The Intelligence With Hands

That distinction separates three things a geopolitical race can compress into one: technical capability, industrial capacity and legitimate authority. A machine might improve the first two without conferring the third. The trophy does not come with a governance system attached.

In an October 12 Foreign Affairs essay, Carnegie researchers Scott Singer and Pavlo Zvenyhorodskyi argue that China's combination of hardware and software could yield an advantage that a U.S. emphasis on AI models misses. Their subject is embodied AI: intelligence joined to machines capable of physical movement.

The essay's text was accessible during reporting. Its central claim remains a strategic argument, not a demonstrated verdict on which country will prevail. The important proposition is narrower than its race framing: success at digital tasks does not automatically establish the capacity to perform physical work.

The Factory Is Part of the Intelligence

Singer and Zvenyhorodskyi describe a potential reinforcing process. Deploy robots, collect information about physical tasks, use that information to improve the systems, then deploy more capable machines. In their account, China's manufacturing base could provide both customers and a training environment.

That is a claim about industrial organization as much as model performance. A useful robot needs more than a persuasive answer. It must function in a particular workplace, with equipment, maintenance and people whose movements do not necessarily resemble a laboratory exercise.

The authors locate an older policy pattern beneath the new machinery. They compare support for robotics with China's development of electric vehicles and solar energy: central priorities pursued through competing local governments, subsidies and other incentives. Their comparison includes waste and overcapacity, not merely successful firms. They also identify strained local finances and uncertain commercial viability as constraints on a robotics boom.

Those qualifications matter. Announcing an industrial priority is not the same as financing it sustainably; financing machines is not the same as making them useful. A strategic plan can contain a remarkably convincing future. The factory still has to operate in the present.

Whose Feedback Improves the System?

The essay's proposed data cycle also raises a political question beyond national competitiveness. If a workplace helps train a machine, what authority do the people in that workplace have over the resulting system?

Consider a hypothetical factory adopting AI-powered robots. Its operators could contribute the adjustments, warnings and knowledge of awkward exceptions that make deployment workable. A procurement contract might nevertheless leave the employer dependent on the supplier for updates and repairs. The people supplying practical knowledge might have little influence over how that knowledge is used.

These are possibilities to investigate in actual contracts and workplaces, not findings about a named Chinese or American deployment. They identify what a national ranking cannot tell us.

Coherenceism's test is whether an arrangement incorporates the feedback of those it affects rather than buying order with their silence. Here that requires distinguishing feedback used to improve a product from feedback empowered to change a decision. A worker's warning should not become merely another training example while the unsafe task continues.

A serious assessment would therefore ask who can suspend operation, who investigates a failure, and whether workers can challenge changes to their tasks without retaliation. These questions do not settle industrial competition. They establish what competition is supposed to serve.

Singer and Zvenyhorodskyi offer a useful warning against confusing software leadership with mastery of the physical economy. But replacing a model leaderboard with a robot count would repeat the measurement error at a larger scale.

If machines gain more capacity to act while affected people gain no capacity to object, the intelligence will have acquired hands. The authority will remain in familiar ones.