Another Baidu AI Developer Conference and Li Yanhong Replaces Lu Qi as CEO
Li Yanhong, the Chairman and CEO of Baidu, replaced Lu Qi, the former President and COO of Baidu at the Baidu AI Developer Conference this year.
Since the launch of its autonomous driving ecosystem platform Apollo, Baidu has faced considerable controversy. While Baidu personnel have claimed that the need for collaboration from partners has surpassed expectations, resulting in prioritized support, others comment on Baidu’s determination, car manufacturers’ hesitation, and negative opinions about Apollo’s business model.
Yesterday, Li Yanhong stepped off the autonomous car and unveiled Apollo 3.0. What direction will this world’s most open autonomous driving ecosystem platform take?
The Persistence of an Elementary School Student, Baidu: Mass Production
Prior to a detailed discussion of Apollo 3.0, it is necessary to review the history of Apollo.
On April 19, 2017, Baidu Intelligent Driving Group (IDG), under the leadership of Lu Qi, unveiled an open platform consisting of four central platforms: vehicles, hardware, software, and cloud services, at the Shanghai Auto Show.
On July 5, 2017, Apollo 1.0 was unveiled; on September 20, Apollo 1.5; and on January 8, 2018, Apollo 2.0, along with Apollo 2.5 on April 19. IDG of Baidu has constantly expanded the open central platform since the unveiling of 1.0 , enhanced the core competencies, and increased the platform’s appeal.
However, just one month following the release of Apollo 2.5, on May 18, Baidu announced that Lu Qi would no longer serve as Baidu’s President and COO from July, with Li Yanhong taking over the first-line business management, including IDG. This made the 3.0 version and the evolution of the Apollo platform more uncertain.
Yesterday, Li Yanhong unveiled the mystery on the stage. He claimed that Baidu would continue with the established strategy and promote the openness and commercialization of Apollo, as he had said to Lu Qi’s employees at a meeting.
What did Apollo 3.0 release? This is the most critical PPT of Apollo 3.0. It shows that the 3.0 version has broken free from the architecture drawn in 1.0 and launched a new category “Mass Production Solution.”
For internet companies like Baidu, promoting mass production of autonomous driving vehicles is a particularly difficult issue.
The most important problem is that Baidu is essentially an outsider to the automotive industry and has no knowledge about a century of technological accumulation and safety standards in the industry. Mass production and being able to put vehicles on the road are completely different difficulties from modifying a few test vehicles to run demos. The most important problem is to solve safety.
The safety culture of automotive manufacturers is safety first. All seemingly cumbersome testing, verification, and project management are based on safety.
The release of Apollo 3.0 made us re-understand Baidu.
First of all, it is more comprehensive and thorough openness, which is the best way to dispel the concerns of partners. In the lower right corner of the PPT of 3.0, there is an “open vehicle interface standard”. What did Baidu open?
This is a biased technical PPT, but it is also true that industry insiders can understand that Baidu has opened all the necessary vehicle interfaces around the mass production of autonomous driving vehicles.
“How to balance the speed of innovation with safety testing, and when there is a conflict, we first ensure safety.” Baidu IDG General Manager Li Zhenyu said. This year, Apollo’s code has grown 6 times from 35,000 lines to 220,000 lines. “Half of the new code is added to improve safety, whereas before it was more for functional and algorithmic improvements. Later, we spent a lot of effort to solve problems such as how to make other sensors work when one sensor fails. We learned and referred to many safety standards in the automotive industry.”
In addition, Baidu has collaborated with Mobileye, the world’s largest ADAS giant under Intel, to bring the latter’s responsibility-sensitive safety (RSS) model to the Apollo platform.
RSS transforms people’s concept of safe driving into a verifiable model equipped with logically verifiable rules and appropriate response behaviors to ensure safe decision-making by autonomous vehicles. You can think of it as a benchmark product launched by Mobileye to address the safety of autonomous vehicles, which Baidu has brought to China.
Finally, Apollo released China’s first and the world’s third autonomous driving safety report. The report covers seven chapters, including takeover mechanisms, hazard response, information security, and remote assistance, elaborating on Apollo’s safety framework. Prior to this, only autonomous driving pioneers Google Waymo and General Motors’ Cruise had released similar safety reports.
Looking back at Apollo over the past year, from initially only daring to “provide a software platform to partners” to step-by-step examination of automobiles and working with OEM research and development teams like Golden Dragon, understanding how to realize mass production of autonomous vehicles, modifying onboard computer line control protocols, how to achieve vehicle regulations for underlying execution mechanisms such as steering and braking, etc. Baidu IDG slowly revealed its differentiation from most autonomous driving startups.
As the platform leader, Baidu can humbly learn the development of the automotive industry, adhere to safety on the eve of commercialization when technology approaches, and lower barriers. This kind of attitude of a humble learner is key to the smooth launch of Apollo 3.0 and the successful commissioning of 100 Level 4 autonomous driving buses, Apolong.
“We just want to lower the threshold”
The 100 Apolong buses will soon be deployed in commercial operations in Beijing, Xiong’an, Shenzhen, Fujian Pingtan, Wuhan, and other places; the autonomous driving logistics vehicles that cooperate with autonomous driving logistics company New Stone Age have been put into operation in Xiong’an and Changzhou, even before the press conference. The common point between these two vehicles is that they are equipped with LiDAR, cameras, millimeter-wave radar, RTK and GPS inertial navigation systems, high-precision maps, and the steering wheel, acceleration pedal, and brake pedal are all removed before leaving the factory.
This is the standard L4 level autonomous driving vehicle, considering the scene limitations, maximum driving speed, and software-level security deployment mentioned earlier, Baidu IDG has minimized the potential accident rate. At this point, the topic of doubting whether Apollo “has a chance or not” has lost its market.
But let’s look back, why did many industry insiders not believe in Apollo at the beginning?
No one has ever attempted the Android of the autonomous driving era. The automotive industry has capital and technology intensiveness, and an extremely long industrial chain. Balancing various interests, resolving differences in data rights, and enlarging the ecology, all are unknowns.Amazon founder Jeff Bezos once said that existing players in the industry will not like you when you do things in a new way. If you want to innovate, you must be prepared to be doubted and misunderstood. In a broad sense, new species like Apollo, TikTok, and Pinduoduo face similar difficulties in public opinion. People do not understand, so defensively do not see it as promising.
But with Apollo 3.0, Baidu has opened everything for mass production. Mass production tool components, autonomous driving kit, safety protection system, human-machine interaction plan, fleet management, map service, OTA upgrade. In extreme cases, as long as you have application scenarios, manufacturing systems, and operation and maintenance systems, you can completely build an autonomous driving fleet even if you lack the genes required for autopilot.
“We just want to lower the threshold. To use an analogy, climbing the mountain of autonomous driving, everyone is building ladders, with some building 3 meters and some building 10 meters, but Baidu opened a 100-meter ladder to help everyone climb faster. Stop building ladders,” said Li Zhenyu.
In fact, as a platform builder, if Baidu’s big flag swings but there are few responders, Baidu will be stuck in the dilemma of being unable to land its technology. “To be frank, this is a win-win situation. If there were no Xinshiqi, Apollo, Baidu would also be in trouble.”
Lowering the threshold also has a greater impact. When Apollo was launched, there was an unfounded rumor on the internet that “Baidu exploded with tens of billions of dollars of R&D investment worldwide.” Today, the release of version 3.0 has put real pressure on autonomous driving start-ups outside the Apollo ecosystem.
Take one example: Xinshiqi’s autonomous driving logistics vehicles. A few days ago, Chehejia’s official website updated its strategic layout, revealing investment in incubating the autonomous driving logistics vehicle development enterprise Xinshiqi.
This is Chehejia’s SEV and Xinshiqi logistics vehicles terminated from listing due to strategic adjustments.
In terms of hardware, Chehejia’s passenger car R&D team has devoted two years of research and development to create the SEV chassis technology and three-electric power assembly, which have come back to life in Xinshiqi’s logistics vehicles.
On the software side, Baidu Apollo provides autonomous driving kits and fleet operation management solutions.This is a highly competitive product that ranks first in the entire logistics robot industry. The comprehensive cost control of the entire vehicle has reached its peak. The supply chain of the three electric powertrain of car and home has been shifted to New Stone Age. The automatic driving kit is directly applied to the price of Baidu Apollo sensor system.
At the same time, both Car and Home and Baidu have accumulated more than two years of technology in terms of overall vehicle performance and reliability.
Now, here’s the question: Why did Car and Home and Baidu collaborate with New Stone Age?
New Stone Age is a company active in the logistics industry chain, with commercially successful products such as express cabinets and handheld terminals for express delivery personnel. In other words, “New Stone Age is a company with huge application scenarios, manufacturing systems, and operation and maintenance systems. Although it has no inherent gene for self-driving, it can still assemble a fleet of self-driving cars.”
Car and Home’s output, Baidu’s model project, and New Stone Age’s blue ocean market all present a win-win-win situation with great prospect.
Moreover, within the Apollo ecosystem, there are at least two startups of autonomous logistic vehicles, competing with New Stone Age – for Baidu, a true autonomous driving ecosystem is taking shape.
Where is Apollo 6.0 heading?
Continuously expanding openness → enhancing ecosystem attractiveness → promoting ecosystem prosperity, promoting the commercial landing of autonomous driving technology in segmented markets, and forming a sustainable commercial closed-loop of algorithms, vehicles, and operations. This is the development concept of Apollo.
Looking at the method of fighting alone, regardless of the investment from 2009 to 2017, Waymo ordered 82,000 autonomous vehicles this year, with a unit cost of over $100,000. Combining with the depreciation and operation cost of the fleet, Waymo needs to spend billions of dollars to maintain a technological advantage by 2020, which is effective but also arduous.
Apollo is a completely different concept. The optimistic side is that Baidu has formed a commercial closed-loop in voluminous blue ocean markets such as freight logistics, delivery and takeout, and floor-sweeping robots, and has shared the R&D investment by opening core capabilities to its partners. By constantly collecting data from landed fleets in segmented markets, positive and negative feedback is formed to improve algorithm technology.
The uncertain side is that even the highly competitive New Stone Age has not yet entered the huge application scenarios of Cainiao Shunfeng, Meituan and Ele.me.
Don’t forget that Baidu and a group of domestic car companies including BYD, BAIC, Chery, Great Wall, and FAW have agreed to produce L3 level autonomous passenger cars in 2020. The indirect tactics are to build up strength for the final confrontation against Waymo.After 441 days, a more proactive and open Baidu Apollo 3.0 is coming.
The keywords for Apollo 6.0 in 2019 were the full landing of a segmented market and the scheduling of mass-produced L3 class passenger cars.
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.