As internet giants all over the world herald their advances in AI, one company has been conspicuously absent—Tencent, the Chinese social media giant, now worth over $300 billion. But the company’s relative silence in matters of artificial intelligence looks set to change quickly.
On May 2 Tencent confirmed that it would open an laboratory in Seattle dedicated solely to artificial intelligence research, to be headed by Yu Dong, a former scientist at Microsoft Research. The news is the latest in a steady stream of announcements indicating the company is making a foray into artificial intelligence. In March the company announced that it had poached Zhang Tong, former head of Baidu’s Big Data Lab in Beijing. And in December 2016, the company confirmed it had established an AI lab in Shenzhen, the company’s home city, earlier that year.
Tencent, at least insofar as it has discussed its efforts publicly, has arrived to AI relatively late. Baidu, China’s Google analog, established its first US-based AI lab in 2013, and has two more in China. Like its American counterpart, the company has poured resources into researching artificial intelligence, image recognition, and self-driving cars.
In the US, Google and Apple have deployed voice assistants on iPhones and Android devices for years. Facebook has been publishing papers on AI since 2014, led by some of the field’s top researchers. Amazon has perhaps enjoyed the most tangible success in the industry, having sold an estimated 11 million Amazon Echo home assistants to date.
But there’s reason to believe that Tencent—which remains barely known outside China—will become the field’s sleeping giant. That’s because its flagship product, WeChat, is uniquely positioned to absorb users’ data and devise algorithms from it.
“Basically, anything you need to do online can be done through WeChat,” says Dr. Andy Chun, a leading AI expert and Associate Professor at City University of Hong Kong. “WeChat is much more ingrained into the average Chinese citizen’s daily life than Alibaba or Baidu. Amazon and Google do not have anything comparable.”
WeChat is China’s most popular chat app, with over 889 million users as of the end of 2016. Its users spend a staggering amount of time in the app—50% of users spend 90 minutes per day inside the app, according to the company. Facebook hasn’t disclosed perfectly analogous data, but in April 2016 it said the average user spends 1 hour per day inside its suite of apps (not including WhatsApp).
A core component of artificial intelligence research lies in teaching machines how to communicate with humans. WeChat is the perfect conduit for this type of research. Since it’s a chat app, WeChat is collecting countless messages each day that real people send to one another.
“Conversational data is richer [than search or shopping data] because you have that many people on the platform performing that many interactions,” says Tak Lo, who manages the Hong Kong-based Artificial Intelligence startup accelerator Zeroth.ai. “Shopping and search engine data show one type of data, which is purchasing or shopping intent, which is valuable, but different. For example, if you were to build out an Natural Language Understanding engine, you would not use search engine data, because no one searches based on complete conversational phrases.”
WeChat has an extra advantage in that a significant portion of its users communicate using voice. Push-to-talk, an neglected feature in most countries, has become a preferred method of chat for many users on WeChat. That means Tencent can draw not just from text, but from recorded speech too.
Like Facebook, WeChat has evolved into one of China’s most popular apps for reading, writing, and sharing blog posts. For years it has let outlets publish directly to WeChat using special URLs, well before Facebook wooed media companies into posting to Instant Articles. It also calculates likes and shares for each article. The app’s influence has fostered a cottage industry of one-person media outlets in China, many of which remain dependent on WeChat for readers’ attention (and in some cases, their livelihood).
Tencent is not a leader in online maps in China, where Baidu and Alibaba (the latter thanks to an acquisition) dominate. According to market research firm iResearch, the company’s standalone mapping app has only 8 million monthly users, compared with Baidu and Alibaba’s 200 million (paywall) on their respective apps. But Tencent is likely getting plenty of data from a popular WeChat feature that lets users share their locations with one another, and then helps these people meet in person using GPS tracking.
The company has also made a few bets in transportation that could pay off later. It’s a major investor in the Uber analog Didi Chuxing, which itself is researching AI and machine learning. It also purchased a stake in HERE (paywall), the mapping unit formerly owned by Nokia, and Tesla, the popular electric vehicle company which itself is looking to increase its presence in China. As long all of these deals include some form of data sharing, Tencent will have a better sense not just of how humans move, but how vehicles move.
This is what makes WeChat different from Facebook and Facebook Messenger, its two closest analogs. The company launched a payments service in the fall of 2013, which quickly grew popular the following year thanks to a viral gifting scheme released during Chinese New Year. While Alibaba’s Taobao and Tmall remain the most popular web destinations for shopping, WeChat has turned into a shopping mall in its own right. Small business owners have opened up online stores there, Alibaba rival JD moved its e-commerce site there (paywall), and brick-and-mortar retailers have adopted WeChat’s payment system as an alternative to cash and credit cards. These forays into shopping have paid off—Tencent now occupies 37% of China’s mobile payment industry.
All of this data can be funneled into a range of different services, Dr. Chun says. WeChat can predict market trends by observing purchasing habits. It can improve the personalized financial services that Tencent already offers to consumers (paywall). It can even predict financial trends and flows.
There are few companies out there with a single app that is capable of tracking such a wide range of data. Meanwhile, Tencent’s domestic competitors remain confined to a relatively narrow field when it comes to their core businesses (Baidu in search, Alibaba in e-commerce), while Tencent’s foreign competitors remain largely absent in China for either legal or competitive reasons. Even if the company remains a latecomer to AI, its newly-poached researchers, talented product engineers, and troves of cash can help it catch up quickly.