A man's choice to solve the hardest problem for the future of?mobility
Zhong Hua, SVP at WeRide.ai told us his personal story from learning alongside the greatest minds in Artificial Intelligence to his current mission with WeRide.
This article is the English adaptation from the original Chinese post on ZhiHu:
In 1990, in order to welcome the Asian Games, Beijing built an 11-kilometer secondary road on the east side of the Summer Palace and named it Zhichun Road. Lenovo and Sina in the 1990s started here. Knowing the spring road, knowing the spring of China’s Internet.
Today, there are still important innovative companies on both sides of Zhichun Road. Countless outstanding young people gather here to watch, chase, and create their own dream. WeRide’s Beijing branch was established here.
Facing the bustling street of Zhichun Road, with floor-to-ceiling windows, sitting in front of the large-screen display is WeRide Technology SVP Zhong Hua. He told us that an important goal in the first half of 2019 is to recruit Beijing’s best AI engineers to fill up the Beijing office.
WeRide was founded in Silicon Valley by Tony Han at Silicon Valley in 2017, and is now headquartered in Guangzhou with RoboTaxi operations running in multiple cities across China and U.S. Relying on first-class autonomous vehicle technology solutions, leading autonomous vehicle independent modification capabilities and rich autonomous driving road test experience, it has become one of China’s leading L4 autonomous driving companies.
Zhong Hua, who came from Guangzhou to Beijing to interview candidates, looks no different from other “code farmers”, wearing a black hoodie with the “WeRide” logo and khaki slacks, speaking with a bit of old Beijing’s ridicule and carelessness. But as soon as he talked about technology, he was energized, his eyes started gleaming, and his speed of conversation accelerated.
As an AI engineer, Zhong Hua’s has a solid resume. He graduated with a PhD in Computer Science from Carnegie Mellon University, under the tutelage of Takeo Kanade, the global leader in computer vision. Zhong Hua was the earliest employee of Microsoft Research Asia. He once worked at Google and Siemens Research Institute and Chief Architect of Autonomous Driving in China. In more than three hours of dialogue, Zhong Hua frankly shared the gains and losses and reflections of each of his experiences, and how these experiences guided his journey to WeRide.
01 From Failure to Enlightenment
Zhong Hua’s father is a professor at Peking University. He grew up in Peking University, from elementary school, junior high school to high school, and has always been among the best. But in retrospect, Zhong Hua thought that he had always been in a state of “bewildered” in his studies, and he didn’t have much of his own thoughts, “the parents can do whatever they want.”
In 1995, Zhong Hua was recommended to Tsinghua University. When choosing a major, because the teacher was worried that Tsinghua’s electronic communications major was too popular, he suggested that Zhong Hua choose to study computer. Zhong Hua’s reaction to this choice was “I loved playing games at the time, and thought it was good to learn computers.”
I really began to think about what I wanted to do and how I should go in the future. It was when Zhong Hua entered Microsoft Research Asia. At the end of 1998, Zhong Hua had the honor to enter the newly established Microsoft China Research Institute, which later became Microsoft Asia Research. As the earliest entrant, Zhong Hua, an intern, saw Li Kaifu personally “choose the colors of the carpets and walls.”
At that time, few people had any predictions about how this institution would have an earth-shaking impact on the Chinese science and technology community. For Zhong Hua, the biggest attraction here is that “there are many people and air-conditioning.” Shen Xiangyang, Wang Jian and other industry legends were working there. Today, twenty years later, recalling this experience again, Zhong Hua said with emotion that it was from Microsoft Asia Research that his life changed a different path and embarked on a different trajectory. “Shen Xiangyang is definitely my life mentor,” Zhong Hua said. “Enlightenment” started when I met Shen Xiangyang. Shen Xiangyang has given Zhong Hua a lot of inspiration and guidance in terms of professional technology and life direction. “I have benefited all my life from meeting him, and I am grateful for that time.”
Shen Xiangyang’s greatest influence on Zhong Hua also came from his leadership style. “He is a very positive person who leads by example, and can inspire many people around him. Seeing his hard work, you naturally want to work hard together.” Zhong Hua said that now he is leading the team to learn Shen Xiangyang’s style, paying great attention to the cultivation and inheritance of talents.
In WeRide, each technical team adopts the method of “veteran” and “novice”, with the ratio controlled at 1:1, and at most 1:2. Even if the pace of recruitment accelerates, WeRide will control the ratio of experienced technical veterans and novice engineers to ensure that the latter can obtain high-quality growth. Three months after the new employees are recruited, the technical team leader will give the new employees a long-term development route. Zhong Hua said that this approach allows employees to better understand their own positioning and is also conducive to long-term technological growth. WeRide’s R&D team in China and U.S. currently has more than two hundred people, 70% of whom are algorithmic engineers. There are not only the autonomous driving industry elites from top domestic companies, but also top engineers from Google and Facebook.
“Our hiring standards are indeed high,” Zhong Hua said frankly. Many people said that our interview test was too difficult for coding. Others said that these are all things from Silicon Valley and are not applicable in China. “But in terms of technology, we Still insist, have everything that should be there.” He proposed two benchmarks for selecting talents. One is to have enough experience, preferably an expert in a certain field. Two is to have strong engineering capabilities. The two criteria are equally important.
This kind of technical persistence is very consistent with Zhong Hua’s profile. In 2001, Zhong Hua went to Carnegie Mellon University (CMU) under the recommendation of Shen Xiangyang, specializing in computer vision. There, he studied under Takeo Kim, a global computer vision leader, and gained the first entrepreneurial experience in his life.
02 First Time Startup in Medical AI
In 1980, Takeo Kanide joined CMU and established the status of CMU in robotics research for many years. In the 2019 American graduate school rankings released by US News, the artificial intelligence major CMU ranked first.
During Zhong Hua’s Ph.D., doctors from Pennsylvania’s largest medical and insurance provider — — Pittsburgh Medical Center (UPMC) approached CMU and put forward the need to use computer vision technology for minimally invasive cardiac surgery positioning.Takeo Kanide handed this project to Zhong Hua. When Dr. Zhong Hua graduated, the accuracy of the ultrasonic navigation system he made was already at the millimeter level, while the best system on the market at that time was still at the centimeter level.
After graduation, Zhong Hua and UPMC doctors used this system to build a startup company. Zhong Hua is the only programmer in it. Medical entrepreneurship projects cost a lot of money, and it is difficult for small entrepreneurial companies to support. Zhong Hua and the others reached an agreement with a medical device company at the beginning of their business, and when the project reached the animal test stage, the company acquired all IP.
For this entrepreneurial experience, Zhong Hua has a deep memory: “I rented a cabin next to CMU. Except for sending my girlfriend to work in the morning and evening, I was just coding all day long.” At UPMC, Zhong Hua also went to an office where he used 3D-printed hearts to do experiments, participated in dozens of heart operations, and taught doctors to collect experimental data. This “simple, pure, technology-focused” environment is very attractive to Zhong Hua. He cherishes similar experiences in his past, and this has also become the background color of Wen Yuan’s technical atmosphere.
Zhong Hua joked that at WeRide, employees as long as they “do not run naked” and there is no requirement for dressing, he often releases himself and wears slippers to work. Under the “typical code farmer” dress style, Wen Yuan Zhixing’s working atmosphere is also very simple and open. “Our company is zero politics,” Zhong Hua emphasized that every employee is very clear about the company’s strategic goals and their own work goals. In this environment, everyone discusses issues very efficiently.
“Every week we will hold company-wide technical lectures, and each technical team will also share technology and study papers,” Zhong Hua said. Engineers can see all the codebases on the server. If the code is deeper, there will be a wiki for explaining the code and recording experiment results, so others will can learn it more efficiently.
03 Solution is the only Answer Engineers Need
“For a real engineer, the only answer is the solution.” Zhong Hua believes in this. The most thorough implementation of this point is the company — -Google, which is the “World Engineering №1” in his mind. Before entering WeRide, the company that had the greatest influence on Zhong Hua was Google. “I learned a lot at Google, especially like the corporate culture there,” said Zhong Hua. Wen Yuan Zhixing is creating an engineer culture like Google to maximize the creativity of the team.
“Find out what the problem you are trying to solve is, and don’t be stuck by the form of it.” This is what Zhong Hua always emphasized in Wen Yuan Zhixing. “You can tell me a hundred reasons why this thing can’t be done. It doesn’t make sense. , What I need is a way to make this thing. “The experience of working at Google has cultivated Zhong Hua’s be a doer attitude. Zhong Hua felt very “shocked” by the way Google solved the problem. There are many problems in engineering, usually using algorithms to reduce the time complexity of the program, thereby saving resources and expenses. In many cases, when the algorithm fails to reach it, the problem seems insoluble. However, at Google, sometimes a problem that cannot be calculated is solved by adding a machine, which seems “crude” but effective. Zhong Hua used 12,000 machines in the data center to run experiments at Google, “Google has this kind of decision, let you do it.”
Upon joining WeRide, he was impressed of the company’s unique way of solving hard problems. “Take LIDR as an example,” Zhong Hua explained. Many people think that the solution of LIDAR , which is worth more than 60,000 US dollars, is not feasible for driverless cars. “It is more expensive than a car.” But technically speaking, LIDAR can do many things that other sensors cannot. “This is very similar to the Google example of 10,000 machines,” Zhong Hua said. Our concern is to solve the problem of “safe and efficient delivery of passengers from point A to point B”.
Of course, behind this is the very comprehensive thinking and trade-offs of the WeRide management team. WeRide CFO, President and Co-founder Lu Qing used to be the CFO of LIDAR giant Velodyne and has a very in-depth understanding of this industry. Lu Qing once gave Zhong Hua a detailed analysis of the cost of using LIDARand his judgment on the future development trend of LIDAR, and these views are also confirmed today. Nowadays, the sensor fusion solution with LIDAR as the core has become the mainstream, and WeRide is one of the first companies to identify this solution. WeRide is using laser radar, camera, millimeter wave radar and other multi-sensor fusion methods to ensure hardware redundancy backup, enhance stability, and adapt to multiple environments.
In June 2018, WeRide’s self-developed autonomous vehicle successfully completed China’s first road test of an autonomous vehicle under heavy rain. In heavy storms, autonomous vehicles will not be affected by the heavy rain, accurately identify traffic lights and illegal vehicles, and steadily traverse a 1.5-kilometer-long tunnel. What attracts Zhong Hua is not only Wen Yuan Zhi Xing’s determination and courage to solve problems, but more importantly: “clear strategy.”
04 Iteration of Strategy is Suicide
Before joining WeRide, Zhong Hua chatted with several senior executives of the company for a long time and asked about the company’s strategy in detail. Because before that, Zhong Hua suffered from “unclear strategy”. Zhong Hua started his second business venture after leaving Google: to be an AI fashion e-commerce company. “Simply put, it is to use AI algorithms to recommend clothing collocations for you. All recommended clothing can be purchased directly by clicking on the link.” Even today, this idea is not out of date.
The technical team composed of four engineers is full of entrepreneurial passion, working six days a week, writing codes desperately. “We launched ios and Android apps in a week, from algorithm to front-end to back-end, I did all the work, not just learn now,” Zhong Hua said. However, this business was not successful. After a year, Zhong Hua, exhausted physically and mentally, left the company and took his wife and children to Hawaii for vacation. “As soon as I got there, I plunged into the sea and found the phone still in my pocket when I got out,” Zhong Hua recalled with a wry smile.
In Hawaii, Zhong Hua, who had no cell phone to cut off contact with the outside world, seriously resumed this entrepreneurial experience. “The biggest problem is that the strategy is not clear,” he said. Changing a strategy in a few weeks will not only make people very tired, they often have to rewrite applications, and there is no time to think deeply about each direction.
At that time, a book called “The Lean Startup” (Chinese translation “Lean Startup”) was very popular in Silicon Valley. This book encourages companies to conduct “validating learning”, first launching minimal prototype products to the market, and then in continuous experimentation and learning, to verify whether the product meets the needs of users in a minimal cost and effective way, and adjust the direction flexibly.
Some tactical things can be solved by high-speed iteration, but when it comes to strategy, it will die miserably. Zhong Hua concluded: “Iterating at high speed is a serious matter, which is equivalent to suicide.” “Think clearly before doing it!” Zhong Hua emphasized that only when you think clearly will you know why you stick to it. The strategy is consistent to go far. For Zhong Hua, WeRide has completely achieved this.
05 All-in Focus
The autonomous landing plan chosen by WeRide is to develop driverless taxi services. Zhong Hua believes that this business model has both “technical barriers” and “early landing”. In the field of autonomous driving, many people think about selling cars, but the premise of this model is to develop a system to ensure that the car can run in all places. This is a huge project and takes too long to land.
And the taxi runs in the urban area of a city, and “the taxi is our own, more controllable, as long as you can run in the urban area of a city, you can make the first dollar,” Zhong Hua Explain this way. At the same time, he believes that by solving the problem of driverless taxis, this solution can be extended to all the moving things on the road that require human operation, and there is a huge potential market behind it.
In November 2018, WeRide and Guangzhou Public Transport Group Baiyun Company jointly launched the country’s first L4 autonomous driving taxi. “We have a strategy meeting every six months, focusing on whether the past strategy makes sense and whether it needs to be changed.” Zhong Hua said. Every time, we find that our strategy is becoming more and more reasonable, and more and more investors are recognized, so we become more and more determined.
When it is determined to be 100%, WeRide will concentrate all resources to promote it. So far, it still maintains the fastest speed among the global driverless companies. Within 39 days of its establishment, WeRide completed the first closed field autonomous driving test, and the first open road autonomous test completed in 81 days. “We got the drive test license of the California DMV in two months, and the car started running in Silicon Valley in another two weeks,” said Zhong Hua. This speed is unimaginable by many companies.
At the same time, this start-up, established in April 2017, has also experienced a lot of turbulence and ups and downs. In December 2017, Baidu announced a lawsuit against Jingchi (the predecessor name of WeRide). With this as a starting point, WeRide experienced many difficulties that the outside world could not fully understand in the following year. Zhong Hua said frankly: “After a very anxious period,” WeRide originally started preparing for the A round at the end of 2017, but because of the dispute with Baidu, all investors said, “We must have a direct relationship with Baidu,” otherwise It is difficult to follow, “At that time, some investors thought our company was dead.” “I would like to thank many people for their help in this process.” Zhong Hua said.
After so many dangerous moments, Zhong Hua’s proudest thing is that WeRide’s technological development has not fallen behind at all. “Our technically this group of buddies is very powerful.” This kind of technical confidence is also a round of investment. Was affirmed. “Our A round investors are of high quality, and we are still looking for investors according to our established strategy.” In October 2018, WeRide announced the completion of Series A financing, strategically led by Renault-Nissan-Mitsubishi Alliance RNM. At the beginning of this year, WeRide’s A round of financing was followed by SenseTime and ABC International of SenseTime. This is the first financing announced in the field of autonomous driving in 2019. It will inject a boost into the industry in the cold winter needle.
We have been looking for partners from major international automakers who have stronger technical strength and can support the development of L4 level. This is an important step for us to land in China. Zhong Hua said that he thinks the Renault-Nissan-Mitsubishi (RNM) Alliance is far more significant than investment. Before investing, Alliance RNM sent an article to WeRide to investigate four or five times, “It is not the same as ordinary investors, they purely look at technology,” Zhong Hua recalled. People from Alliance RNM came to WeRide early in the morning and usually followed the experiment. The car ran on the road for several hours in the morning rush hour. After returning, the problem was listed and discussed on the big whiteboard.
“They have their own laboratory, they understand this matter very well, and they have seen many companies, and finally selected WeRide. This is a very important affirmation in technology,” Zhong Hua said.
Alliance RNM is the world’s largest automotive alliance. This is the alliance’s first investment in the Chinese market. It also makes WeRide the first L4 autonomous driving company in China to receive strategic investment from a global car manufacturer.
06 The Culture of G.R.I.T.
Out of the trough, WeRide plans to complete the modification of 500 autonomous vehicles in 2019, accumulate 5 million kilometers of autonomous driving mileage, and gradually conduct normalized trial operation and commercialization with partners in cities such as Guangzhou and Anqing. Attempt to achieve the goal At the same time, another important task is to attract a group of “like-minded” AI talents.
At present, the demand for artificial intelligence talents is in short supply domestically and even globally. Zhong Hua also said: “Excellent talents are hard to find.” However, in addition to high technical standards, Zhong Hua also repeatedly emphasized that he hopes to find people who are consistent with WeRide’s corporate culture and values. “What I evaluate most is drive and values, and these two points are the most important.” Zhong Hua said, he asked many HRs: “What kind of person is recruited and found that he is the most successful?” Everyone’s answers are unified, that is, values and The company is matched and the driving force is very strong, even if the technology may not be the strongest, but the development in the company is the best.
WeRide summarized the corporate culture and values as the word “GRIT”, which not only means “perseverance”, but also an acronym for the English acronyms of the keywords “Growth, RESULT, INNOVATION and TEAMWORK”. In Zhong Hua’s eyes, “GRIT” is the essence of WeRide’s corporate culture. He believes that corporate culture is actually a set of ideas in your mind, “When you encounter problems or need to make choices, this set of ideas can support you to make the right choice.”
Zhong Hua gave an example of the company’s demonstration for investors some time ago. The route that was originally selected to run through was temporarily unavailable, and he had to change to a new route within seven days. “On the afternoon of the first day of the plan change, the technical team found the second route. All the engineers rushed to make preparations and ran through this more difficult route within three days.” We were able to pull everyone in to do the seemingly impossible task, way before the deadline.
“At a critical juncture, one can act immediately or blame others,” Zhong Hua believes that this is the place where corporate values are embodied. No one said, “It’s definitely not a problem, it can’t be solved,” and no one shirks responsibility and blames each other. When faced with a challenge, do you take the challenge or not? How to make the hardest choice? For Zhong Hua, for everyone who is part of WeRide now and in the future, the answer should be the same.