Newsweek/Getty/Shenzhen Story/YouTube
By David Averre
0
Share
Newsweek is a Trust Project member
See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.
The Ultimate Fighting Championship event at the White House in June delivered some spectacular knockouts, but nothing close to the savagery that occurred in a mixed martial arts bout in China last week.
One fighter’s head was quite literally hanging from their shoulders during the fight in Shenzhen on Thursday, yet somehow they kept swinging until the final bell. A normal human could not hope to withstand such punishment. Or any human, in fact.
The fighter in question was a robot—one of dozens of full-scale humanoids that, based on all available photo and video evidence, are programmed to never give up in a war. This surreal display of science *non-*fiction, though impressive, offers a concerning preview of a much broader conflict with catastrophic potential: the U.S.-China race for AI supremacy.
Read More on AI
Right now, neither competitor has thrown a real strike; they’re circling each other in the cage, observing their opponent’s strengths and weaknesses and refining their strategies. But the first punch is rapidly approaching.
America’s Next Top Model—Or China’s?
The lingering narrative that China produces cheaper, more accessible AI models but cannot hope to match the raw power and capability of America’s frontier products is both outdated and dangerously reductive.
While the likes of Mythos and Fable—the leading models offered by American AI lab Anthropic—are believed to be the world’s most powerful systems, the gap to China is closing fast.
Kimi K3, the latest model released by Beijing-based Moonshot AI last week, is going toe-to-toe with the best Anthropic and OpenAI have to offer across a variety of coding tests and research tasks (though Kimi K3’s developers have said Anthropic and OpenAI’s top models remain a little more polished).
A better lens through which to observe the battle of the AI superpowers is an ideological one.
...
Fighting Philosophies
America and China have distinct methods of pursuing AI dominance. They are partially dictated by the realities of resource access and geopolitical conditions, but more so by the fundamentally different philosophies of where AI will lead.
The U.S.’s leading labs are concentrating maximum force. They have the very best chips, the largest reserves of computing power and a seemingly endless flow of capital from America’s deepest pockets to churn out ever more capable models at breakneck speed. But these models are closed-weight: customers pay high prices to access the model through an app or cloud interface but can’t see the code hiding behind the user interface.
In other words, the ingredients to the secret sauce remain just that.
China’s leading labs, meanwhile, are pursuing efficiency and reach. Their models may remain less capable than their American counterparts in extreme-use cases, but they are designed to achieve more with constrained hardware and, most importantly, are open-weight.
This means developers worldwide can download the AI models, run them on their own infrastructure, adapt them for local languages, and integrate them into their own public and private systems. Beijing’s companies don’t give away every morsel of training detail and line of code, but access to their products is largely available to all, for far less money.
Even Moonshot, whose hotshot CEO Yang Zhilin moved back to China after earning his PhD at Carnegie Mellon University in Pittsburgh, said Kimi K3’s weights will be released for public access by the end of the month.
Think of it like this: America’s labs are producing luxury supercars. They’re comfortable, made of premium materials, and are brutally fast—but only a select few can afford them. China’s automobiles might have slightly fewer horsepower, but you’d be hard-pressed to tell the difference, and they’re for sale in your local dealership.
...
A Cold War, for Now
This underpins China’s broader long-term strategy on AI: making powerful models cheap and easily accessible is an effective way to increase market share.
When middle- and low-income powers that could not hope to develop their own robust AI programs go in search of tools to help digitize their societies and they’re forced to choose between very expensive private systems and cheap, readily available open-source ones, many are likely to look east.
Every government department, public organization or private company that builds around a Chinese model creates demand for compatible cloud services, developer tools, updates and technical expertise—and steadily becomes locked into the Chinese ecosystem.
Market share is only part of the plan. As AI continues to shape the trajectory of humankind, Beijing also understands that by bringing more countries under its wing, it gains more influence over future AI governance frameworks and the decisions on how AI is regulated.
That ambition took institutional form in Shanghai on Thursday, July 16, when 29 countries signed the founding agreement for the World Artificial Intelligence Cooperation Organization at a conference attended by a slew of prominent world leaders, among them United Nations secretary-general António Guterres.
According to the Chinese Communist Party, WAICO will coordinate international AI development and governance initiatives - including 5,000 training opportunities for developing countries and new AI cooperation centers linked to Global South conglomerates like the African Union and the Arab League.
Related Story
To Compete with China, US Must Win Next Generation Wireless | Opinion
WAICO is only days old, but the comparison between it and BRICS is obvious. BRICS—a group of countries designed to foster economic development between emerging economies led by China and Russia—has long been touted as an effort to usurp Western dominance over the global financial system.
Unlike BRICS, which has thus far had little to no impact on the dollar’s status as the world’s reserve currency, WAICO may prove to be a much more effective tool for the CCP, helping it dictate the terms by which most of the world uses AI.
For now, each side holds an edge over the other that keeps them locked in stalemate.
America and its partners retain control over the most advanced chips and specialized manufacturing tools that power the best AI models. But China is the leading refiner of 19 of the 20 critical minerals needed to keep manufacturing this hardware, according to the International Energy Agency.
That mutual vulnerability is helping keep the AI war cold. If Washington secures mineral supply chains beyond Beijing’s reach, or China learns to produce the most advanced chips without Western technology, the cold war could turn hot very quickly.
View 0 Comments
Submit Correction
View Editorial & AI Guidelines