In the summer of 2017, during the live broadcast of the Baidu AI Developer Conference, Robin Li took a driverless car to the fifth ring and then ate a ticket.
Not only was he not depressed, but he was happy. Later, when talking about this ticket, he shouted: "The driverless ticket has come, will the mass production of unmanned vehicles be far away?" ”
According to Li Yanhong's expectations at that time, the mass production of unmanned vehicles and the realization of unmanned driving on fully open roads are also things that will happen in the next three to five years.
"As long as you drive on the highway, you don't have to care, you can eat hot pot and sing songs..."
But apparently, this is idealized after all.
In 2022, only the footsteps will be heard, and no one will come down.
Even the sound of footsteps has weakened a lot compared to the past.
The first is that "automatic driving" has replaced "unmanned driving" as the mainstream caliber of manufacturers, and secondly, the "automatic driving" promoted by manufacturers is itself exaggerated.
According to the Society of Automotive Engineers classification, autonomous driving is divided into five levels, including L1-L5.
In simple terms, in the L1, L2, L3 levels, the driver is still dominant, until L4, L5, the vehicle control is completely owned by the automatic driving system.
A very bone-chilling reality is that the current general development stage of the industry is also around L2-L3, that is, most of them are only in the stage of automatic driving assisted driving systems, far less than the L4-L5 level.
At the same time, Tesla, Weilai and other markets from time to time related to automatic driving accidents, but also continue to use bloody lessons to break people's illusions about this technology that represents the future.
The once-excited autopilot world is no longer so noisy and has begun to be cautious. People began to wonder what year and month the so-called "hands off the wheel" could be truly realized.
But Robin Li remains optimistic.
At last year's Baidu AI Developer Conference, Robin Li judged that intelligent transportation will be a major change that will affect the next 10-40 years, so that China's first-tier cities will no longer need to restrict purchases and traffic within 5 years, and basically solve the problem of congestion within 10 years.
At the end of the same year, he also published a book with the same proposition.
In the book, Robin Li expects that from 2024 to 2025, autonomous driving will begin to gradually enter China's urban roads; from 2027 to 2028, large-scale autonomous driving fleets will begin to operate in cities; by 2030, the proportion of autonomous vehicle fleets will begin to increase, and it is expected to reach 30%-40%.
Even if the past has been ups and downs, the road ahead is still difficult, but for this very good technological future, Robin Li is still working hard.
It is better to give automatic driving and give Baidu a little more time to see if his optimistic prediction can be fulfilled as scheduled this time.
Dreamers
Among those practitioners who dream of autonomous driving, Baidu is undoubtedly the most thorough, and it can also be called a desperate bet.
First of all, it is natural for Robin Li to shake his arm.
In the past decade, whether it is the various public occasions that Robin Li has participated in or the conferences held by Baidu itself, Robin Li has worked hard to bring AI and depict emerging technologies including autonomous driving and deep learning to the public.
Even brought to the scene of the two sessions, actively on the promotion of AI, automatic driving and other scenarios of the application of suggestions and suggestions.
Moreover, Baidu is deeply cultivated.
Public information shows that as early as 2010, Baidu began the layout of AI, becoming the earliest pioneer enterprise in China to deeply lay out AI; in 2013, Baidu Deep Learning Research Institute was established, which is also the first research institute in the global business community to use "deep learning" to be named; in April 2014, Baidu established a big data laboratory, and in May of the same year, it established a Silicon Valley artificial intelligence experiment; in 2017, it added a business intelligence laboratory, a robot and autonomous driving laboratory, and so on.
In addition to technology, Baidu has also been building an autonomous driving ecosystem.
In 2017, Baidu launched the Apollo plan to open the autonomous driving platform, which is the first system-level opening of autonomous driving technology in the world, Baidu's idea is to create "Android in the field of automatic driving", hoping to gather the global automotive and technology industry forces to accelerate the landing process of autonomous vehicles.
Since its development, Baidu has gained a lot in the field of automatic driving.
Baidu previously released Apollo self-driving travel service data that as of the first half of 2021, Baidu Apollo's self-driving travel service has received more than 400,000 passengers, tested more than 14 million kilometers, and the number of autonomous driving patents has exceeded 2,900.
And according to the "Beijing Autonomous Vehicle Road Test Report (2020)", in this year's Beijing, Baidu's 1.1253 million kilometer test mileage ranked first, 20 times more than all other players combined.
It is worth noting that such test miles are still completed in different test areas and road grades.
Previously, the difficulty of road testing in Beijing was divided into 4 levels (R1-R4). As of 2020, only Baidu has one company that has carried out R4 level testing. In addition, the country's highest technical level, the highest standard, the most difficult test scenario of the open road test qualification certification (Beijing T4 license), only Baidu to get.
In terms of automatic driving, Baidu obviously has enough say.
It's just that these distances completely illuminate the dream of autonomous driving into reality, or it is not enough.
The reality dilemma
To a certain extent, the development of Baidu's automatic driving is not as expected.
From 2017 to 2018, as Baidu's ai and autonomous driving stories gradually entered the hearts of investors, Baidu's market value hovered around 100 billion US dollars for a long time.
But then there was a shock, falling back to $50 billion so far.
The reason lies in the imagination of autonomous driving, which is gradually defeated by reality.
On the one hand, Baidu overestimated Apollo's open strategy.
In the past few years, the output of Baidu Apollo's autonomous driving technology has been less than expected, and although there are dozens of cooperative car companies, few models are fully equipped with its autonomous driving technology and have achieved large-scale mass production. Having nowhere to display a martial art becomes an embarrassment for Baidu.
This directly led to the failure of many of Robin Li's promises.
On the other hand, it is also a challenge from technology.
Even if many companies, including Baidu, have mastered autonomous driving technologies such as L3/L4, this does not mean that they can go directly to the road.
Take Tesla, for example. Amid an investigation by road safety regulators after multiple accidents in the U.S. related to Tesla owners' excessive trust in autopilot systems, Musk said mastering the technology was harder than he expected.
Musk acknowledges the industry's long-standing consensus that self-driving car technology is far more difficult than it is theoretical because cars must learn, detect, and avoid variables encountered in countless real-world scenarios as they travel.
In addition to having bike intelligence that can detect lanes, other vehicles, traffic signals, and pedestrian movements, fully autonomous vehicles must also learn to preemptively guard against the behavior of other drivers and pedestrians like humans, which is almost an impossible task at this stage.
More critically, the immaturity of the entire industrial chain, as well as various supporting environments, ethical regulations, etc. are still not yet equipped with the conditions for large-scale application of automatic driving.
All of this has also directly led to Baidu's commercialization of automatic driving.
In the past decade, Baidu has been using advertising profits to transfuse blood to drive autonomously, hoping that the latter can grow into Baidu's new growth engine after commercialization. But the business model of autonomous driving has not worked.
Obviously, for Baidu, if you want to promote the real arrival of the era of automatic driving, these mountains must be gradually broken.
A shift in thinking
Fortunately, while Li Yanhong insisted, he was also adjusting his strategy.
Apollo opens up the way to rely on car dealers, because the attitude of car dealers is not as positive as imagined, then they will build their own cars.
So there is setness.
Although Robin Li said in an interview in 2018,
"Baidu doesn't build its own cars. We don't open production lines, design engines, study fluid mechanics or industrial design, we only do what we do best in the research and development of autonomous driving technology, and open up, and all the enterprises that are interested in unmanned vehicles, have ideas, and have the ability to do it together. ”
But in the face of slow progress, Baidu still has to get in the car.
According to Robin Li, in the first half of 2022, Jidu will announce the first concept car, and in 2023, the first automotive robot will be mass-produced and delivered.
The public can't see or touch automatic driving, so push the "radish fast run" to test operation in the city.
In February this year, Baidu announced that its self-driving brand "Radish Run" will open up manned test operation services in Shenzhen, which is another city where Baidu autonomous driving has entered after six cities, including Beijing, Shanghai and Guangzhou. The relevant person in charge of Baidu revealed that by 2025, the "Radish Run" will open up manned test operation services in 65 cities in the mainland.
Users can call a car through the "Radish Run" App, and they can experience the real self-driving travel service.
Industrial support and transportation ecology are not mature, then automatic driving will be embedded in the framework of "intelligent transportation", and the latter will be exported to local cities as the main selling point.
In April 2020, Baidu Apollo officially launched Baidu ACE Intelligent Transportation Engine 1.0, becoming the first full-stack intelligent transportation solution at home and abroad that integrates vehicles and roads. A year later, it was upgraded to Baidu ACE Intelligent Transportation Engine 2.0 again, forming a "1+3+N" overall architecture with cars, roads, clouds, and figures as the digital base, Apollo automatic driving, vehicle-road collaboration, and MaaS behavior intelligent engines, and becoming a systematic platform for the industry to truly achieve the three goals of "automatic driving, vehicle-road collaboration, and efficient travel" represented by "ACE".
It can be seen that although autonomous driving is still far from people's ideal scenarios, more and more users experience autonomous driving travel services, more and more roads begin to open the test of autonomous vehicles, and more and more cities begin to accept intelligent transportation solutions to gradually build ecological support...
Autonomous driving may also become less out of reach than imagined.
Robin Li also maintained enough confidence.
In his new book" Intelligent Transportation: Major Changes Affecting Mankind in the Next 10-40 Years, Robin Li believes that the inflection point of the era of intelligent transportation has arrived, and he is also prepared for a long process.
He quotes Stanislav Lem, who many call "futurists": "We always love to extend the future development of new technologies in a straight line... History has nothing to do with this simplified presentation. It will never show us any linear development path, but will only use tortuous and roundabout lines to show the non-linear evolutionary trajectory. ”
Yes, technological transformation of the world is always tortuous, but gradually approaching. It depends on whether Baidu can eventually reach the other side of the dream of automatic driving.
(Source|.) AI Blue Media Collection Author |Ye II)