China's AI Ambitions

M.E. Strickland

In July, China’s State Council issued a proposal laying out a national strategy for AI, the “Next Generation AI Development Plan.” Or perhaps strategy is not quite the word. As is usual with such policy memoranda, it is more of a wishlist than a plan, a set of goals without much explanation of how they are to be arrived at.

For example, the proposal lays out a timeline from 2020 to 2030, with the objectives that the “core AI industry” should grow to about $150 billion by that end date, with related industries to reach the gargantuan scale of $1.5 trillion. That is a tall order, and the proposal does not define what exactly should be counted as part of the AI industry, let alone give details on how those numbers are to be reached.

Still, the proposal deserves outside observers’ attention precisely because it shows where the policymakers’ attention now lies. Like most such statements from China’s central government, it is a signal to local governments and industry about where to focus their energy—and where the government itself might choose to throw its weight.

For those interested, a very fine translation of the plan has been made available here. But for those who lack the will (or masochism) to wade through pages of official Chinese government prose, we have made a straightforward statistical breakdown of the text to help give a different perspective on what’s in it and what might be the real priorities.

(A slightly pedantic disclaimer: pulling a text apart like this does lose most of the nuance from the context, but in return it helps to show the relative weight that’s actually given to the topics discussed).

Simply put, the proposal does two things: it lists technologies and research directions to pursue, and economic sectors or applications to concentrate on.

The first includes things like natural language processing, VR, “hybrid intelligence” and “swarm intelligence,” sensor equipment and so on. The frequency with which these different subjects are mentioned, however, suggests they are not all valued equally, with (big) data, robots, quantum computing, and “brain-like” (or neural or neuromorphic) technologies receiving the most emphasis.

(As a basis of comparison, the term “AI” appears 349 times in the document.)

The emphasis on robotics isn’t too surprising, as it’s a fairly general category of technology. Notable here, though, is that the proposal cites “service robots” seven times, while “industrial robots” are only mentioned twice. The reason for that may have to do with concerns about elder care (more on that in just a moment).

The mentions of “[processor] chips” also shouldn’t be overlooked, as the government has made building a domestic chipmaking industry a priority in its separate Made in China 2025 policy. Yet on the other hand things like drones and natural language processing get surprisingly little attention.

But then there’s the question of what applications the government hopes for AI to have. Outside of general mentions of the economy and infrastructure, the main priorities seem to be education, medicine, elder care, and smart cities. Smart agriculture (in the form of sensor equipment and big data) is listed as well, though it seems to get shorter shrift.

Logistics, strangely, doesn’t get many mentions either, though that is certainly a focus for Alibaba.

Equally important are the absences. The focus on education, medical care, and old age care serves two segments of the population, the youth and the elderly. What does not receive much discussion is the middle segment of the population—working age adults. And what has been the preoccupation in the US, the possible existential threat that AI and robotics pose to human labor, is not on display here. There is a great deal of talk of cultivating research and engineering talent (“professional talent” is mentioned no less than 34 times), but only a bare handful of mentions of such things as “changes in employment.” Remarkably, the terms “automation” and “unemployment” are each only mentioned once.

But there’s also something less obvious here. Strewn throughout the proposal are a set of other, more abstract themes that maybe reflect some of the deeper concerns about what AI can, and will, do to society.

“Open” and “open source” (distinct words in the Chinese text) show up quite a bit, as does “ethics” (“privacy” not so much). But one of the biggest themes by far here is “security,” with one in six instances in the form of “national security.”

And there is one final theme here, so ubiquitous that it could almost forms a background and escapes notice. While this is a national strategy for AI, the words “world,” “international,” and “global” collectively apear 59 times, often in phrases like “international influence” and “world leading.” Technical breakthroughs and societal gains aren’t the only aim here; the whole of the policy is framed against “international standards.”

It may not be clear yet exactly how China will build an AI industry—it’s not even very clear just what an AI “industry” is or might look like. But the new plan makes it exceedingly clear that the government doesn’t want just a domestic AI industry. It wants a global one, and it wants China to lead it.

(Image credit: Lorie Shaull)