers,” Hannas, Chang and Chou wrote.
Newsweek contacted the institutes and leading Chinese AI scientists for comment, including the Tsinghua University AI Institute for International Governance which addresses ethical issues surrounding AI, but did not hear back. The Chinese embassy in Washington, D.C. did not respond to a request for comment.
Testbed Wuhan
The joint effort in Wuhan is led by the Chinese Academy of Sciences’ Institute of Automation (CASIA), Peking University’s PKU-Wuhan Institute for Artificial Intelligence, and the technology giant Huawei. It aims to “infuse” Wuhan’s industrial and commercial enterprises with AI and to deploy a “social simulator” that expands AI’s reach into all aspects of daily life, the report said.
Other institutes are part of an interlocking research landscape that aims for “an interpretable, trustworthy, and evolvable multimodal AI infrastructure platform for open and complex environments,” or “a new generation of AI infrastructure”, according to the Yangtze Daily.
The U.S. and China were both dedicating vast resources to achieving AGI, said Valentin Weber of the German Council on Foreign Relations Center for Geopolitics, Geoeconomics, and Technology.
“The US’s relative strength lies in AI chips and computing centers. But China has rolled out AI in most of its cities for urban governance purposes, while also pursuing the development of LLMs like the US. China is taking multiple pathways toward AGI. It is arguably farther in deploying AI in the real world. This increases its chances to reach AGI first,” Weber said.
Why Wuhan?
Wuhan is a research and transportation hub in China, with dozens of interlocking institutes, three national development zones, four technology parks, and 2,000 research and high-tech enterprises supported by multiple networks of computing centers and grids whose total capacity is projected to reach 30 exaflops this year, creating China’s highest municipal concentration of compute power, the report says, citing Chinese state media such as the Yangtze Daily.
While it is difficult to draw direct comparisons, one of the fastest supercomputers in the U.S. – El Capitan at Lawrence Livermore National Laboratory in California – has a capacity of 1.7 exaflops, or 1.742 quintillion calculations per second, according to the laboratory.
In a striking development, in May, a consortium of Chinese companies launched the first 12 satellites of a planned space-based supercomputing platform for AI, Star Compute, according to the China Aerospace Science and Technology Corporation (CASC).
Signs of Wuhan’s ambition abound. Earlier this year the municipal government announced plans to boost the city’s AI industry in 2025 by targeting five key areas: “AI plus robots, AI plus automobiles, AI plus PC and server, AI plus mobile phones, and AI plus eyeglasses,” it said in a notice published on Mar. 12.
“Social Governance”
The goal is a “value-driven” form of AGI set to become “the operating model of a coexisting smart society” and a tool for social governance, the authors cite the AI scientist Wu Zhiqiang, the co-manager with Zhu Songchun of the PKU-Wuhan AI institute, as saying.
Said Weber: “Social governance is a euphemism for social control. Achieving social control is a central theme of China’s AGI pursuit. The closer China gets to reaching AGI, the higher are its chances of preserving regime security.”
“China has a track record of achieving supremacy in technological areas that its leadership deems worth pursuing,” Weber said, pointing to batteries, electric vehicles, and facial recognition.
Both Wu and Zhu were recipients of numerous U.S. Government grants from defense-oriented agencies during their long tenures in the United States. Newsweek has previously reported on more than $30 million in federal grants awarded to Zhu.
“Given today’s global collaboration network, China doesn’t need to be at the cutting edge to score big wins, although now it often is. Diffusion and application matter more, where China has a commanding lead,” Hannas said.
The U.S. was paying “scant attention” to China’s high-tech research and development. “China has incredible insight into U.S. AI research, whereas we know just enough about theirs to know what we don’t know.”
The post China’s Next-Level AI Could Overtake US: New Report appeared first on Newsweek.
