If you get into a car accident in China in the near future, you'll be able to pull out your smartphone, take a photo, and file an insurance claim with an AI system.
That system, from Ant Financial, will automatically decide how serious the ding was and process the claim accordingly with an insurer. It shows how the company—which already operates a hugely successful smartphone payments business in China—aims to upend many areas of personal finance using machine learning and AI.
The e-commerce giant Alibaba created Ant in 2014 to operate Alipay, a ubiquitous mobile payments service in China. If you have visited the country in recent years, then you have probably seen people paying for meals, taxi rides, and a whole lot more by scanning a code with the Alipay app. The system is far more popular than the wireless payments systems offered in the U.S. by Apple, Google, and others. The company boasts more than 450 million active users compared to about 12 million for Apple Pay.
Ant’s progress will be significant to the future of the financial industry beyond China, including in the U.S., where the company is expanding its interests. The company’s approach goes around existing institutions to target individuals and small businesses who lack access to conventional financial services. Ant said in April of this year that it is buying the U.S. money-transfer service MoneyGram for $880 million. The deal is subject to regulatory approval and should close in the second half of this year. The company could well apply the technologies it is developing to its overseas subsidiaries. A spokesperson for the company says it hasn’t brought Alipay to the U.S. because existing financial systems provide less of an opportunity.
Yuan (Alan) Qi, a vice president and chief data scientist at Ant, says the company’s AI research is shaping its growth. “AI is being used in almost every corner of Ant’s business,” he says. “We use it to optimize the business, and to generate new products.”
The accident-processing system is a good example of how advances in AI can flip an existing system on its head, Qi says. It has become possible to automate this kind of image processing in recent years using a machine-learning technology known as deep learning. By feeding thousands of example images into a very large neural network, it is possible to train it to recognize things that even a human may struggle to spot (see “10 Breakthrough Technologies 2013: Deep Learning”).
“We use computer vision for a job that is boring but also difficult,” Qi says. “I looked at the images myself, and I found it pretty difficult to tell the damage level.”
Qi speaks a mile a minute, which seems appropriate given how quickly his company seems to be moving. Dressed in a smart shirt and dress pants on a sweltering afternoon in Beijing this May, shortly after giving a speech at a major AI conference, Qi explained that the company considers itself not a “fintech” business but a “techfin” one, due to the importance of technology.
Ant already operates a range of other financial services besides Alipay. For instance, it provides small loans to those without a bank account. It assesses a person’s creditworthiness based on his or her spending history and other data including friends' credit scores (see “Alipay Leads a Financial Revolution in China”).
Ant’s creditworthiness system also provides a high-tech way to obtain various services, such as hotel bookings, without a deposit. Qi says that Ant uses advanced machine-learning algorithms and custom programmable chips to crunch huge quantities of user data in a few seconds, to determine whether to grant a customer a loan, for instance.
A recent hire offers some measure of Ant’s intent to apply artificial intelligence to finance. This May the company announced that Michael Jordan, a professor at the University of Berkeley and a major figure in the field of machine learning and statistics, would become chair of the company’s scientific board.
Qi is no slouch, either. He got his PhD from MIT and became a professor in the computer science department at Purdue before joining Alibaba in 2014. Once there, he developed Alibaba’s first voice-recognition system for automating customer calls.
“We built a system, based on deep learning, to carry on conversations; to provide answers to your questions,” Qi says. This chatbot system also taps into a knowledge base of information created by Ant, and is an example of how researchers are increasingly combining cutting-edge machine-learning techniques with conventional representations of knowledge. “Human language is still very hard for a machine to understand,” Qi says.
In March this year, the chatbot system surpassed human performance in terms of customer satisfaction, says Qi. “There are many, many chatbot companies in Silicon Valley. We are the only one that can say, confidently, they do better than human beings,” he says.
Ant’s success to date has certainly been impressive. Credit Suisse estimates that it manages 58 percent of mobile payments in China. A key competitor has emerged in recent years with WeixinPay, from the mobile chat giant Tencent, now accounting for almost 40 percent of the market. Ant remains enormously valuable, though. Earlier this year, a Hong Kong investment group valued the company at $75 billion. The company was expected make an initial public offering this year, but that now looks more likely to happen in 2018.
Ant is also increasingly looking to expand its interests overseas. The company has invested almost $1 billion in Paytm, an Indian payments company. It has also invested in Ascend, a Thai online payments business, and M-Daq, a Singaporean financial business. Ant apparently also sees investments and acquisitions as a way to bolster its technological prowess. Last year the company acquired EyeVerify, a U.S. company that makes eye recognition software.
Why Meta’s latest large language model survived only three days online
Galactica was supposed to help scientists. Instead, it mindlessly spat out biased and incorrect nonsense.
DeepMind’s game-playing AI has beaten a 50-year-old record in computer science
The new version of AlphaZero discovered a faster way to do matrix multiplication, a core problem in computing that affects thousands of everyday computer tasks.
A bot that watched 70,000 hours of Minecraft could unlock AI’s next big thing
Online videos are a vast and untapped source of training data—and OpenAI says it has a new way to use it.
The White House just unveiled a new AI Bill of Rights
It's the first big step to hold AI to account.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.