Let’s go back to the Baidu AI Developers Conference in 2017.
COO Lu Qi took the lead in making his debut. At 10 o’clock in the morning, he stepped onto the stage on time, occupying the field of vision of 5000 spectators, and blasting a value judgment into the minds of on-site and off-site users:AI will bring mankind into a new era just like the previous three industrial revolutions.
Lu Qi’s voice is firm, powerful and full of confidence, and there is no doubt: "We can definitely lead!"
"AI is an opportunity for Baidu, All In AI."With a wave of his right hand, Lu Qi put his index finger up, calmly "stud" and bet on the whole Baidu.
Luqi
At the end of the meeting, Liu Ji wanting more, announced on the spot:In 2020, Baidu can realize automatic driving of high-speed and urban roads.
Ok, now we can travel back to May 2020. At this time, Lu Qi has left Baidu for two years, and the situation of the artificial intelligence industry has changed greatly.Automatic driving on the high speed is far away.
Wave Computing, a foreign star AI company, laid off its employees and filed for bankruptcy protection. Horizon, a domestic AI head company, was laid off by 50%. At the beginning of this year, the report of Cheetah Global Think Tank showed that the financing amount of AI enterprises experienced a cliff-like decline after 18 years of continuous increase. Compared with 2018, the amount of financing in 2019 decreased from 148.45 billion to 96.727 billion, a decrease of 34.8%, and the amount of financing decreased from 737 to 431, a decrease of 40%.
Since the second half of 2019, IPO plans have been successively spread from Defiance Technology, Shangtang Technology, Yuncong Technology, Excellent Choice and Cambrian. Regrettably, after six months’ application for listing, the status of Defiance was still invalid. Shang Tang was delayed to go public and turned to private equity market for financing. According to the "investment circle" report, Defiance and Shang Tang were frustrated in listing one after another because the valuation was not recognized. CAMBRIAN was asked 20 questions by Shanghai Stock Exchange after applying for listing in science and technology innovation board.
The capital market, which was greeted with smiles before, was completely faceless. Gone are the days when artificial intelligence companies could get high valuations by putting AI tags on PPT two years ago.
At this time, Baidu was caught in a full-scale war with ByteDance. Li Yanhong should have discovered it long ago.All In AI is not an opportunity for Baidu, but a big pit for Baidu.
Today’s situation is particularly clear:The era has emerged, that is, the great migration of content competition from graphic to video. The new traffic generated by this great migration will profoundly affect the entire Internet industry in China. Which big company dominates this great migration will be able to hold the gate of traffic distribution in its own hands, just as Tencent has done in the past 10 years.
Baidu has fallen behind from BAT, which has long been a recognized fact. Today, the market value of Meituan exceeds 100 billion US dollars, and the most unpleasant thing may be Li Yanhong. Although the news of "Baidu delisting" was officially denied, it may not be without reason.
If you want to restore your former glory and re-tie with Erma,The ongoing content ecological war is just one of the few opportunities for Li Yanhong, if not the only one.
Li Yanhong has obviously realized this, so he will vigorously support the live broadcast platform. Baidu executives announced that they will give live broadcast of 10 billion traffic+500 million cash support. However, because it is too late to launch, live broadcast and even the whole short video field, it is already an almost impossible task for Baidu to make a preemptive strike.
If in 2017, before Aauto Quicker and Tik Tok were exposed, Baidu All in was not AI, but content ecology, would the result be different?
The slogan "All in AI" reveals the background color of Lu Qi’s character.
At the end of last year, Lu Qi accepted a late interview and made no secret that he was too confident and was considered "self-righteous" by his wife.
This is Lu Qi’s reflection after he returned to China for three years and suffered setbacks in his two career choices.
But in 2017, ALLINAI-this almost gambling action is undoubtedly a stimulant in front of the lively capital market.
Lu Qi, who put forward this slogan, has been well-known in the Chinese and American science and technology circles for many years, and his resume can’t be picked out-postdoctoral fellow at Carnegie Mellon University, executive vice president of Yahoo and global executive vice president of Microsoft, known as "the most powerful Chinese in Silicon Valley".
It seems that the time has indeed come: in 2016, the "man-machine war"-Alpha Dog defeated Li Shishi in the ninth stage of Go, turned AI, which was originally confined to the fields of science and science fiction, into a popular hot topic on everyone’s lips.
The geographical location has also been created: outside the Fifth Ring Road in Beijing, Lu Qi sat in the Baidu AI conference venue, and the real owner of Baidu, Li Yanhong, performed a stunning scene of unmanned driving outside the Fifth Ring Road. Despite being issued a ticket, compared with the seemingly upcoming driverless era, such a price is completely insignificant.
That night, Baidu’s market value rose by nearly $18 billion.
Half a month later, the State Council released the "New Generation Artificial Intelligence Development Plan", which added another fire to the development of AI at the policy level, and the market quickly detonated. Ma Yun founded the Dharma Institute, Ma Huateng proposed AI in all, and Huawei released the full-stack full-scenario AI strategy. Investors throw money into the wind.
In July of the same year, Shangtang Technology raised $410 million, which was the highest financing record in the world at that time. Three months later, Defiance Technology completed $460 million in financing. At this point, AI Company entered the financing race mode and its records were constantly updated. In May 2018, Shangtang Technology raised $620 million, and Ubisoft completed $820 million. In June, Yunzi raised $1 billion from technology, in July, Shang Tang received $1 billion from Softbank, and in November, Horizon raised $1 billion …
Under the fierce fire, the market soon proved that everyone overestimated the commercial liquidity of AI. At the beginning of 2018, Kai-Fu Lee bluntly said in an interview with China International Television: The bubble of AI will burst. "Every entrepreneur wants to package his startup into an AI company, and every VC wants to flaunt himself as an AI investor. But AI investment is actually not suitable for novices to participate. If you don’t understand the AI ? ? technology itself, you are eager to participate in the early stage of AI, and such people will lose even their pants. "
Baidu’s complex organizational structure casts a shadow over the All in AI strategy.
Lin Ning, a former Baidu cloud worker, said frankly: "Baidu feed stream, driverless and cloud technology all belong to AI, but all departments are running in parallel and are not harmonious with each other."
On the one hand, Li Yanhong led a team to do the familiar and profitable search +feed streaming business, on the other hand, Lu Qi focused on betting on driverless driving.
Both of them are commercialized. Even by 2020, looking at the world, the goal of unmanned commercial mass production is still at the conceptual stage.
Taking Waymo, a driverless company owned by Google, as a reference, Waymo started research and development in 2009, and completed the first phase-10 billion mileage test in 10 years. The mass production stage of the second stage is still stuck.
According to Waymo’s good news at the beginning of this year, it raised 2.25 billion US dollars to produce 80,000 cars. However, according to a survey from the media "high-tech smart car", Waymo’s mass production is unrealistic. Waymo invested more than 5 billion US dollars in research and development in the first ten years. Considering the hardware and software costs of 80,000 vehicles, 2.25 billion US dollars can only be regarded as a drop in the bucket.
Back in the domestic market, the situation of driverless driving is even less optimistic. A Baidu employee told the alphabet list that at present, Baidu’s driverless driving is only used in a limited range. "The domestic landing conditions are immature and the social infrastructure development is not matched, so it is difficult to land on a large scale."
According to Li Deyi, an academician of China Academy of Engineering,Driverless driving will be in the product incubation period before 2025, and mass production is expected to be in 2060.
Driverless driving is difficult to realize, and the profit model of search +feed has already run through, and Li Yanhong’s choice does not need to be guessed at all.
Q1 financial report released in April 2018 shows that revenue increased by 31% year-on-year and net profit was RMB 6.7 billion, up by 277% year-on-year. The unmanned business has no revenue, and search and feed streams have become profit anchors. Asked about the commercialization of AI in the earnings conference call, Lu Qi’s response was vague, and he hoped for the future between the lines.
If you pull the camera back to April 2017 and Lu Qigang joined Baidu’s three-month earnings conference call, his response will be much simpler-face the commercialization of driverless technology raised by analysts and give a clear answer:Realize the commercialization of driverless cars in 2020.
When he joined Baidu in early 2017, Lu Qi shared an experience: jumping on the right boat at the right time. By the late interview at the end of 2019, his quotation added one more sentence: take a step back and broaden the horizon.
In Lin Ning’s view,It is inevitable for Lu Qi to leave Baidu. "Lu Qi chose a track that Baidu is not good at."
Lin Ning said frankly that driverless driving is a very good AI landing scene, and Baidu’s technology is also very good, but the Apollo program is just a famous sound, and the products can’t land at all. "BMW, Mercedes-Benz, Baidu and Apple are all driverless, and what do Internet companies take? Compared with car manufacturers?"
More importantly, this led Baidu to take a fork in the road in strategy and completely fall behind in content layout.
The core of Baidu is search, and the essence of search is content.Baidu tried to rely on AI to make a turnaround, so that it was underinvested and lacked foresight in the main channel of the times-content ecological competition.Throughout 2016, Ali announced the creation of "super consumer media". Tencent has Daily Express and Tencent News to seize the consulting market. The headline is 1 billion, and it is necessary to focus on supporting short videos …
Although Baidu has already made preparations. In 2016, he also laid out 1 billion layouts of 100 numbers. At the media meeting to welcome Lu Qi in January 2017, Li Yanhong emphasized the importance of information flow. "I am confident that users can rely on Baidu to obtain information." In the end, the center of gravity fell on AI. Li Yanhong received a ticket at the Fifth Ring Road, Lu Qi used the Apollo plan to ripen the AI bubble, and the All in AI strategy can be said to be of no help in revenue.
On the other hand, in 2017, advertising revenue reached 15 billion yuan, accumulating hundreds of millions of creators around the world, and hitting the target of 50 billion in 2018. At that time, there were 1 million creators in the same starting point, and the advertising revenue of information flow was less than 10 billion.
In January 2018, before Lu Qi left Baidu, Li Yanhong attended the geek conference, in addition to emphasizing himself.Never said "All in AI"In addition, information flow is specifically mentioned. "Information flow is the focus of Baidu. Most of Baidu’s resources are on Baidu search and Baidu feed stream."
Li yanhong
Looking back now, reassessing Lu Qi’s choice to leave Baidu can be said to be the best of both worlds.
Lu Qi became an investor and joined YC. Later, he founded Miracle, laid out AI consistently and did what he liked.
When Lu Qi left, public opinion fell to the ground and felt sorry for Lu Qi. With data, Baidu couldn’t keep people. In 10 years, he left at least 10 vice presidents and more than 20 executives. Rumors such as internal palace fights and couples seizing power are rampant. Coupled with the bad impression of hemophilia bar and Wei Zexi incident to the public, Baidu was simply sprayed into a sieve.
But I have to say,It is wise for Baidu to loosen the throttle and win Lu Qi in AI strategy, at least it is better late than never.
After Lu Qi left, the voice of Baidu All in AI disappeared. In the Q3 financial report of 2018, the revenue of search and feed stream business accounted for 80% of the total revenue. At the Baidu World Congress held in the same year, the whole package revolved around the content ecology, such as empowering creators, supporting short videos and landing applets, and upgrading Baidu APP into a "comprehensive content consumption and service platform". It also played a tit-for-tat with the headlines, and the frequency of infringement lawsuits by both sides was much higher.
A few days ago, Baidu launched the "Gathering Energy Plan" to supplement the live broadcast business, which can be said to be the foundation of a comprehensive return to the content ecology.
At the same time, the implementation of Baidu’s driverless strategy is obviously more rational. "Driverless gambling is the future, and the return is too far. At present, we can only reduce the research and development costs slowly." A former Baidu employee said.
At the Baidu AI Developers Conference in 2018, Baidu announced the official mass production of the Apollo 3.0 version "Apollon"-"the world’s first L4-class self-driving car". However, compared with Waymo’s mass production of 80,000 vehicles, Baidu’s mass production of "Apollon" is a little shy, with only 100 vehicles. The activity space can only be limited to relatively closed and fixed-line environments such as parks, airports, scenic spots and factories.
Social supervision also needs Baidu to take it slowly. In 2019, Baidu L4 driverless taxis were put into trial production in Changsha. According to national business daily, drivers are on standby at the driver’s position of driverless taxis. These drivers, also known as "security officers", are prepared in accordance with official requirements to ensure the safety of passengers.
According to the unmanned management measures introduced in Shanghai last September, such as vehicles must be equipped with drivers, enterprises must not charge passengers, and must not engage in transportation or business-related activities. This also means that it will take N years for the enterprise’s driverless products to be truly commercialized.
It is not difficult to understand that although Baidu will definitely mention AI in every financial report, people in the industry know that Baidu’s revenue can only rely on content business.
At the end of February this year, the annual financial report with a total revenue of 107.4 billion yuan in 2019 was released, and the search +feed stream sang the leading role. The unmanned appearance only emphasizes business progress, such as cooperating with 23 cities and getting 40 autonomous driving licenses in Beijing. Revenue data has not been disclosed so far.
By May 19th, the latest Q1 financial report with revenue of 22.5 billion yuan and net profit of 3.1 billion yuan, the main force of Baidu’s revenue has not changed, and the financial data of AI business is still missing. The driverless business is completely playing soy sauce in the financial report, and the cloud business has been taken over. The only AI product that is emphasized is only a small one, a smart speaker.
Baidu’s development path in AI strategy is almost the commonality of domestic AI companies.
"Technical improvement is difficult to completely solve the fundamental defects of current AI, which determine that its application space is limited to specific fields-most of them are concentrated in image recognition and speech recognition." This is the conclusion that Zhang Yue, president of Tsinghua University Institute of Artificial Intelligence, accepted in Economic Observer last May.
At the same time, another conclusion of Zhang Yue is that artificial intelligence based on deep learning has technically touched the ceiling.
This seems hard to understand. As far as we can see, AI products appearing on the market, except intelligent robots, face recognition, intelligent customer service and intelligent speakers, have hardly seen products with more advanced technology research and development. Why is the ceiling of AI so low?
However, the reality is in front of us. From the "man-machine war" in 2016, there has been a lot of talk about AI replacing labor. Kai-fu Lee once asserted that 50% of human work will be replaced by AI within 15 years, but four years later, the development process of domestic AI is not optimistic.
At the AI International Conference in June last year, Ma Ziya, Intel’s technical director, said frankly: Emerging technologies such as big data and AI are difficult to land. At present, only 2% of the data is really used to help human production and life.
Back to the AI commercial market. Even in Ali and JD.COM, which have scene advantages, only basic products such as robots and intelligent customer service have been widely used, and more intelligent technologies are still in the conceptual stage.
A typical example is Iflytek, the leading brother in the field of intelligent voice. In September, 2018, I was questioned about "AI simultaneous interpretation fraud". Finally, Liu Qingfeng, the chairman of the board, delivered the hammer himself, claiming that the intelligence level of the current machine could not replace simultaneous interpretation for the time being. By October, CCTV revealed that Iflytek had illegally occupied the Chinese alligator National Nature Reserve in Xuancheng City, Hui Province, and was suspected of using land to build villas under the guise of "AI".
The financing boom that was rolled up like a storm in the past is now flying everywhere.
AI Unicorn Shangtang Technology has raised nearly $3 billion in 9 rounds in 6 years, and its valuation once exceeded $7 billion. Its business covers smart cities, smart phones, pan-cultural entertainment, smart cars, smart health, smart business, education and advertising. Since 2018, it has frequently exposed packaged and sold assets, which has been questioned by observers and has not formed core technical barriers so far.
The company, which competed with Shang Tang for the dominant position in financing, set a new loss record year by year, from a loss of 343 million in 2016 to a direct loss of 5.2 billion in the first half of 2019, and was forced to fall on the road of IPO of Hong Kong stocks.
The Cambrian lost nearly 1.6 billion yuan in three years. In March this year, it submitted an application for IPO in science and technology innovation board, and planned to issue 40.1 million shares, raise at least 2.8 billion yuan, and value itself at 28 billion yuan. Even it was not afraid to show its cards in the prospectus: it could not guarantee profitability in the next few years, and it might face the risk of delisting after listing.
……
Looking back at AI now, the attitude of investors is the most intuitive. Wu Shu, the founder of Qushi Capital, has been paying attention to AI since 2011. He told the alphabet list (ID:wujicaijing): "Many AI directions are promising and valuable, and business models have gradually emerged, butPart of AI was speculated by a group of irrational investment institutions in the early stage, which led to the high valuation deviating from the value, and continued investment became a take-over man. In the end, no one could take it."
Lu Qi, who is also an investor, not only admitted that he was overconfident, but also listed the investment projects he participated in, such as robotics, biotechnology, telecommuting and cloud computing technology.Only there is no unmanned driving.
The last time I responded to the question about driverless driving was in a late interview at the end of last year. When I was asked about the three most optimistic AI landing scenarios, Lu Qi only gave two answers-smart place and new mobile (robot driving).
Facing the reporter’s innuendo on the issue of unmanned driving, Lu Qi said:It is a long process for unmanned driving to completely land in commercialization.
(Lin Ning is a pseudonym)
[Some References]
Dialogue with Lu Qi: It is inevitable to do something long-term and valuable and not be understood, later, December 13, 2019.
"Baidu executives interpret the first quarter financial report, Lu Qi said that it is necessary to strengthen AI innovation", Sina Technology, April 28, 2017;
"Lu Qi: An" outsider "who broke through the fog of Baidu", people, June 28, 2018;
"Shenwang | Baidu’s" new "era after Lu Qi’s curtain call", Shenwang, June 7, 2018.
"Lu qi’s reform is not completed. Baidu returns to the old road? ",Beijing News, May 23, 2018
"Unmanned driving enters the" mass production era "? Dawn at dawn, or a crazy daydream, high-tech smart car, March 8, 2020.
Li Deyi: mass production of driverless cars is expected to be in 2060, China Auto News, May 22, 2019.
"AI Company, Began to Bankruptcy", investment community on May 6, 2020.
"exclusive interview with academician Zhang Wei: deep learning touches the ceiling, and the AI miracle is hard to continue", Economic Observer, May 28, 2019.
Shangtang Technology was exposed to IPO: the listing is uncertain after nine rounds of financing, Blue Whale TMT, March 27, 2020.