Editor's note: this article is from the official account "Chief Business Review" (ID:CHReview) of Wechat, author: Lao Dao36氪Released by authorization.

Lian is quite old, can he still make a living? In the market capitalizationMeituan、JingDong、PingduoduoAfter the rising stars surpass, "Baidu"already old" seems to be a periodic after-dinner topic, setting off a round of discussion from time to time.

In recent years, Baidu's position is indeed very awkward. The current market capitalization of Baidu is 44.4 billion US dollars, less than 1/4 of Meituan's.AliThe market capitalization gap is even bigger, about 1/18 of Ali's. The three giants of BAT, Baidu, have long been unworthy of being one of the top three of the Internet in China.

For the capital market, why is Baidu so unfavored? Chief (WeChat account: chreview) from a more objective and rational, and calm and simple point of view, there are only two main reasons:

First, Baidu's main operating income is relatively single, although relatively stable but full of risks. Most of Baidu's revenue comes from advertising, and Baidu's profit model has been criticized by the rules of bidding ranking, especially the medical advertising that Baidu once referred to.

Second, Baidu's ecology is not yet in place, its business units are far less diversified than the other two giants, and it has not created new traffic entry-level products (such as Wechat and short videos) in the mobile Internet trend. Although Baidu has also invested in some enterprises, such asiQIYI、CTRIPEtc., but these services can not achieve mutual empowerment and flow superposition for the entire ecological construction of Baidu, and Baidu's sense of existence in the invested enterprises is on the low side.

So, at least in the previous stage, Baidu made people have no big expectations for its future. WhenTencentAnd Ali each based on their own ecology, layout of e-commerce, scramble for payments, refurbished games, and even in the same city retail, local life, audio-visual entertainment and other areas almost all-inclusive, Baidu in addition to the original Baidu app, only active in maps, network disk and other smaller market segments.

Robin Li once said that from 2002 to 2003, he was reminded to pay attention to the mobile market. At that time, he ignored it. "by the time the mobile Internet really came around 2011, a lot of people would have laid out it, and we didn't have a layout. That's too bad. I really panicked at that time. "

However, Robin Li redefined a new label for Baidu's future: "the arrival of the AI era is indeed relatively leisurely." Because we began to lay out in the technology market of AI about six or seven years ago, or even a little longer. Even today, even if everyone still has some hope for Baidu's future, they also think that Baidu's future opportunity lies in AI.

Can AI help Baidu to overtake around the corner one day?

1Baidu's Apollo Program

In an interview with Baidu in 2017, connected magazine, Robin Li revealed that Baidu has used all its latest AI technology on the "Apollo" project.

In 2017, Baidu launched the Apollo program for the first time to provide an open software platform for partners in the automotive industry and self-driving field to build a complete autopilot system in the industry.

After Changsha and Cangzhou, Baidu announced the full opening of its self-driving taxi service in Beijing on October 10.

At the self-driving taxi stations in Haidian and Yizhuang, there is no need to make a reservation and directly place an order for a free trial ride in the self-driving taxi service. Baidu released relevant data on Oct. 14, saying that in only 12 days, the cumulative order for the whole day reached 2608, and the peak number of orders on a single site reached 600. the operators of Baidu Apollo said they were "stunned" by the order volume.

At present, Baidu self-driving taxis still face problems such as long waiting time to place orders, difficulties in taking a taxi and the need for security guards to take over under special circumstances, according to time Weekly. Faced with the doubt that there are security officers in the "unmanned car", Baidu said that at present, its vehicle is still in the manned test stage, and the safety guard for the main driver is a requirement of the existing autopilot manned test policy. But in fact, Baidu has technically realized the unmanned driving of the main driver.

At the Baidu World Conference held in September this year, Li Yanhong proposed the ultimate destination of self-driving-5G cloud driving, that is, in self-driving vehicles, long-distance drivers will replace in-car security guards.

Li Yanhong has predicted that in the next five years, self-driving technology will enter the stage of large-scale commercial use. China's first-tier cities will no longer need purchase and traffic restrictions, and the congestion problem can even be basically solved within 10 years.

October eighthGoogleWaymo, the parent company's self-driving subsidiary, announced that it will launch a self-driving taxi service without security guards in Phoenix, Arizona, which will replace in-car security officers with remote supervision.

In April this year, a research report released by the China Electric vehicle Association of 100 pointed out that although current technology can deal with 90% of the conventional problems of traffic conditions, the remaining 10% of the problems have a huge impact. Traffic travel is related to the safety of every passenger, not to mention that there may be a 10% problem, even if there is 1%, it still means 100% for passengers who have accidents, so it will take longer to solve unpredictable situations. The long tail effect is obvious.

At present, there are still two problems that autopilot needs to be separated from the safety officer in the car, one is the division of responsibility after an accident when assisting driving, and the other is that although the current test data can cover most application scenarios, however, compared with manual reliability, it still needs a lot of data for in-depth comparison and research.

2Can AI save Baidu?

Self-driving is just one part of Baidu's AI plan. It is clear that what self-driving means is not just about optimizing a driver. More importantly, self-driving is bundled with intelligent transportation, and assuming that one day this technology is highly mature, it has the potential to solve two big problems:

First, it is safer. Self-driving, or AI driving, is safer than manual driving. Because in the supercomputing system of AI, when facing the unexpected situation on the road, it can respond more quickly than manual, and the processing scheme is better.

Second, when all the vehicle information is interconnected, the vehicles on the road can be predicted and diverted more scientifically, and a better route can be planned, so as to reduce traffic congestion.

But the premise is that the inherent intelligent processing system of self-driving is to get a breakthrough development, which is more scientific and efficient than manual driving. In order to realize this dream, mankind still seems to have a huge gap to cross.

In his book "intellectual Revolution", Robin Li pointed out:

Artificial intelligence can be divided into "narrow AI" (Narrow AI) and "generalized AI" (Strong AI). "Generalized AI" is a system that uses the same algorithm to solve a large class of problems. In principle, the "generalized AI" system can learn and adapt to solve new problems without human intervention. The "narrow AI" system uses specific algorithms to solve specific problems, such as playing chess, recognizing pictures, and so on.

At present, almost all the companies that can be seen doing artificial intelligence are doing "narrow AI". "narrow AI" only solves one problem, or one or two relatively narrow problems. Playing go, playing cards or driving a car is all "narrow AI".

Correspondingly, "generalized AI" can solve all problems with the same system, which is similar to human intelligence. "Generalized AI" is the long-term goal of the development of artificial intelligence, and its real realization will take at least 20 or 30 years. Now, Baidu, Google,Microsoft、FacebookAnd other companies are working in the direction of "broad AI".

Here, Robin Li defines driverless AI as AI in a narrow sense because he currently only defines driving as stable driving under regular procedures, according to traffic rules, without considering how to deal with emergencies, how to optimize traffic jams and other ever-changing and complex problems.

Zhang cymbal, academician of the Chinese Academy of Sciences and director of the Institute of artificial Intelligence of Tsinghua University, is more conservative. He believes that the current human artificial intelligence based on deep learning has reached the ceiling in technology. A breakthrough from a narrow AI to a broad AI requires a huge amount of time and even subversive and revolutionary innovation for today's human beings.

In Zhang cymbal's view, at present, the business community and some academic circles all over the world are too optimistic about deep learning technology, and artificial intelligence urgently needs to be promoted to a new stage, which is destined to be a long process. it depends on the breakthrough of the underlying theory in combination with mathematics and brain science.

The breakthrough of artificial intelligence needs three links: the first is big data, the second is computing power, and the third is algorithm. The latter two are the more difficult bottlenecks.

So far, artificial intelligence can do three things (speech recognition, image recognition, go) because it meets five conditions, that is, as long as these five conditions are met, the computer can do it well. as long as any one or more conditions are not met, it is difficult for the computer to do.

The first is that there must be sufficient data; the second is certainty.

The third is the most important, which requires complete information. Go is a game with complete information, and cards are games with incomplete information.

The fourth is static, including the law of evolution according to certainty, which is the problem of predictability, which is not satisfied by autopilot in complex road conditions. In fact, it satisfies neither certainty nor complete information.

The fifth is a specific field, which he cannot do if it is too wide. A single task, that is, artificial intelligence software for playing chess, is playing chess and can do nothing else.

Zhang cymbal believes that the characteristics of the work that meet these five conditions are very obvious, that is, the four words "do things according to the rules" and there is no need for flexibility.

At present, Baidu's self-driving system is still in its infancy and can only "follow the rules" to a certain extent. It is still a long way off to be able to actively judge and make correct decisions under special circumstances.

3The pain of Baidu

In the era of PC, Baidu has become the most important traffic distribution platform in China. Especially after Google withdrew from the Chinese market), Baidu is dominant in the search field and has no competitors.

Over the past decade, Baidu has carried out persevering and repeated expansion of new business boundaries, such as investing in a large number of vertical subclassification websites (such as iqiyi, Qunar, etc.), and exploring the e-commerce of C2C. And then to the expensive building of O2O.

From an operational point of view, when Baidu holds a large amount of traffic, invests in vertical websites, and uses traffic to pry investment income, this is the best time for Baidu. However, with the arrival of the mobile Internet, Baidu's dominant position in traffic continues to dilute, investment enterprises have experienced "shutdown and merger", O2O business has ceased and financial split, and after competitors have built strong barriers, Baidu has shifted its focus to Baibai, sponsoring the Spring Festival Gala, improving user stickiness and maintaining the bottom line of advertising revenue.

On the morning of August 14, Baidu released its unaudited financial results for the second quarter ended June 30, 2020. According to the financial report, in the second quarter of 2020, Baidu achieved revenue of 26.03 billion yuan, operating profit (Non-GAAP) of 5.6 billion yuan, an increase of 187% over the same period last year, and net profit (Non-GAAP) of 5.08 billion yuan, an increase of 40% over the same period last year. Revenue and profit indicators have exceeded the average market expectations.

Data show that Baidu App still had 204 million daily active users in June 2020. The three pillars of mobile ecology, which are composed of Baijia account, Mini Program and managed pages, maintain a high growth trend: the number of content creators of Baijia account is about 3.4 million, up 52% from the same period last year; the number of monthly active users of smart Mini Program reached 339 million, up 25% from the same period last year. The number of smart Mini Program has more than five times that of the same period last year; revenue from Baidu managed pages accounts for about 30% of Baidu's core online marketing service revenue.

Baidu's Q2 financial report for 2020 is still quite good. But there are two hidden problems:

First, compared with other Internet giants horizontally, Baidu originally has a low base, although the increase is very large, but the total plate still can not keep up with the footsteps of others.

Second, in the composition of revenue, advertising business accounts for more than half, the source of income means that the business unit is relatively single.

So far, advertising still accounts for 83% of Baidu's core revenue ratio, which means Baidu is still an "online marketing company". The innovation business is still in the period of infrastructure investment and loss, at this time, the advertising business not only shoulders the important task of maintaining growth, but also bears the mission of diluting the loss of innovation business.

The 2019 China Internet Advertising Development report of Beijing normal University once disclosed:

In 2019, advertisements from e-commerce platforms accounted for 35.9% of the total, ranking first, up 3% over 2018; search platform advertising still ranked second with a share of 14.9%, but down from 21% in 2018; revenue from video platforms increased by 43% year-on-year, replacing news and information platforms to become the third largest Internet advertising platform.

The status of e-commerce and search advertising has been challenged, and the sudden rise of video, especially short videos, has not only inherited the main increments of online advertising, but also squeezed the market share of established companies represented by Baidu.

4The difficulty of Baidu

In January 2017, Lu Qi, who had high hopes, joined Baidu. He served as President and Chief operating Officer of Baidu Group. In February, he was appointed Director of Baidu and Deputy Chairman of the Board. Since March, he has also served as General Manager of Baidu Smart driving Group.

A major adjustment of Li Yanhong is that the executives who originally reported directly to Li, such as Hailong, Zhang Yaqin, Zhu Guang and Wang Jin, reported to Lu Qi instead. It can be seen that Li Yanhong's trust and trust to Lu Qi.

Lu Qi, a skinny man, danced with his hands above his head after swinging his right arm at Baidu's AI development conference on July 5, 2017, prompting cheers from thousands of AI developers. At that moment, the 56-year-old Lu Qi looked like a teenager, and in front of him was a giant named AI, who had just opened his eyes.

Zhihu @ Zhang Yuncong wrote:

In my opinion, Qi is really a very capable and practical technician. And he really regards Baidu as his own company, which is very important. In the past, many senior executives never considered the interests of the company, only their personal gains and losses, and only knew how to report things well, rather than thinking about how to do things well. Layers of managers were forced by these executives to consider only short-term interests, not long-term interests, and then forced on technicians, the whole company has a particularly short-sighted feeling of doing things. Baidu has a lot of elite troops technically, but it can't win a battle. With the arrival of Qi, at least Baidu has taken a turn for the better.

Lu Qi

The addition of Lu Qi opened the road to the revival of Baidu. He drastically cut off medical advertising, takeout and other unfocused businesses, put forward the "All in AI" strategy, condensed Baidu's energy into the best technology areas, and found a direction for the confused Baidu. According to Baidu insiders, Lu Qi specially set up a "fresh air meeting" to discuss the company's direction and business with Baidu employees on a regular basis in response to difficulties in communication between Baidu internal departments, serious internal friction, and middle-level leaders who reported good news but not bad news. Won the support of grass-roots employees.

In Lu Qi's more than 400 days in power, Baidu's share price has risen by nearly 60% since the beginning of his tenure.

Some people think that Baidu's internal structure is too complex and huge, and everything should be done in accordance with established procedures. Serious bureaucracy has lost the strategic direction that the company should have focused on. And Robin Li, whether in public or in private, has always been reluctant to admit that Baidu has corporate politics.

At that time, Lu Qi left Baidu for two main reasons: first, Lu Qi's reform in Baidu entered the deep water zone, he could not shake the interests and rights of some people, and the reform was blocked; second, Li Yanhong may have hesitated again in the strategy of "ALL in AI" and could not give Lu Qi full support. Because of his wavering judgment on the company's strategy in the mobile era, he has made some internal employees lose confidence in their management ability.

Within the politically complex Baidu, Lu Qi had to leave on May 18, 2018, 486 days after joining Baidu.

A year after Lu Qi left, in May 2019, Xiang Hailong also left Baidu. Xiang, who is an old man of Baidu, joined in 2005. In April 2007, Xiang Hailong became the vice president of sales of Baidu, responsible for the national sales management of the company's bidding ranking business. In 2011, Xiang Hailong was transferred to vice president of business operation system of Baidu Company.

In other words, Xiang Hailong has always been the main person in charge of Baidu's sales business, that is, the main cashier of Baidu to make money.

After Lu Qi took office, he laid off the entire Baidu Medical Division, but with his departure, medical advertising has made a comeback. One of the rectification measures made to Hailong is to force the landing page of medical advertising to be hosted on Baidu's server. Baidu can directly intervene in the promotion content.

At that time, after Lu Qi parachuted to Xianghai Dragon head, Xiang was very confused and did not know where his position was. However, when Lu Qi left, Xiang became the exit of all contradictions.

However, whether it is Lu Qi or Xiang Hailong, they never held the financial power of Baidu during their greatest power. This is the bottom line for Robin Li to "step back behind the scenes." CFO must be independent of the system and report to him separately.

On January 20, 2018, Robin Li's wife, Ma Dongmin, announced that she would return to Baidu as a special assistant to CEO, responsible for Baidu's investment, human and financial work-to put it bluntly, Baidu's money and people are in the hands of the boss's wife.

Therefore, from the point of view of outsiders, Baidu's corporate politics and bureaucracy, as well as the control of personnel and financial power by Li and his wife, may be an important reason why Baidu keeps leaving, whether it is a new minister or a veteran.

5Has Baidu been underestimated?

In March this year, the Ministry of Industry and Information Technology issued a Chinese version of the classification standard for autopilot, which is divided into L5, L4 and L3 levels from high to low. If you reach the L4 level, the car can basically achieve a high degree of autopilot, while the L5 level can achieve full scene autopilot.

Among several new energy car dealers that are very popular in the capital market this year.TeslaAutopilot is still at L2 level, domesticWeilai和Ideal carAlthough it has been publicly announced that a large amount of research and development costs will be spent on L4 self-driving technology within two years, at least it is still at the conceptual stage.

Baidu cooperated with FAW to achieve mass production of L4 self-driving cars at this stage. Weimar also announced that next year it will mass produce a model with L4-level self-driving technology, which has functions such as automatic parking and summoning, and the self-driving technology is jointly developed by Weima and Baidu.

The latest information revealed by Li is that R & D investment in the field of artificial intelligence has accounted for 19% of Baidu's total revenue.

Many people believe that Baidu is undervalued to some extent. Expectations for the future of Baidu are concentrated in the field of "autopilot", which is the strategic information that Baidu reveals to the outside world from time to time.

However, judging objectively, self-driving will still be a relatively long new business, involving all aspects of resource integration, especially at the national policy level, and even the revision of relevant laws on transportation. It is unlikely that autopilot will be fully available in the market for three to five years, not to mention many technical problems that are increasingly difficult to overcome.

Regardless of whether Baidu is undervalued or not, at least for three to five years, Baidu has no new business areas that can bring substantial growth. Betting on AI is a major test of Baidu's strategic patience, and the more worrying question is whether Baidu can always maintain its leading position in this long period without being surpassed and subverted by the emerging waves.

references

1, Times weekend, "self-test Baidu self-driving: manual driving still requires a safety officer to switch between empty cars", October 16, 2020

2. Science and technology said, "how did Baidu become a fashion?" ", 21 May 2020

3. Lei Feng net, "Lu Qi left Baidu because he was dissatisfied with the combination of Robin Li and Ma Dongmin?" "May 19, 2018