Goldman identified four key areas where development is needed to create value in AI: talent, data, infrastructure and computing power. The bank concluded China has the talent, data and infrastructure needed to fully embrace AI.
Because AI is a relatively new technology, finding adequate number of talented individuals is a perennial problem. Experts have argued that more needs to be done to train people in new AI-related skills.
To get around talent scarcity in any particular location, U.S. tech giants are opening research labs around the world, according to Goldman. Chinese companies are also following their lead by opening Silicon Valley research labs and offering comparable salaries, Goldman said.
Earlier this year, Baidu snagged Microsoft executive Qi Lu as part of a push into AI. Meanwhile, Tencent tapped up former Microsoft scientist Yu Dong to head up its AI research facility in Seattle.
China’s vast population, much of which is connected to the internet, gives the country an advantage in generating data. Moreover, China’s large internet companies have comprehensive online ecosystems increasingly penetrating more of the daily lives of the country’s internet users, generating volumes of data, according to Goldman.
“China understandably generates (about) 13 percent of the digital information globally. By 2020, we expect this to grow to around 20 percent to 25 percent as China’s economy emerges as the world’s largest,” the bank said. It predicted China would generate about 9 to 10 zettabytes of data; one zettabye is about 1 trillion gigabytes.
When it comes to infrastructure, most major companies involved in AI research have adopted open-sourced platforms to attract resources and talent into their ecosystems.
Chinese companies are also following the trend, said Goldman. For example, Baidu has an open-sourced machine-learning platform called PaddlePaddle that stands for Parallel Distributed Deep Learning. Baidu also announced project Apollo, another open-sourced platform to develop autonomous driving.
AI algorithms and their performances are also limited by computing power that depends on the processing unit. Goldman noted that China had been “heavily dependent on foreign suppliers” for processing chips, but there was some “encouraging progress” in its domestic semiconductor industry.
The bank said that it expected China’s dependency on foreign suppliers to decrease over time.
