Baidu should not rush into making driverless cars

Recently, at the China Cloud Computing Conference, Baidu Senior Vice President and Chairman of the Technology Strategy Committee Wang Jin revealed that Baidu will launch driverless cars in the second half of this year, which will make comprehensive use of Baidu's existing big data, maps, artificial intelligence, Baidu Brain and other technologies. He said that Baidu will cooperate with third-party car manufacturers to manufacture driverless cars - Baidu will not build cars itself, but will build the "soul system" of cars.

In fact, Baidu's work on driverless cars is not news. A few months ago, there was news that Baidu and the Ministry of Transport were testing driverless cars. However, after reading this news, the author still feels a little worried.

Not long ago, Volvo's self-driving car suddenly accelerated and rushed towards the crowd after parking. This terrifying video went viral on the Internet, causing people to have doubts about self-driving cars.

You should know that Google started its self-driving car project in 2011 by acquiring a startup with core technology, and has not officially launched a self-driving car since its testing. Now Baidu is going to develop self-driving cars in the second half of 2015, and in China, a country with relatively weak automobile technology, it feels like a great leap forward.

1. What does Baidu have?

Baidu has invested a lot in artificial intelligence, big data, and maps in recent years. Not long ago, Robin Li demonstrated at Lenovo's technology conference that Baidu's software can recognize objects through cameras, select information, and complete the function of interacting with humans. This ability to recognize and distinguish image information is already very high. Baidu's Andrew Ng once revealed that in terms of face recognition, Google's 6,000 pairs of faces had an error rate of 0.37%, while Baidu's was 0.16%.

Although Baidu was accused of cheating in the ImageNet AI test not long ago, debugging the program more times than required, the final result was still excellent. For users, it is enough for the program you finally come up with to have a high recognition rate. How many times you debugged it is none of my business.

These things at least show that Baidu is relatively reliable in image recognition.

In terms of speech and semantic recognition, Baidu's current translation software has a very good accuracy rate, which is the result of in-depth research on natural language. It can separate words, organize the meaning of sentences based on words, and then understand and finally translate. This is also one of Baidu's assets.

In addition, through years of operation, Baidu Maps has, on the one hand, obtained enough geographic information, and it is becoming more and more perfect through continuous updates. On the other hand, Baidu has obtained a large amount of location information and O2O information through maps, and can provide users with a variety of services. Baidu also has group buying and payment, and users can consume directly, making it easier to establish a business model.

Artificial intelligence's recognition of images and voices, high-precision map information, and the business model behind the O2O payment chain are Baidu's advantages in driverless driving.

2. What is Baidu missing?

A driverless car is actually equivalent to a robot. The process of driverless driving is first to perceive external commands through various sensors, then the computer makes a judgment on these commands and gives orders to the car's power system, and finally the mechanical or electronic control part of the car responds according to the computer's commands.

Baidu's driverless test has just entered field testing in cooperation with the Ministry of Transport, but it is still far from practical use.

The power of Baidu's artificial intelligence lies only in the part where it makes judgments based on external instructions, and this judgment is still a simple judgment and is not combined with driving logic.

Baidu's image recognition is powerful and can identify pedestrians and vehicles on the road, but autonomous driving requires judging the route, speed and direction based on the location, speed and trajectory of these pedestrians and vehicles. In addition, real urban autonomous driving requires facing a lot of information and filtering out useful information. In fact, Google has been able to achieve autonomous driving on highways for a long time, and the tests in recent years have been testing how to deal with complex road conditions.

Baidu's speech semantic recognition is powerful, but when Robin Li demonstrated it at the Lenovo conference, there were problems. The reason was not that the recognition was poor, but that the mobile phone microphone did not filter the speech well enough in noisy and echoing conditions. Baidu is engaged in artificial intelligence, not mobile phone microphones. Mobile phone microphones affect the recognition rate. What Baidu can do is to replace them with better microphones.

The environment inside and outside the car is sometimes worse than that of a press conference, especially for domestically produced mid- and low-end cars with poor sound insulation. Coupled with the different dialects found throughout China, no matter how advanced Baidu's technology is, it will take some time to cope with this complex environment. It seems a bit too hasty to get the system into the car in the second half of 2015.

The biggest problem is the mechanical part. The driving control of the car depends on the engine, gearbox, transmission decoration, brake system, and steering wheel. At present, the core technologies of these domestic cars are basically in the hands of foreign suppliers.

The engine calibration was done by foreigners, the gearbox shifting logic was done by foreigners, the ABS brakes and the body stability system were done by foreigners...

Most of China's automobile manufacturers do not actually have core technologies. If Baidu cooperates with them, gives them control signals, and then comes up with the correct engine, gearbox, and braking actions, the reliability will be very worrying.

There are rumors that Baidu is cooperating with BMW, but China's BMW Brilliance does not have much core technology, and the gearbox of Germany's BMW is also supplied by ZF. Baidu does not cooperate with upstream ZF and Denso and directly cooperates with the vehicle manufacturer for the mechanical part, which will be a bit difficult.

Moreover, the actual road conditions are so complicated that any problem can be a matter of life and death.

3. AI is on the rise, taking small steps

As for driverless cars, the automotive industry has actually been developing in stages, starting with computer-assisted driving, then moving step by step to things like automatic following and high-speed automatic cruising, and finally driverless driving.

Baidu wants to develop driverless cars, the "soul system" of car manufacturing. This ideal is very ideal. Once it is developed, it will be equivalent to the Windows in the car, and both the pre-installation income and the entrance income will be lucrative. However, cars are related to personal safety, so it is not advisable to take too big a step.

In fact, Baidu can do it step by step. The mechanical part is more difficult, so it can be left unconnected for now. First, we need to master the image, voice, map, and O2O. The reliability of the sensor can also be tested slowly.

Therefore, Baidu and automobile companies can start with additional functions when developing driverless cars. Ordinary cars can be equipped with various sensors, replaced with high-performance car computers, and equipped with various intelligent functions.

For ordinary cars, you have to set the GPS when you go to unfamiliar places. With the intelligent system installed by Baidu, you can tell it by words, "I want to go to Yansha," and it will automatically detect the route and traffic congestion and set the GPS for you.

If you want to eat near your destination, just tell the intelligent system and it will automatically recommend restaurants and group purchases to you. If you are satisfied, it will automatically reserve a table and even pay directly.

While driving, if there is a slow car in front and you are driving fast, the intelligent system will sense it through sensors and then use voice to remind you to pay attention to your speed; if you turn on the turn signal, the intelligent system will use radar to tell you the speed of the car in front, whether there are any vehicles on the side or rear, how far away you are, how fast you are, and whether you can change lanes; if there is an old lady crossing the road, the intelligent system will detect it and prompt you to avoid her; if a vehicle from behind is rushing towards you quickly, the intelligent system will warn you of the possibility of a rear-end collision and detect the surrounding environment to prompt you to speed up or change lanes.

Simply put, autonomous driving can take small steps and move forward quickly. First, voice recognition can be added to the car to improve the user experience, O2O can be added to the car to meet user needs, and sensors, image recognition, and intelligent judgment can be added to the car to help the driver drive safely, and it will continue to improve in the process.

Once the voice recognition rate is high enough, the sensors, image recognition, and intelligent judgment have been tested sufficiently on vehicle and road conditions, and the control suggestions given are 100% correct, then we can consider combining the intelligent system with the vehicle's mechanical system.

Autonomous driving concerns the life safety of car owners and should not be rushed.

As a winner of Toutiao's Qingyun Plan and Baijiahao's Bai+ Plan, the 2019 Baidu Digital Author of the Year, the Baijiahao's Most Popular Author in the Technology Field, the 2019 Sogou Technology and Culture Author, and the 2021 Baijiahao Quarterly Influential Creator, he has won many awards, including the 2013 Sohu Best Industry Media Person, the 2015 China New Media Entrepreneurship Competition Beijing Third Place, the 2015 Guangmang Experience Award, the 2015 China New Media Entrepreneurship Competition Finals Third Place, and the 2018 Baidu Dynamic Annual Powerful Celebrity.