“Artificial intelligence (AI) will profoundly change human society… and change the world…” according to a 2017 report from China’s State Council.  Similar language appears in a 756 page US report in 2020 from the National Security Commission on Artificial Intelligence, (p. 1, 7) “AI is… world altering… [and includes] the most powerful tools in generations for expanding knowledge, increasing prosperity, and enriching the human experience.” 

So, what are the two countries doing about it?  The Chinese report set this goal: “by 2030, China’s AI theories, technologies, and applications should achieve world-leading levels.”  China has taken this recommendation very seriously.  Its most recent five year plan calls for a total investment of almost $1.4 trillion in “building ‘new infrastructure’ through AI, data centers, 5G, the Industrial Internet, and other new technologies.”

According to a global AI index compiled by Stanford University, in 2021 the US was still number one and China was number two. But while China is racing ahead, “America is not prepared to defend or compete in the AI era” according to the US 2020 commission (p. 3).

So, if AI is so important, what is it exactly?  There are two main schools of thought. 

Theoreticians who support a broad view of general AI see it as an attempt to program computers to function like the human brain, only much much better.  Unfortunately, to date, the field of trying to program general brain-like intelligence has been littered with failed predictions.  For example, as Melanie Mitchell noted in her excellent introductory book Artificial Intelligence (p. 19), in the early 1960s “the future Nobel laureate Herbert Simon predicted, ‘Machines will be capable, within twenty years, of doing any work that a [person] can do.’”  Oops, never mind.

Mitchell went on to explain that these days (p. 276) “Several surveys given to AI practitioners, asking when general AI or ‘superintelligent’ AI will arrive, have exposed a wide spectrum of opinion, ranging from ‘in the next ten years’ to ‘never.’ In other words, we don’t have a clue.”

In contrast, the more practical narrow approach to AI uses computers to solve just one problem at a time and has been astonishingly successful in the last few decades. According to Kai-Fu Lee, the former president of Google China, [in his book AI Superpowers, p. 19; italics added for emphasis] narrow AI programs “can now do a better job than humans at [a number of tasks including] identifying faces, recognizing speech, and issuing loans.”

Not to mention chess.  When world champion Gary Kasparov was defeated by IBM’s Deep Blue AI in 1997, he was so “stunned… that he accused the IBM team of cheating” (Mitchell p. 44).  If Kasparov felt that bad, he should have challenged the machine to play checkers, Go Fish, or “Duck, Duck, Goose” which he would have easily won.  The only thing Deep Blue could do was play chess.

Narrow AI accomplishes these feats with machine learning. 

Traditional computer programs are based on algorithms, standard sets of instructions to solve problems such as adding a set of numbers or finding its average.  If you run the program 100 times, the algorithm will always use the same steps. 

But an AI algorithm learns from experience, by applying massive amounts of computing power to huge sets of data, and modifying its approach based on results.  According to AI Superpowers (p. 18), “The data ‘trains’ the program to recognize patters by giving it many examples, and the computing power lets the program parse these examples at high speeds.” 

Kai-Fu goes on to explain (p. 24) that “Harnessing the power of AI today—the ‘electricity’ of the twenty-first century—requires four analogous inputs: [big] data, hungry entrepreneurs, AI scientists, and an AI-friendly policy environment.”  Both China and the US have all four.

But the more big data one has access to, the more accurate AI programs can learn to be. China’s biggest AI advantage is its access to the largest databases in the world.  The country’s population is nearly four times larger than the US, and its culture and laws do little or nothing to protect privacy. 

A few weeks ago, the New York Times published the results of a year-long study of “China’s expanding surveillance state.”  The conclusions were ominous: “Phone-tracking devices are now everywhere [in China]. The police are creating some of the largest DNA databases in the world. And the authorities are building upon facial recognition technology to collect voice prints from the general public.”

According to the cybersecurity website Comparitech 54 percent of the world’s 1 billion [surveillance] cameras are located in China.  And over a billion Chinese citizens use the messaging app WeChat (AI Superpowers, p. 27) to “[send] text and voice messages to friends, pay for groceries, book doctors’ appointments, file taxes, unlock shared bikes, and buy plane tickets, all without ever leaving the app… [This] wealth of information on users—their location every second of the day, how they commute, what foods they like, when and where they buy groceries and beer—will prove invaluable in the era of AI implementation.” 

One company to benefit from this is the Alibaba group, China’s answer to Amazon.  The two companies have somewhat different business models.  While Amazon sells directly to consumers and maintains an ever-growing number of warehouses to fulfill orders, Alibaba serves as a middleman to allow businesses to reach consumers directly.  Despite these differences, the two companies are in a race to attract to dominate global online markets.

One place this competition can be seen is in the AI algorithms used to offer buying recommendations.  At Alibaba, historically recommendations were based only on each person’s past purchases and browsing history.  In 2019, the company introduced a new “Artificial Intelligence Recommendation (AIRec) engine” to offer a more diverse group of recommendations.  The details are of course a secret, but Alibaba claims the new AI based system “outperforms… [traditional] algorithms by 20-100%.”

This race between Alibaba and Amazon to improve buying recommendations may make a difference to revenues, profits and stock price.  But whether Chinese consumers or American consumers buy more stuff they don’t need doesn’t make a difference to the future of world politics. 

However, it does matter a great deal if China’s advantage in AI research leads to improvements in two key areas Xi Jinping identified in 2019: “Big data should be used as an engine to power the innovative development of public security work and a new growth point for nurturing combat capabilities.”

We will talk about military uses of AI in a future post, but we have already seen the effects of improved “public security” in a number of posts in this blog.  My post on China’s social credit system describes how big data is being used to develop systems to reward “good citizens” — which could mean anything from serving in an important Party position to just getting to work on time every day.  Rewarded citizens may be able to rent an apartment without a deposit, get a better interest rate at banks, skip hospital waiting rooms, and even get more matches on dating websites.  On the other hand, dissenters can be punished by preventing them from getting a good job, staying at a luxury hotel, or even buying a plane ticket.  At a more threatening level, as described in two other posts, mass surveillance is being used to imprison activists among the Uighurs, an ethnic group of about 11 million Sunni Muslims who live in Western China.

The New York Times investigation lists many other applications of AI, such as warning “the police if… a drug user makes too many calls to the same number [or]…   each time a person with a history of mental illness gets near a school.”

And if these applications do not seem menacing enough, consider the way local officials can misuse these AI systems.  Perhaps you’ve heard about the ongoing scandal in which Chinese savers have been unable to withdraw their money from four Chinese rural banks, and the government’s attempts to prevent the savers from publicly demonstrating.  One of the techniques local officials applied was to misuse China’s AI based covid tracking system. 

Some bank protestors found that “mobile apps used to identify and isolate people who might be spreading Covid had turned from green — meaning safe — to red, a designation that would prevent them from moving freely…. [local] authorities [who were] under pressure to account for the episode, later punished five officials for changing the codes of more than 1,300 customers.”

Whether used for good or bad, the race to be first in AI will have enormous political implications.  In a Foreign Affairs article entitled How artificial intelligence will reshape the global order, Nicholas Wright describes the threat of “digital authoritarianism… [in which] new technologies will enable high levels of social control at a reasonable cost… [and empower dictators to] make their citizens rich while maintaining control over them… AI will offer authoritarian countries… the first… plausible alternative to liberal democracy… since the end of the Cold War.”

Let’s hope Mr. Wright is wrong.