A.I. in China: TikTok is just the beginning
This article is part of a Fortune Special Report on Artificial Intelligence.
Andy Warhol was wrong. The pop artist famously asserted that in the future everyone would be famous for 15 minutes. But what Warhol failed to imagine back in the 1960s was TikTok, the dangerously addictive video-sharing app that today doles out global renown in 15-second increments.
With some 1.5 billion downloads worldwide—including 124 million in the U.S. as of November—TikTok is racing ahead in the battle for eyeballs, where its short clips compete with the likes of Facebook, Instagram, Twitter, and YouTube. In the world of TikTok, a group of tweens doing a choreographed dance routine can go viral overnight; a dancing ferret can garner millions of likes; or a clip of a man shaking his groove thing in a plushy rabbit costume might turn into a meme. (There’s a lot of dancing on TikTok.)
While the content that powers TikTok is created by the app’s obsessive user base, the company’s A.I.-powered recommendation algorithms—which customize picks for users with increasingly uncanny insight—are what makes the platform so truly enthralling. That secret sauce is provided by TikTok’s parent company, Beijing Bytedance. And thanks to TikTok’s runaway success, Bytedance has emerged as a somewhat unlikely player in another global contest: the A.I. Arms Race.
The privately held Bytedance was founded in 2012 by former Microsoft engineer Zhang Yiming. Six years later, a $3 billion fundraising round led by SoftBank valued Bytedance at $75 billion, ranking it among the world’s most valuable startups. That was before TikTok crossed the billion-user threshold.
Given its prowess in a hot technology like artificial intelligence and its wild popularity outside China, Bytedance should, in theory, be a darling of policymakers at home. Instead, it often appears to be out of favor with the government. The company has been made to apologize several times in recent years owing to content shared over its various apps. (In addition to TikTok, Bytedance owns other video-sharing platforms and a news-aggregating app called Toutiao that utilizes A.I. in a way similar to TikTok.)
There are other reasons that Bytedance might not have been Beijing’s first choice to carry the banner for Chinese A.I. The central government intends for artificial intelligence to become a long-term driver of economic supremacy, a means for social governance, and a key tool in developing greater military strength. In short, Beijing has a plan for what A.I. development should look like in China—and frivolous videos are not a core part of it.
But if Bytedance is in some ways an outlier in the fast-growing Chinese A.I. ecosystem, in other ways it is a model for the sort of unpredictable success that’s likely to be replicated given the vast sums of public and private capital flowing into both primary A.I. research and A.I.-based startups in China.
“The industry is definitely hot right now, and I don’t think it will abate,” says Henrik Bork, managing partner of Beijing-based consultancy Asia Waypoint. “We are only just beginning to see A.I. getting started in China.”
It was a board game that helped kick off the current A.I. boom in China. In 2016, Lee Sedol, a South Korean master in the ancient Chinese game of Go, played a best-of-five-game competition against an A.I. system developed by DeepMind, an A.I. startup owned by Google parent company Alphabet. The object of Go is to surround more territory than your opponent on a board of 324 squares by laying down pebble-like pieces, and it’s considered to be exponentially more complex than chess. Some 280 million people in China tuned in to watch Lee’s televised showdown with the algorithm, called AlphaGo. Lee was confident he would beat the algorithm. He lost all but one game.
For policymakers in Beijing, witnessing an A.I.-powered Western computer defeat a human in the ultimate Chinese game was a wake-up call—described by many as a “Sputnik moment” for A.I. in China. According to Luciano Floridi, the director of the Digital Ethics Lab at Oxford University, before 2017 A.I. was treated by Beijing as just one among many technologies in which China should improve its expertise. That soon changed.
The following year Beijing released the Next Generation Artificial Intelligence Development Plan (AIDP). The pivotal policy document singled A.I. out as a strategic technology and set a target for China to become the world leader in A.I. by 2030, with A.I. contributing $150 billion to Chinese GDP.
Then Beijing did what it typically does when it wants to kick-start a given industry: It engaged in a bit of central planning.
In 2017, the central government selected five leading tech companies as “national champions” in A.I., instructing each to pursue a specific avenue of A.I. research. In return for picking up the mantle, the champions receive government support, such as access to finance, preferential contract bidding, and sometimes even market share protection. China’s leading search engine provider, Baidu, was tasked with developing autonomous driving, for example. E-commerce giant Alibaba, meanwhile, has been charged with advancing technology needed to build smart cities. In the past few years, the number of A.I. champions has risen to at least 15.
The pairings between champion and designated area of expertise aren’t always as random as they might seem. Even before the AIDP was released, Baidu was investing heavily in autonomous vehicles (AVs). The year before that, Alibaba launched its smart-city software service, City Brain, which uses big-data analysis to assist with urban planning and management.
Despite the big policy push from Beijing, China’s approach to A.I. is only semi-centrally planned. Plenty of the heavy lifting is being left to local governments. Some $15 billion of new funding for A.I. has been pledged by Shanghai. Tianjin, another coastal metropolis, is raising $16 billion in funds for A.I. And the city of Beijing is currently building a $2.2 billion industrial park for A.I. research.
“There’s a lot of bottom-up,” says Jeffrey Towson, a private equity investor and former professor at Peking University. “There are a lot of entrepreneurs, lots of venture capital, and half the time the government is trying to figure out what’s going on.”
Just as AlphaGo’s victory caught the attention of Chinese officials, the dollar amounts being invested by the Chinese to pursue their A.I. goals have set off alarm bells in the U.S. amid concern that Washington isn’t investing enough in the technology. In November, for instance, Congress’s National Security Commission on Artificial Intelligence, chaired by former Google CEO Eric Schmidt, published an interim report in which it warned that China is outpacing the U.S. in A.I. spending. That was followed by a report from the Center for a New American Security urging the U.S. government to fund $25 billion of R&D in A.I. At a debate in December, Democratic presidential candidate Andrew Yang warned that China is “leapfrogging us in A.I. because they have more data than we do and their government is subsidizing it to the tune of tens of billions of dollars.”
The Trump administration, meanwhile, has imposed export controls on several categories of A.I. technology in an attempt to curb the flow of key A.I. intellectual property to China. But the Chinese are making gains in the research arena as well. Judging by the number of A.I. patent filings and published academic papers on A.I., China is now neck and neck with the U.S. (See graphic below.)
China’s rapid progress in A.I. has done much to even the playing field with the West. But in many fundamental ways China still has ground to make up. Jeffrey Ding, a researcher at Oxford’s Future of Humanity Institute who studies China’s A.I. strategy, recently conducted a systematic comparison of the U.S. and China A.I. ecosystems. “My conclusion was that the U.S. is still far ahead,” says Ding.
One area in which the U.S. has a clear lead is in hardware. China currently consumes 58% of the world’s semiconductor supply. But according to the Center for Strategic and International Studies (CSIS), only 16% of chips used in China are produced within the country, and only half of those are made by Chinese companies. Beijing has a plan to meet 70% of its domestic demand by 2025 and has pledged more than $100 billion to that end. However, China’s leading chipmaker, Semiconductor Manufacturing Industry Corp., still lags far behind its Western rivals. SMIC began producing 14-nanometer chips only last year; industry leaders have all moved on to 7-nanometer chipsets. (In chips, smaller is better.) “China still depends heavily on Western technology,” says Lian Jye Su, principal analyst at ABI Research in Singapore. “That explains the sense of urgency from Beijing to accelerate China’s R&D into these sectors.”
One area in which the U.S. and China have begun to square off: the war for talent. In November, a Senate Committee on Homeland Security and Governmental Affairs labeled China’s talent acquisition strategies—the committee counts some 200 plans—a threat to national security, claiming that “China unfairly uses the American research and expertise it obtains for its own economic and military gain.” The report notes that the U.S. lacks a “comprehensive strategy to combat this threat” of brain drain.
One of the factors most often cited when discussing China’s advantages in developing artificial intelligence is its massive treasure trove of data. A.I. algorithms feed on data in order to learn. And it’s certainly true that, for the right company, data is available in abundance in China. That’s not only because of the country’s roughly 1.4 billion people, generating multitudes of data points, but also because of a, shall we say, cooperative regulatory environment. And China has a massive video surveillance system with millions of A.I.-enabled cameras.
In this setting, the development of facial recognition technology has thrived. China is home to two of the world’s most valuable computer-vision companies—Hong Kong–based startup SenseTime, which is valued at more than $7 billion, and Beijing’s Megvii, which is valued at just over $4 billion and plans a Hong Kong IPO this year. Megvii operates the world’s largest open source database for training other facial recognition algorithms, Face++, which has over 300,000 users. It has reportedly used government data banks to help compile its training program.
China has faced criticism for its use of facial recognition to track citizens—particularly the Muslim Uighur population—and its experiments with a “social credit system” that would incorporate facial recognition to monitor behavior. As of December, anyone purchasing a new SIM card in China is required by law to submit to a facial recognition scan in order to confirm their identity.
But a massive library of its citizens’ faces alone doesn’t assure China of a long-term advantage in A.I. more broadly. “I think a lot of the arguments about how China’s data advantage means it’s going to be ahead on deploying A.I. applications are overstated,” says Oxford’s Ding. “More facial recognition data, for instance, doesn’t benefit other verticals, like smart manufacturing, precision medicine, or autonomous vehicles.”
Given all the anxiety in the West about China’s A.I. ambition, even TikTok has come under scrutiny. In November, the Committee on Foreign Investment in the United States (CFIUS) opened an investigation into whether TikTok posed a national security threat, prompted by reports alleging that TikTok shares user data with the Chinese government. Bytedance, which declined to be interviewed for this article, has denied the accusations and says its U.S. data is stored on servers in either the U.S. or Singapore.
But for most of TikTok’s billion-plus users, such concerns aren’t on their radar. They are focused on chasing fame, 15 seconds at a time.
National champions of A.I.
When China proposes a national policy for industrial advancement, it doesn’t leave success to chance. Instead, the government selects leading enterprises—state-owned or otherwise—and designates them “national champions.” Those selected are expected to work toward the government’s goals and are often rewarded with preferential policy treatment and easier access to financing. In 2017, Beijing selected five companies to spearhead its A.I. program. That number has grown to at least 15. Here are five of the most important:
1. Champion: Baidu
Arena: Autonomous driving
Stats: “China’s Google” launched an open source platform for autonomous driving software, called Apollo, in 2017. Baidu stylizes Apollo as the equivalent to Google’s Android operating system, but for autonomous vehicles (AVs).
2. Champion: Tencent
Arena: Computer vision for medical diagnoses
Stats: The operator of the WeChat super-app has been an active investor in digital health care since at least 2014. Its Medical A.I. Lab is currently working on a computer vision system designed to diagnose Parkinson’s disease.
3. Champion: Huawei Technologies
Arena: A.I. infrastructure and software
Stats: Huawei, the global telecom equipment giant that has been embroiled in the U.S. vs. China trade battle, was added to China’s roster of A.I. national champions last September. It is tasked with boosting “device, edge, and cloud” capacity.
4. Champion: Alibaba Group
Arena: Smart city initiatives
Stats: China’s largest e-commerce conglomerate has pursued its designated mission by developing the City Brain initiative, which uses cloud computing and big data to improve urban planning and is deployed in 23 cities including Kuala Lumpur.
5. Champion: SenseTime
Arena: Computer vision
Stats: SenseTime began as an academic project at the Chinese University of Hong Kong in 2014 and is now a leader in the field of computer vision with a valuation of over $7 billion. Its technology is used in some government surveillance projects.
A version of this article appears in the February 2020 issue of Fortune with the headline “TikTok: China Starts the Clock.”
More from Fortune’s special report on A.I.:
—Inside big tech’s quest for human-level A.I.
—A.I. breakthroughs in natural-language processing are big for business
—Facebook wants better A.I. tools. But superintelligent systems? Not so much.
—A.I. is transforming HR departments. Is that a good thing?
—Medicine by machine: Is A.I. the cure for the world’s ailing drug industry?
Subscribe to Eye on A.I., Fortune’s newsletter covering artificial intelligence and business.