GBA Brief wants to win the robo taxi race in Guangzhou – and in California too

Special zone Nansha is trying to establish itself as a testing ground for new ideas in artificial intelligence. Local start-up – already a unicorn – is one of its pioneering companies in driverless cars.

As the trade row between Washington and Beijing enters its second year, the tech world has been drawn directly into the conflict. Talk of a new cold war is going too far but the rivalry between the two superpowers is getting more confrontational, with chatter about an ‘arms race’ in artificial intelligence as well.

Perhaps this was one reason why the Chinese press has picked up on news that Lee Kai-fu’s Sinovation Ventures has just set up a new subsidiary in Guangzhou to invest in AI businesses in the Greater Bay Area.

Lee is a high-profile investor because he was formerly president of Google China and before that an executive at Microsoft and Apple. That gives him a foot in both camps of what he describes as “the great AI duopoly”, a topic discussed in a book he published last year. 

The one-sentence summary is that Lee thinks the Americans are ahead in the raw science behind AI but that the Chinese are set to take the lead in how the technology is applied in real life. 

He also believes that China has key advantages in the sheer volume of data that it is generating. The bigger the dataset fed into the algorithms, the better they will perform. Having a monopoly on the best and brightest scientists loses some if its edge, he warns.

In testing out applications for artificial intelligence, Chinese firms have also been getting a helping hand from local governments that want to seed hi-tech clusters on their home patch. 

A good example is Nansha, a special zone on the outskirts of Guangzhou, where there are already more than a hundred 100 AI firms experimenting in applications from voice recognition to traffic management. 

Another area of focus is autonomous driving, where local unicorn is testing out a fleet of robo taxis. Founded in 2016 by former executives from search engine giant Baidu, moved to Nansha last year, where it is trialling a programme in which riders hail cars at predetermined points with their smartphones. 

Journeys are navigated through its PonyAlpha network of sensors, lidars, radars and cameras (footage of previous testing shows cars moving at speed through town).

Sinovation’s Lee says that the Americans are some distance ahead in ‘autonomous AI’ (the type of research that steers self-driving cars). But he thinks that Chinese companies will catch up with support from the state, including help with transport corridors that are tailor-made for next-generation vehicles.

Along similar lines,’s chief executive James Peng told reporters that the trials in Nansha have been crucial in generating data that improves how vehicles respond to a wider range of scenarios. 

All of this fuels the narrative about which of the two superpowers will win the race for AI riches, although Lee’s view is that differences in business models and customer habits are going to generate two very distinct ecosystems. 

“It is very difficult, if not impossible, for any American company to try to enter China’s market or vice versa, “ he told the Washington Post last year. “It’s like two different jigsaw puzzles. You can’t take a piece from one and try to fit it into the other – everything is different.”

For the moment, at least, there is a two-track approach. Like local competitor AutoX, which set up an R&D centre in Shenzhen this year, is running a parallel operation in California, where both companies have just got the green light for passenger-carrying trials on public roads. came fifth in California’s ‘miles per disengagement’ rankings earlier this year, or the frequency in which human drivers are forced to take control of the self-driving vehicle. At 1,022 miles per disengagement, its score was higher than its Chinese rivals Baidu (206), AutoX (191) and WeRide (173). But it was still some way back on Alphabet subsidiary Waymo, which reported one disengagement every 11,017 miles. will be planning to close that gap – and it will be hoping that it isn’t closed out of the American market as the AI arms race picks up pace.