Google is already at it, as are Uber, Tesla and several other world-class technology companies. They’re all testing and deploying self-driving cars on public roads — what promises to be the next frontier in the ultra-competitive auto industry.
But in China, those companies are unlikely to make a dent. It’s Baidu that may end up having a bigger say in an auto-driving market — potentially the world’s largest — that may be worth more than $100 billion by 2025 — thanks to its algorithm, commercial partnerships and heavy state support that push out western competitors, analysts say.
The search giant, known as China’s Google, has been trying to turn its artificial intelligence (AI) research into money-spinning ideas, as its core online advertising revenue is still reeling from a 2016 scandal that led to a regulatory crackdown on its most profitable healthcare ads. Autonomous driving presents the brightest prospect: Baidu’s other AI-based initiatives such as its DuerOS voice assistant and automated news feed face stiff local competition, as many firms have the tech prowess and supporting services to train robots to play music, place online orders and study people’s reading habits.
According to research firm Canalys, Shenzhen-based iFlytek is now the leader in China’s crowded field of smart voice assistants, beating Baidu and fellow tech-titan Alibaba in the total number of mobile devices installed. As for automated news feeds, a startup called Jinri Toutiao, or “Today’s Headlines,” currently dominates the market, with its 33-year-old founder Zhang Yiming recently becoming one of the youngest billionaires ever on the FORBES China Rich List.
Autonomous driving stands out because Baidu has better algorithms than local peers to teach cars to recognize objects and avoid collisions — a point of pride for Baidu as safety is the top concern for this futuristic technology, according to Zhang Xueru, an analyst at Shanghai-based research firm 86 Research. Deng Zhidong, a professor of computer science at Tsinghua University, isn’t as optimistic but agreed that Baidu is among China’s handful of internet companies capable of catching up with Google, as it has the country’s biggest self-driving research investment that includes two research labs in Silicon Valley.
“Baidu’s advantage mainly lies in deep learning, including algorithm and software,” he said.
Baidu is also accelerating swiftly in commercialization. Although related laws aren’t entirely in place, it is currently working with state-controlled King Long Motor Group to put a partially autonomous bus in an industrial park as early as next year, with several other companies including China’s FAW Group and Chery Automobile using Baidu’s software to manufacture vehicles with self-driving features such as in-car driving assistance and autonomous parking, according to announcements made at Baidu’s recent annual technology conference.
Blessing From Regulators
Aside from warning Baidu Chief Executive Robin Li when he rode an autonomous vehicle in July without permission, regulators haven’t intervened much. This is partly because the company’s moves fit into the government’s goal of leapfrogging over the west in frontier technologies, with mandates from the Made In China 2025 policy initiative requiring domestic firms to have the lion’s share of ten strategically important industries including AI and robotics.
“They [regulators] are re-examining current traffic laws,” said Xu Gang, a Shanghai-based partner at Boston Consulting Group. “Automakers also want to attract tech-savvy young consumers by promoting self-driving features.”
According to eMarketer, 82% of Chinese consumers polled in an August survey wanted to ride an autonomous vehicle. This compares with 59% in Singapore and 43% in Japan.
This means that by 2025, the market for sales of cars with self-driving functions as well as related services will balloon to $114 billion, with Baidu generating $6.2 billion in additional revenue from selling its auto-driving software in that year, according to 86 Research.
But significant barriers remain as the company eyes a fully autonomous future.
First and foremost, the technology for 100% driverless cars is still relatively new — though some early advances such as autopilot, which helps human drivers to change lanes or set routes, are ready for consumer adoption. Google, with more than three million miles in road tests conducted since 2009, reported a high-profile accident in 2016, when its autonomous SUV collided with a bus at low speed. Still, Tsinghua’s Deng said Google’s driverless cars are the safest worldwide, with an accident rate of about 1.25 per 10,000 kilometers — not yet as good as a human but getting close.
Safety-wise, Baidu said it hasn’t had an accident since it started the autonomous driving project in 2013, but declined to reveal total miles of road tests conducted. “The goal for fully autonomous cars is one accident or computer interference per 10 million kilometers, or 100 times safer than a human driver,” Deng said. “There is still a long way to go.”
Meanwhile, Baidu has figured out a way to make its robot cars smarter. In July, the company unveiled project Apollo, where it shares some self-driving codes and data with more than 70 partners, including auto makers Chang’an, Daimler and chip maker Nvidia. Baidu, in return, can get its hands on valuable road test data from vehicles using the codes, and harness them to train better robots.
“In the future, it will be data that is more important than algorithm.” Li Zhenyu, general manager of Baidu’s intelligent driving group, said in an interview with FORBES. “You have to let robots review as many accidents as possible, otherwise it is hard to simulate similar environments for cars to learn. This is why we think it is better that everyone can share data.”
Still, Li concedes that robot drivers can’t match humans yet. In China, where more than 260,000 people die from road traffic every year — almost eight times than that of the U.S. — complicated traffic conditions can be difficult for machines to fathom. For example, a police may beckon cars to move forward even when the red light is on. Although the gesture is easy for humans to understand, robots are unlikely to follow because they simply can’t recognize it.
“Sometimes robots just look retarded,” Li said. “You’d have to tell it repeatedly to look at the police, not the traffic light.”
What also stands in the way of mass adoption is the cost of completely driverless cars. With Lidar, a critical part of such vehicles that measures distance through laser light, alone costing up to $30,000, price is likely to deter a lot of people, analysts said.
But analysts agree that the market potential is bigger in China. “The Chinese government has stronger decision-making ability and execution powers,” said Yang Ming, a professor at Shanghai Jiao Tong University. “China also has good internet technologies. These are all favorable factors for autonomous cars.”