AI Arms Race and what it means

In the spring of 2016, an artificial intelligence system called AlphaGo defeated a world champion Go player in a match at the Four Seasons hotel in Seoul. In the US, this momentous news required some unpacking. Most Americans were unfamiliar with Go, an ancient Asian game that involves placing black and white stones on a wooden board. And the technology that had emerged victorious was even more foreign: a form of AI called machine learning, which uses large data sets to train a computer to recognize patterns and make its own strategic choices.

Still, the gist of the story was familiar enough. Computers had already mastered checkers and chess; now they had learned to dominate a still more complex game. Geeks cared, but most people didn’t. In the White House, Terah Lyons, one of Barack Obama’s science and technology policy advisers, remembers her team cheering on the fourth floor of the Eisenhower Executive Building. “We saw it as a win for technology,” she says. “The next day the rest of the White House forgot about it.”

In China, by contrast, 280 million people watched AlphaGo win. There, what really mattered was that a machine owned by a California company, Alphabet, the parent of Google, had conquered a game invented more than 2,500 years ago in Asia. Americans don’t even play Go. And yet they had somehow figured out how to vanquish it? Kai-Fu Lee, a pioneer in the field of AI, remembers being asked to comment on the match by nearly every major television station in the country. Until then, he had been quietly investing in Chinese AI companies. But when he saw the attention, he started broadcasting his venture fund’s artificial intelligence investment strategy. “We said, OK, after this match, the whole country is going to know about AI,” he recalls. “So we went big.”

In Beijing, the machine’s victory cracked the air like a warning shot. That impression was only reinforced when, over the next few months, the Obama administration published a series of reports grappling with the benefits and risks of AI. The papers made a series of recommendations for government action, both to stave off potential job losses from automation and to invest in the development of machine learning. A group of senior policy wonks inside China’s science and technology bureaucracy, who had already been working on their own plan for AI, believed they were seeing signs of a focused, emerging US strategy—and they needed to act fast.

In May 2017, AlphaGo triumphed again, this time over Ke Jie, a Chinese Go master, ranked at the top of the world. Two months later, China unveiled its Next Generation Artificial Intelligence Development Plan, a document that laid out the country’s strategy to become the global leader in AI by 2030. And with this clear signal from Beijing, it was as if a giant axle began to turn in the machinery of the industrial state. Other Chinese government ministries soon issued their own plans, based on the strategy sketched out by Beijing’s planners. Expert advisory groups and industry alliances cropped up, and local governments all over China began to fund AI ventures.

China’s tech giants were enlisted as well. Alibaba, the giant online retailer, was tapped to develop a “City Brain” for a new Special Economic Zone being planned about 60 miles southwest of Beijing. Already, in the city of Hangzhou, the company was soaking up data from thousands of street cameras and using it to control traffic lights with AI, optimizing traffic flow in much the way AlphaGo had optimized for winning moves on the Go board; now Alibaba would help design AI into a new megacity’s entire infrastructure from the ground up.

On October 18, 2017, China’s president, Xi Jinping, stood in front of 2,300 of his fellow party members, flanked by enormous red drapes and a giant gold hammer and sickle. As Xi laid out his plans for the party’s future over nearly three and a half hours, he named artificial intelligence, big data, and the internet as core technologies that would help transform China into an advanced industrial economy in the coming decades. It was the first time many of these technologies had explicitly come up in a president’s speech at the Communist Party Congress, a once-in-five-years event.

In the decisive span of a few months, the Chinese government had given its citizens a new vision of the future, and made clear that it would be coming fast. “If AlphaGo was China’s Sputnik moment, the government’s AI plan was like President John F. Kennedy’s landmark speech calling for America to land a man on the moon,” Kai-Fu Lee writes in his new book, AI Superpowers.

Meanwhile, as Beijing began to build up speed, the United States government was slowing to a walk. After President Trump took office, the Obama-era reports on AI were relegated to an archived website. In March 2017, Treasury secretary Steven Mnuchin said that the idea of humans losing jobs because of AI “is not even on our radar screen.” It might be a threat, he added, in “50 to 100 more years.” That same year, China committed itself to building a $150 billion AI industry by 2030.

Only slowly, pushed mainly by the Pentagon, has the Trump administration begun to talk about, and fund, national AI initiatives. In May, secretary of defense James Mattis read an article in The Atlantic by Henry Kissinger, who warned that AI was moving so quickly it could soon subvert human intelligence and creativity. The result, he warned, could be the end of the Enlightenment; he called for a government commission to study the issue.

Many AI experts pooh-poohed Kissinger’s article for extrapolating too broadly and darkly from the field’s narrow accomplishments. Mattis, however, pulled the article into a memo for President Trump. That month, Michael Kratsios, Trump’s top adviser on technology, organized a summit on the subject of AI. In an interview with WIRED this summer, Kratsios said the White House was fully committed to AI research and to figuring out “what the government can do, and how it can do it even more.” In June, Ivanka Trump tweeted out a link to the Kissinger piece, praising its account of “the ongoing technological revolution whose consequences we have failed to fully reckon with.”

But if the Trump White House was relatively slow to grasp the significance and potential of AI, it was quick to rivalry. By midsummer, talk of a “new cold war arms race” over artificial intelligence was pervasive in the US media.

At the dawn of a new stage in the digital revolution, the world’s two most powerful nations are rapidly retreating into positions of competitive isolation, like players across a Go board. And what’s at stake is not just the technological dominance of the United States. At a moment of great anxiety about the state of modern liberal democracy, AI in China appears to be an incredibly powerful enabler of authoritarian rule. Is the arc of the digital revolution bending toward tyranny, and is there any way to stop it?

After the end of the Cold War, conventional wisdom in the West came to be guided by two articles of faith: that liberal democracy was destined to spread across the planet, and that digital technology would be the wind at its back. The censorship, media consolidation, and propaganda that had propped up Soviet-era autocracies would simply be inoperable in the age of the internet. The World Wide Web would give people free, unmediated access to the world’s information. It would enable citizens to organize, hold governments accountable, and evade the predations of the state.

No one had more confidence in the liberalizing effects of technology than the tech companies themselves: Twitter was, in one executive’s words, “the free speech wing of the free speech party”; Facebook wanted to make the world more open and connected; Google, cofounded by a refugee from the Soviet Union, wanted to organize the world’s information and make it accessible to all.

As the era of social media kicked in, the techno-optimists’ twin articles of faith looked unassailable. In 2009, during Iran’s Green Revolution, outsiders marveled at how protest organizers on Twitter circumvented the state’s media blackout. A year later, the Arab Spring toppled regimes in Tunisia and Egypt and sparked protests across the Middle East, spreading with all the virality of a social media phenomenon—because, in large part, that’s what it was. “If you want to liberate a society, all you need is the internet,” said Wael Ghonim, an Egyptian Google executive who set up the primary Facebook group that helped galvanize dissenters in Cairo.

It didn’t take long, however, for the Arab Spring to turn into winter—in ways that would become eerily familiar to Western countries in a few years. Within a few weeks of President Hosni Mubarak’s departure, Ghonim saw activists start to turn on each other. Social media was amplifying everyone’s worst instincts. “You could easily see the voices in the middle become more and more irrelevant, the voices on the extremes becoming more and more heard,” he recalls. The activists who were vulgar or attacked other groups or responded with rage got more likes and shares. That gave them more influence, and it gave otherwise moderate people a model to emulate. Why post something conciliatory if no one on Facebook will read it? Instead, post something full of vitriol that millions will see. Ghonim began to become dispirited. The tools that had brought the protesters together, he said, were now tearing them apart.

Ultimately, Egypt elected a government run by the Muslim Brotherhood, a traditionalist political machine that had played little part in the initial Tahrir Square groundswell. Then in 2013 the military staged a successful coup. Soon thereafter, Ghonim moved to California, where he tried to set up a social media platform that would favor reason over outrage. But it was too hard to peel users away from Twitter and Facebook, and the project didn’t last long. Egypt’s military government, meanwhile, recently passed a law that allows it to wipe its critics off social media.

Of course, it’s not just in Egypt and the Middle East that things have gone sour. In a remarkably short time, the exuberance surrounding the spread of liberalism and technology has turned into a crisis of faith in both. Overall, the number of liberal democracies in the world has been in steady decline for a decade. According to Freedom House, 71 countries last year saw declines in their political rights and freedoms; only 35 saw improvements.

While the crisis of democracy has many causes, social media platforms have come to seem like a prime culprit. The recent wave of anti­establishment politicians and nativist political movements—Donald Trump in the United States; Brexit in the UK; the resurgent right wing in Germany, Italy, or across Eastern Europe—has revealed not only a deep disenchantment with the global rules and institutions of Western democracy, but also an automated media landscape that rewards demagoguery with clicks. Political opinions have become more polarized, populations have become more tribal, and civic nationalism is disintegrating.

Which leaves us where we are now: Rather than cheering for the way social platforms spread democracy, we are busy assessing the extent to which they corrode it.

In China, government officials watched the Arab Spring with attentiveness and unease. Beijing already had the world’s most sophisticated internet control system, dynamically blocking a huge swath of foreign web domains, including Google. Now it garlanded its Great Firewall with even more barbed wire. China developed new ways to surgically turn off internet access in zones within cities, including a major block of downtown Beijing where it feared demonstrations. It also digitally walled off the entire province of Xinjiang after violent protests there that spread via the internet. Beijing may even have dabbled with creating a nationwide internet “kill switch.”

This bowdlerized version of the internet doesn’t sound at all like the original dream of the World Wide Web, but it has thrived nonetheless. By now, there are roughly 800 million people who surf the internet, exchange chat messages, and shop online behind the Great Firewall—nearly as many people as live in the United States and Europe combined. And for many Chinese, rising middle-class prosperity has made online censorship considerably easier to bear. Give me liberty, the line might go, or give me wealth.

China’s authoritarianism, which has doubled down under Xi’s leadership, certainly hasn’t hindered the Chinese tech industry. Over the past decade, China’s leading tech companies have come to dominate their home markets and compete globally. They’ve expanded through acquisitions in Southeast Asia. Baidu and Tencent have set up research centers in the US, and Huawei sells advanced networking equipment in Europe. The old silk road is being strung with Chinese fiber-optic cables and network equipment.

More than any other country, China has shown that, with a few adjustments, autocracy is quite compatible with the internet age. But those adjustments have caused the internet itself to start to break apart, like two continents cracking along a shelf. There’s the freewheeling, lightly regulated internet dominated by the geeks of Silicon Valley. And then there’s China’s authoritarian alternative, powered by massive, home-grown tech giants as innovative as their Western counterparts.

Today, China doesn’t just play defense against viral dissent by redacting troublesome parts of the internet; the government actively wields technology as a tool of control. In cities across China, including in Xinjiang, authorities are trying out facial-recognition software and other AI-powered technologies for security. In May, facial-recognition cameras at Jiaxing Sports Center Stadium in Zhejiang led to the arrest of a fugitive who was attending a concert. He had been wanted since 2015 for allegedly stealing more than $17,000 worth of potatoes. China’s Police Cloud System is built to monitor seven categories of people, including those who “undermine stability.” The country also aspires to build a system that will give every citizen and every company a social credit score: Imagine your FICO score adjusted to reflect your shopping habits, your driving record, and the appropriateness of your politics.

The fundamental force driving this change—this pivot from defense to offense—is a shift in how power flows from technology. In the beginning, the communications revolution made computers affordable to the masses. It wired devices together in a giant global network and shrank them down to the size of your hand. It was a revolution that empowered the individual—the lone programmer with the power to create in her pocket, the academic with infinite research at his fingertips, the dissident with a new and powerful way of organizing resistance.

Today’s stage of the digital revolution is different. That supercomputer in your pocket is also a homing device. It’s tracking your every “like,” keeping a record of everyone you talk to, everything you buy, everything you read, and everywhere you go. Your fridge, your thermostat, your smartwatch, and your car are increasingly sending your information back to headquarters too. In the future, security cameras will track the ways our eyes dilate, and sensors on the wall will track our body temperature.

In today’s digital world, in China and the West alike, power comes from controlling data, making sense of it, and using it to influence how people behave. That power will only grow as the next generation of mobile networks goes live. Remember how it felt like magic to be able to browse real web pages on the second-generation iPhone? That was 3G, the mobile standard that became widespread in the mid-2000s. Modern 4G networks are several times faster. 5G will be vastly faster still. And when we can do things faster, we do them more, which means data piles up.

It’s already hard for most people to comprehend, much less control, all the information collected about them. And the leverage that accrues to data aggregators will just increase as we move into the era of AI.

Vladimir Putin is a technological pioneer when it comes to cyberwarfare and disinformation. And he has an opinion about what happens next with AI: “The one who becomes the leader in this sphere will be the ruler of the world.”

In a way, Putin’s line is a bit overwrought. AI is not a hill that one nation can conquer or a hydrogen bomb that one country will develop first. Increasingly, AI is simply how computers work; it’s a broad term describing systems that learn from examples—or follow rules—to make independent decisions. Still, it’s easily the most important advance in computer science in a generation. Sundar Pichai, the CEO of Google, has compared it to the discovery of electricity or fire.

A country that strategically and smartly implements AI technologies throughout its workforce will likely grow faster, even as it deals with the disruptions that AI is likely to cause. Its cities will run more efficiently, as driverless cars and smart infrastructure cut congestion. Its largest businesses will have the best maps of consumer behavior. Its people will live longer, as AI revolutionizes the diagnosis and treatment of disease. And its military will project more power, as autonomous weapons replace soldiers on the battlefield and pilots in the skies, and as cybertroops wage digital warfare. “I can’t really think of any mission that doesn’t have the potential to be done better or faster if properly integrated with AI,” says Will Roper, an assistant secretary of the US Air Force.

And these benefits may compound with interest. So far, at least, AI appears to be a centralizing force, among companies and among nations. The more data you gather, the better the systems you can build; and better systems allow you to collect more data. “AI will become concentrated, because of the inputs required to pull it off. You need a lot of data and you need a lot of computing power,” says Tim Hwang, who leads the Harvard-MIT Ethics and Governance of AI Initiative.

China has two fundamental advantages over the US in building a robust AI infrastructure, and they’re both, generally, advantages that authoritarian states have over democratic ones. The first is the sheer scope of the data generated by Chinese tech giants. Think of how much data Facebook collects from its users and how that data powers the company’s algorithms; now consider that Tencent’s popular WeChat app is basically like Facebook, Twitter, and your online bank account all rolled into one. China has roughly three times as many mobile phone users as the US, and those phone users spend nearly 50 times as much via mobile payments. China is, as The Economist first put it, the Saudi Arabia of data. Data privacy protections are on the rise in China, but they are still weaker than those in the US and much weaker than those in Europe, allowing data aggregators a freer hand in what they can do with what they collect. And the government can access personal data for reasons of public or national security without the same legal constraints a democracy would face.

Of course, data isn’t everything: Any technological system depends on a whole stack of tools, from its software to its processors to the humans who curate noisy inputs and analyze results. And there are promising subfields of AI, such as reinforcement learning, that generate their own data from scratch, using lots of computing power. Still, China has a second big advantage as we move into the era of AI, and that’s the relationship between its largest companies and the state. In China, the private-sector companies at the cutting edge of AI innovation feel obliged to keep Xi’s priorities in mind. Under Xi, Communist Party committees within companies have expanded. Last November, China tapped Baidu, Alibaba, Tencent, and iFlytek, a Chinese voice-­recognition software company, as the inaugural members of its “AI National Team.” The message was clear: Go forth, invest, and the government will ensure that your breakthroughs have a market not just in China, but beyond.

During the original Cold War, the US relied on companies like Lockheed, Northrop, and Raytheon to develop cutting-edge strategic technology. Technically, these companies were privately owned. In practice, their vital defense mission made them quasipublic entities. (Indeed, long before the phrase “too big to fail” was ever used to describe a bank, it was applied to Lockheed.)

Fast forward to today, and the companies at the forefront of AI—Google, Facebook, Amazon, Apple, and Microsoft—don’t exactly wear flag pins on their lapels. This past spring, employees at Google demanded that the company pull out of a Pentagon collaboration called Project Maven. The idea was to use AI for image recognition in Defense Department missions. Ultimately, Google’s management caved. Defense Department officials were bitterly disappointed, especially given that ­Google has a number of partnerships with Chinese technology companies. “It is ironic to be working with Chinese companies as though that is not a direct channel to the Chinese military,” says former secretary of defense Ashton Carter, “and not to be willing to operate with the US military, which is far more transparent and which reflects the values of our society. We’re imperfect for sure, but we’re not a dictatorship.”

The Cold War wasn’t inevitable in 1945. The United States and Soviet Union had been allies during World War II, but then a series of choices and circumstances over a five-year period set the conflict on its self-perpetuating track. Similarly, as we can now see in the cold glare of hindsight, it was never inevitable that the digital revolution would inherently favor democracy. Nor is it inevitable today that AI will favor global authoritarianism to the permanent disadvantage of liberalism. If that scenario comes to pass, it will be because a series of choices and circumstances precipitated it.

In the original Cold War, two ideological foes created rival geopolitical blocs that were effectively non-interoperable. The US was boxed out of the Soviet bloc, and vice versa. The same could easily happen again, to disastrous effect. A new cold war that gradually isolates the Chinese and American tech sectors from each other would starve the US of much of the fuel it now relies on for innovation: American companies depend heavily on the Chinese market for their profits and for engineering and software talent. At the same time, it could actually create the kinds of dangers that hawks warn about now: It would increase the risk that one side could surprise the other with a decisive strategic breakthrough in AI or quantum computing.

Right now, maintaining a degree of openness with China is the best defense against the growth of a techno-­authoritarian bloc. That’s not the way American leaders are headed, though.

A little over six months after Donald Trump’s inauguration—and his invocation of “American carnage”—the administration launched a sweeping investigation into China’s trade practices and alleged theft of US technology via cyberspace. That investigation has mushroomed into a steadily escalating trade war, with the US launching tariffs on billions of dollars of Chinese goods and new investment and export restrictions on technologies that China considers key to AI and to its advanced manufacturing ambitions.

The confrontation is about much more than trade. The Trump administration has made it official US policy to protect the “national security innovation base”—White House shorthand for America’s leading technology and talent—from China and other foreign economic predators. In January, Axios published a leaked White House presentation that recommended the US work with its allies to build a 5G network that excludes China, to prevent Beijing from grabbing “the commanding heights of the information domain.” The presentation likened the 21st-century struggle for data dominance to the WWII-era race to construct an atom bomb. Then in April, the US Commerce Department hit ZTE, a leading Chinese telecommunications equipment firm that was gearing up to work on China’s 5G network, with a seven-year ban on doing business with US suppliers; the department said ZTE had violated the terms of a sanctions settlement. (The US later lifted the ban.)

For US security hawks, the prospect that China might dominate both 5G and AI is a nightmare scenario. At the same time, Washington’s escalating pushback against China’s tech ambitions has made Xi even more determined to wean his country off Western technology.

This is a very different philosophy from the one that has guided the technology sector for 30 years, which has favored deeply enmeshed hardware and software supply chains. Shortly before Trump’s inauguration, Jack Ma, the chair of Alibaba, pledged to create a million jobs in the United States. By September 2018, he was forced to admit that the offer was off the table, another casualty in the growing list of companies and projects that are now unthinkable.

Global work in AI has long taken place in three spheres: research departments, corporations, and the military. The first sphere has always been marked by openness and cooperation; to a lesser extent, so has the second. Academics freely share their work. Microsoft has trained many of China’s best AI researchers and helped nurture many promising AI startups, and Alibaba, Baidu, and Tencent employ US engineers at their research hubs in Silicon Valley and Seattle. An AI-driven breakthrough in Shanghai—say, in diagnosing disease through more accurate scans of medical images—can save lives in Shawnee. But national security concerns have a way of overriding commercial considerations. For now, the political momentum appears to be driving the two countries’ tech sectors apart to such a degree that even collaboration between researchers and corporations could be stifled. The schism could well define how the struggle between democracy and authoritarianism plays out.

Imagine it’s 2022: America’s confrontational economic policies have continued, and China has refused to yield. Huawei and ZTE have been banned from the networks of the US and key Western allies. Through investment and theft, Beijing has reduced its reliance on US semiconductors. Rival tech superpowers have failed to develop common standards. US and Chinese academics increasingly deposit their cutting-edge AI research in government safes instead of sharing it at international conferences. Other countries—like France and Russia—have tried to build homegrown technology industries centered on AI, but they lag far behind.

The world’s nations can commit to American technology: buying Apple phones, using Google search, driving Teslas, and managing a fleet of personal robots made by a startup in Seattle. Or they can commit to China: using the equivalents built by Alibaba and Tencent, connecting through the 5G network constructed by Huawei and ZTE, and driving autonomous cars built by Baidu. The choice is a fraught one. If you are a poor country that lacks the capacity to build your own data network, you’re going to feel loyalty to whoever helps lay the pipes at low cost. It will all seem uncomfortably close to the arms and security pacts that defined the Cold War.

And we may be seeing the first evidence of this. In May 2018, about six months after Zimbabwe finally got rid of the despot Robert Mugabe, the new government announced that it was partnering with a Chinese company called CloudWalk to build an AI and facial-recognition system. Zimbabwe gets to expand its surveillance state. China gets money, influence, and data. In July, nearly 700 dignitaries from China and Pakistan gathered in Islamabad to celebrate the completion of the Pak-China Optical Fibre Cable, a 500-mile-long data line connecting the two countries through the Karakoram Mountains, built by Huawei and financed with a loan from China’s Export-Import Bank. Documents obtained by Pakistan’s Dawn newspaper revealed a future plan for high-speed fiber to help wire up cities across Pakistan with surveillance cameras and vehicle-monitoring systems, part of a “Safe Cities” initiative launched in 2016 with help from Huawei and other Chinese firms. China has effectively constructed its own Marshall Plan, one that may, in some cases, build surveillance states instead of democracies.

It’s not hard to see the appeal for much of the world of hitching their future to China. Today, as the West grapples with stagnant wage growth and declining trust in core institutions, more Chinese people live in cities, work in middle-class jobs, drive cars, and take vacations than ever before. China’s plans for a tech-driven, privacy-invading social credit system may sound dystopian to Western ears, but it hasn’t raised much protest there. In a recent survey by the public relations consultancy Edelman, 84 percent of Chinese respondents said they had trust in their government. In the US, only a third of people felt that way.

No one can be certain what happens next. In the US, in the wake of controversies surrounding the 2016 election and user privacy, a growing number of Republicans and Democrats want to regulate America’s tech giants and rein them in. At the same time, China has stiffened its resolve to become an AI superpower and export its techno-­authoritarian revolution—which means the US has a vital national interest in ensuring that its tech firms remain world leaders. For now, there is nothing close to a serious debate about how to address this dilemma.

As for China, it remains unclear how much digital intrusion people there will tolerate in the name of efficiency and social cohesion—to say nothing of people in other countries that are tempted by Beijing’s model. Regimes that ask people to trade freedom for stability tend to invite dissent. And Chinese growth is slowing. For the past century, democracies have proven more resilient and successful than dictatorships, even if democracies, particularly in an age of algorithms, have made some stupid decisions along the way.

It is at least conceivable that Trump’s aggressive policies could, counterintuitively, lead to a rapprochement with Beijing. If Trump threatens to take something off the table that China truly cannot afford to lose, that could pressure Beijing to dial back its global tech ambitions and open its domestic market to US firms. But there is another way to influence China, one more likely to succeed: The US could try to wrap Beijing in a technology embrace. Work with China to develop rules and norms for the development of AI. Establish international standards to ensure that the algorithms governing people’s lives and livelihoods are transparent and accountable. Both countries could, as Tim Hwang suggests, commit to developing more shared, open databases for researchers.

But for now, at least, conflicting goals, mutual suspicion, and a growing conviction that AI and other advanced technologies are a winner-take-all game are pushing the two countries’ tech sectors further apart. A permanent cleavage will come at a steep cost and will only give techno-authoritarianism more room to grow.

Nicholas Thompson (@nxthompson) is editor in chief of WIRED. Ian Bremmer(@­ianbremmer) is a political scientist and president of the Eurasia Group.