For years, Kai-Fu Lee would share his venture capital firm’s office space in Beijing with startups he was backing so they could get advice and collaborate with each other. But about three months ago, the last of the startups packed up and moved out. The desks were promptly taken over by employees of Sinovation Ventures’ latest growing endeavor, an in-house AI Institute.
“It makes more sense to have our AI team together in the office – there’s more synergy,” says Lee, a veteran of both Google and Microsoft Corp. in China, during an interview at his headquarters in the city’s Haidian District. The AI Institute now has about 30 full-time employees with plans to grow headcount to about a hundred within the year.
China’s government has declared its ambition to lead the world in artificial intelligence by 2030. That ambition, set forth in state plan issued in July, creates a hospitable environment for AI entrepreneurs and investors. Lee’s first aim is not world domination, but more pragmatic. About 80 percent of his shop’s funding and attention is now focused on achieving practical business efficiencies, with the other 20 percent lavished on potentially more disruptive if speculative AI-related technologies.
While flashy AI-related projects like driverless cars tend to grab the most headlines, Lee says he’s first focused on “the low-hanging fruit” of improving the back-ends of many business and industrial operations. The trick is convincing China’s lumbering state-run industrial giants, not typically seen as on the leading technology edge, that their operations can benefit significantly from AI.
“We’re first taking what’s known, what’s been proven in the academic research, and applying it to areas where huge amounts of data exist and are accurately and objectively labeled,” says Lee. “Every bank, insurance company, every loan company, every e-commerce site has longitudinal customer data.” Tailored algorithms can help loan officers improve risk assessment, help banks detect fraudulent credit-card activities, and help sales agents narrow down lists of potential customers “to determine which ones you’ll apply tele-sales prospects.”
One of the firm’s investments is a startup called SmartFinance, which uses AI to help determine how to make loans over the internet. The company’s app uses as many as 1,200 data points collected by its smartphone app to assign credit ratings for would-be customers.
While China’s leading internet companies – Baidu, Alibaba, Tencent, and others – all have significant in-house AI teams, Lee recognizes that state-run insurance companies, investment firms, logistics managers, and others won’t necessarily have the reputation, swanky offices, or high salaries to attract top-notch AI scientists, whose numbers remain scarce in China. So he’s configuring his AI Institute to provide solutions or services directly, like a high-tech consultancy.
“We’re building an AI solutions team to talk to big businesses” about designing customized recommendation engines, risk management tools, etc., explains Lee.
The goal is to “design reusable AI toolkits,” says John Woo, a vice president at the AI Institute. He and Lee expect the financial industry in China will be among the first to be radically transformed by the application of AI. “For picking stocks, the future of mutual funds will all be made this way,” by algorithms, says Lee. “Today you have all these Morningstar ratings of funds – but A.I. will replace all these one day.”
On the August morning that Bloomberg News visited Sinovation Venture’s offices in northern Beijing, about three dozen students from universities in the U.S. and China were putting final tweaks on their summer camp projects. The inaugural class of DeeCamp students – ranging from undergrads to PhD candidates – had been selected from more than 1,000 applicants, says Woo. The objective is to get more students interested in AI, especially in China, and perhaps identify Sinovation’s future interns or employees.
One student team designed a buggy out of parts ordered on Amazon and Taobao, which uses a spinning lidar sensor mounted on top to develop a map of the office layout. It recognizes ceiling pipes and doorways, although it has more trouble recognizing feet and other obstacles directly in front of it. Still, for about five weeks labor, it’s an impressive effort. “It reminds me of a small child,” says Woo, who has two children. “It is still teaching itself about the world.’’
Lee set up the summer camp because of the talent shortage in the industry. He anticipates bringing in such student groups multiple times a year in the future, and will probably end up hiring some because he needs the brainpower.“AI is the most important part of our investment portfolio,’’ he says.