Shulou(Shulou.com)06/02 Report--

Https://www.toutiao.com/a6643767307745100296/

Is there a company in the TMT field that has dared to admit that it has not studied, invested in and laid out AI-related businesses in the past two years? Is there a worker of the right age who pays minimal attention to media topics who has not heard or thought about the possibility that AI will replace him?

We are used to seeing the "AI popular science" in the clouds, and those philosophical discussions and technical analyses that are too high-level often make us resist the understanding of the development of AI during the difficult reading, feeling that this remote thing has nothing to do with us.

Furthermore, it blurs our understanding of the AI competition, and many people must have thought: what AI? This is not a new round of variants O2O, group buying, sharing economy, live broadcast, Internet finance. Do you gamble in groups?

However, AI competition is completely different from the above-mentioned "track craze", which is not some kind of "to VC" fooling, nor is it a simple "copy to China" innovation. The embrace of AI by Chinese TMT enterprises begins with the challenges posed by market factors and environment to every business organization.

1. Dividend disappears: the huge demographic dividend and mobile Internet traffic dividend in the past have enabled TMT companies to become natural market winners as long as they simply and rudely run fast. Now China's consumer Internet is becoming increasingly weak, and getting customers online is so expensive that companies point out that they can live a good life without promoting a taxi-hailing APP so that the whole people can pay with a QR code. How to improve business efficiency has become the most anxious thing for Chinese entrepreneurs who are used to "solving problems while developing".

two。 Higher employment costs: the extremely low employment costs in the past have allowed labor-intensive enterprises to lead the tide of opportunities. Forty years on, industrial upgrading has been accompanied by a declining working-age population and a growing reluctance of young people to engage in repetitive, monotonous jobs. And those business owners located in office buildings also have to pay for the rising cost of living and living in the city, and the cost of employment is rising.

3. The "traditional industries" outside the Internet are in urgent need of reform: industrial industries, including manufacturing, have always been forgotten by the "first half" reform of the Internet, in an era when Internet companies are overeating bonuses. Can only become news background and Internet upstart advertisers. This is not just like the new and hate the old, mainly because the previous Internet technology can not fundamentally empower the traditional industry, at best can only become a kind of marketing channel. But these traditional industries are still the foundation of the national economy and people's livelihood in terms of volume and value, and determine the background of an economy.

For this reason, AI has become a development path that can not be ignored by almost all TMT practitioners. Because whether it is reducing operating costs, improving business efficiency, or reforming traditional enterprises (developing industrial Internet), AI is almost the best solution for human society at present.

In 2013, Wu Enda and others, then a professor at Stanford University, published a paper saying that they used a cluster of graphics card chips that were not scarce and expensive to train computers to identify the "cat" species after watching YouTube videos for a week.

Since then, the road of human AI has finally embarked on the right path. After all, if the unique way of human cognition and thinking can be taught to computers through training, and as the amount of training increases, computers can beat the top human masters in complex operations such as go, then why not leave part of the work of a business organization to the computer?

This is the dawn that almost all TMT and business owners in traditional industries look forward to.

Hand-to-hand combat panorama: the fighting methods of the four giants

As a result, we saw Robin Li running up the Fifth Ring Road in Beijing in an unmanned car; Jack Ma called for 100 billion yuan in three years and set up a "Dharma House" to explore the future of science and technology; Ma Huateng inculcated in his internal letter: "in the past, it is often 'application demand to find technical support'. In the future, I believe that more and more innovations will come from 'technological breakthroughs to seek the landing of products'." Xu Zhijun, chairman of Huawei, directly dished out two AI chips in the fourth quarter of 2018, leading foreign competitors in terms of performance data.

We have also seen all kinds of industries, such as banking, insurance, cars, home appliances, security, retail, medical care, and even smartphones, which are a bit "traditional" at the moment, are embracing AI in an all-round way. If you open the APP of the bank where your payroll card is located, you will certainly jump out of the recommendation of "smart investment consultant"; many e-commerce sellers have used AI customer service to communicate with you; the development process of new cars and self-driving is vigorous; more and more appliances at home can be remotely controlled, password controlled and programmed by you; even straight men who no longer take pictures can easily use AI mobile phones to take photos to satisfy their girlfriends.

How to feel the development of AI tide in China more intuitively? Which enterprises are outstanding competitors, and how do the giants lay out strictly?

In order to help readers answer the above questions, Hu olfactory collected materials including Deloitte Consulting, Everbright Securities and other research newspapers, recent news reports of various enterprises, and so on, to draw up a panoramic picture of the AI war in China.

In this picture, we divide China's AI industry into three layers: application, technology and foundation. the lower layer is the premise and guarantee of the upper layer, and the upper layer is the practice and application of the next layer. In each layer, it is divided into several vertical areas, each of which is a representative enterprise. Among these enterprises, there may be investment and strategic partnerships arising from AI, which are shown in different colors in the picture.

Undoubtedly, the most conspicuous are the four giants in the center of the picture, as well as a large number of lines of the same color, as well as the enterprise circle of the same color at the end of the line. That symbolizes the AI positions built by the giants with money and technology: Ali, Baidu, Huawei and Tencent, which almost constitute the main force of AI in China. Completely "independent development", there are very few enterprises that do not have dealings with these four giants.

Nearly 65% of the AI participants in the picture come from these four powerful families. And the separatist princes outside the rich families, only JD.com, iFLYTEK, Shangtang, good future and other enterprises, and the advantages of these enterprises are very single.

In contrast, Baidu alone has occupied 48 seats in the chart, accounting for nearly 1/4; Tencent followed closely in second place, with 37 seats, reaching 1/5; Ali system accounted for 16%; and Huawei system accounted for only 4%. However, there is a solid layout in the upper, middle and lower levels of the industrial chain, and it is still a very strong participant.

Next, we will explain the panorama of the battlefield in detail in a hierarchical way.

AI application layer: the eight Immortals cross the sea, each hugging their thighs

Application layer, to put it simply, is the specific AI products that we ordinary users are most likely to feel intuitively. For example, the "AI key" on the mobile phone, the "smart speaker" chatting with you in the living room, traffic lights that can be dynamically adjusted according to traffic flow, vehicles that can drive themselves under certain conditions, and so on.

Although the layout of BAT in this layer is very extensive, there is still a clear sense of "biased science" in it.

For example, Alibaba's most advantageous AI areas are retail, finance and entertainment marketing, which is obviously closely related to the group's main business. Whether Ali cooperates with traditional retail formats such as Yintai and Da Runfa, or supports new species such as box horses, it hopes to prove that the online e-commerce model verified by Tmall on Taobao can efficiently empower offline. The AI investment in the field of entertainment marketing is like a large number of short video, social networking and live streaming applications that can be seamlessly plugged into Alimama and Taobao links, more like Ali's attempt to grab new traffic.

Since the figure of Robin Li on the Fifth Ring Road has been widely remembered by the outside world, Baidu's relationship with self-driving and newly built cars has become a solid label impression. Of course, in fact, this is indeed an area in which Baidu has an absolute advantage in the application layer of AI. Whether it is the mass production of L4 bus, or the large investment and cooperative planning for automakers, it is an important project that Baidu has unveiled in public in recent years. In addition, Baidu also has a good cooperative relationship in home appliances, mobile phones and other fields: Li Yanhong and Yu Chengdong hand in hand, Baidu enable Haier and other home appliance enterprises, have produced landing results.

Although Huawei's advantages in the application layer are mainly concentrated in the field of mobile terminals, the huge market volume and volume advantages of Huawei and Glory have made the industry fully believe that Huawei has occupied a favorable position in the palm of ordinary consumers in the path of "transfer from mobile Internet to AI Internet".

While Tencent is as happy as Baidu to hold hands with car companies, it also devotes a lot of energy to the field of education. Of course, there is some path dependence: Wechat and QQ are like unlimited pools of young users, enough to generate a lot of low-cost, high-quality traffic for AI education, and these users, after being gilded, can further feed Tencent's social and other value-added services.

In fact, in the current AI application layer war, we can clearly find a rule: although it looks like the eight Immortals cross the sea, they are actually "hugging their thighs." Ali's retail financial advantage is the thigh of e-commerce business, Tencent's educational advantage is social business, Huawei's terminal advantage, JD.com 's retail advantage, iFLYTEK's educational advantage and so on. Most enterprises still continue their original corporate genes, and then carry on the corresponding AI.

On the other hand, Baidu is an "outlier" in this layer, almost in a completely different way, giving birth to the fruits of self-driving and newly built cars from a gene that started with the recommendation of search engines and information flows.

Why did this happen? Let's move on to the next level, looking for answers.

AI Technical layer: the difference between Qizong and Jian Zong

In Jin Yong's novel The Smiling、Proud Wanderer, Huashan School is divided into two groups, one is Jian Zong and the other is Qi Zong. As the name implies, Jian Zong advocated swordsmanship first, and he could defeat the enemy by practicing swordsmanship. On the other hand, Qizong put more emphasis on "using Qi to control the sword", that is, to practice the internal force first and then learn the tricks before sharpening the knife without mistakenly cutting firewood.

Jian Zong is easy to become quick, and Qizong takes time-"Jianzong and Qizong have each practiced kung fu for ten years, so Jianzong must have the upper hand." after 20 years of practice, it was difficult for each of them to win. It was only after 20 years that Qizong's kung fu became stronger and stronger. By the time 30 years have passed, the kung fu of Jianzong will no longer be able to watch the back of Qizong. "

In the war chart of the AI technology layer, we can clearly find that the Chinese AI giants have separated Jian Zong and Qizong from this layer.

For example, Ali's advantage in the technical level is machine vision. Although Ali occupies the most seven seats on this floor, three of them belong to the field of machine vision (Shantang, absent sight, Yitu). Tencent, together with its younger brother Sogou, occupies a favorable position in the field of machine translation.

Machine vision and machine translation, both belong to "swordsmanship", the effect is very accurate, the speed is also fast. Machine vision can not only directly enable face recognition, security and other projects, but also assist Ali's idea of "unmanned retail". Machine translation can be directly used in Tencent's social products, and it is a natural partner of Tencent in the field of AI education application.

On the other hand, Baidu and Huawei belong to the representatives of Qizong. Although Baidu also has direct advantages such as "map-enabled autopilot" at the technical level, more advantages also belong to open platforms (such as Apollo) and operating systems (such as DuerOS). Huawei is devoting more energy to open platforms and the Internet of things.

From an intuitive point of view, the investment in these technical fields can not directly produce some kind of application results, still less can they directly endow their own advantage projects. However, this accumulation of technology can provide other enterprises or developers with a complete set of mature AI solutions, eliminating a lot of the pain of "repeating wheels". Even Baidu has put forward the concept of "rapid delivery in 30 days", believing that the process of upgrading an ordinary car to a smart car in the future will only take 30 days with the help of the Apollo platform. Huawei's recent main concept is also "inclusive AI": let AI step down from the altar and really become the basic productivity, so that every industry can afford to use it and use it well.

Robin Li once said, "our competitors do not buy technology companies, they only pay for products and user scale, not for technology, only Baidu pays for technology." Although this is absolute, the meaning is very clear. Jianzong is more willing to invest in technologies that directly empower products and gather business flow, while Airbender is more willing to delve into underlying and universal technology. is there any more appropriate "internal training" than building an open platform and operating system?

AI basic layer: who is Tuhao and who is moving bricks?

Entering the bottom of the AI war, we can instantly feel the power of capital. In the field of chip alone, Ali has shot four major players: Cambrian, Deep reference, Zhongtianwei and Nengneng. And Tencent's money is more invested in the field of big data, including Mingluo, carbon Cloud and Sipai.

These names may still be too unfamiliar to the average consumer. But what is no stranger to ordinary people is that if you take a walk through the major airports in China's first-tier cities, you are sure to see huge advertising posters of Aliyun and Tencent Cloud.

In fact, in October 2018, the Daily Telegraph reported that China's largest technology companies had surpassed Silicon Valley's technology giants in artificial intelligence and machine learning investment competitions-- the eight largest US and Chinese technology companies invested about $14 billion in AI, of which BAT accounted for $12.8 billion, while Tencent alone accounted for $10.5 billion (including overseas investment).

While there are tuhao, there are "bricks", this is not to say poor, but relatively more focused on self-research and internal technology iteration.

For example, in the field of algorithms, PaddlePaddle, the open source framework for deep learning, which is China's only independent intellectual property right, and Baidu brain, China's largest and most mature AI open neural network, is owned in the field of computing. Although Baidu does not have as high market share as Ali Yun, nor does it have a huge investment in the chip field, it actually monopolizes the other two core areas of the basic layer.

Huawei adheres to the idea that "what is not done is unspeakable". In October last year, Huawei directly threw out the AI chip with better performance than foreign competitors. Coupled with the Kirin chip, which has long been familiar to the industry and used in Huawei's mobile phone terminals, Huawei has become a domestic AI chip boss.

Ali's expectation is to digest and mature the bought technology and then promote it downstream. The idea of Baidu Huawei is to create the core underlying technology independently. This different AI basic layer path is in fact the root cause of the different performance of each enterprise in the above layers.

Overall, the performance of Chinese enterprises in the basic layer of AI still needs to be improved. At present, international giants such as NVIDIA, intel and Qualcomm still firmly control most of the market share represented by the chip field.

The advantage of Chinese companies is data, according to IDC estimates, the total amount of global data is expected to reach 44ZB in 2020, China's data volume will account for 18% of the global data. Such a huge amount of data can feed Chinese technology giants enough material to improve the accuracy of algorithms, models and products.

"All in AI" and "AI in All"

In 2017, Baidu and Tencent respectively put forward the concepts of "All in AI" and "AI in All" at two meetings hosted by themselves. At that time, many people feel that the formulation of Baidu is more or less solemn and heroic, and has some feelings of chicken blood. In contrast, the formulation of Tencent appears to be relatively light and appropriate.

Nearly two years later, combined with the panoramic picture of the current AI war, when we look back on these two routes at the beginning of 2019, we do feel the consistency of goals and results. Among China's four technology giants, Baidu is almost the only one that has really studied, invested in and developed the AI industry, and has achieved solid results at all levels. Although Huawei has its own outstanding areas at each level and its work is very solid, it started relatively late and the areas involved are relatively limited, so it will take time for a full exchange of fire between AI and Baidu.

(CB Insights: Baidu's AI strategy is earlier than Google, Ali and Tencent)

As for "Jian Zong" Ali and Tencent, they have fully implemented the method of using AI to strengthen their inherent advantages. The primary value of AI to these two giants is to optimize the business efficiency of itself and its partners, and then grab more traffic and consolidate traditional areas of advantage.

For example, Baidu hopes to become the best wheat grower, flour processor and bread baker through research on bread making, and form strategic alliances with seed dealers, fertilisers, builders and retailers in the process. Then Alibaba and Tencent prefer to buy bread products that suit their own taste, and quickly improve their e-commerce efficiency and social efficiency, and then further consider how to enable other enterprises in the same industry.

Of course, Baidu does not completely break away from its traditional advantages and go to AI. After all, since I started as a search engine in 2000, it has been Baidu's job to master data algorithms that are closely related to people, and to understand user needs and web content through natural language processing technology. Just as Google became the first American Internet giant to pry up the value of AI, AI may not be a search, but the search giant naturally has the AI gene.

However, Baidu and Huawei's "Qizong" route is not necessarily the best choice, at least for Baidu is relatively "at a loss". Because most of its technology is to B, with the exception of small speakers and self-driving cars, it is difficult for ordinary consumers to feel the AI value of Baidu. Huawei's concept of Pratt & Whitney AI also needs more reporting and testimony in its promotion. And Ali uses its own retail channel advantages to push the sales of Tmall series of intelligent hardware to the forefront of the country. Tencent's VIPKID, Youbixuan and other enterprises have good C-end awareness and media exposure.

In short, AI is the dawn of the second half of the Internet, and the current route and situation have been clear, let's see how the giants fight to the end and keep their advantages until the dawn.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.