The hottest in-depth interpretation of 2018 artifi

In depth interpretation of 2018 artificial intelligence: polarization, talent competition upgrading, internationalization, policy sinking

looking back over the past year, in the context of Sino US trade war, national financial deleveraging, and even under the whirlwind of blockchain at the beginning of the year, it did not fall in the cold winter like several waves of technology waves such as intelligent hardware and VR, but rose against the trend in the wind and rain

polarization: the industry pattern of "one super and many strong" has gradually formed.

reviewing the history of Internet development, the rise of a round of new air outlets will experience the process of incubation period, outbreak period and stability period. For example, in the early smart market blowout, the traditional Nokia, blackberry, LG, HTC and emerging, Xiaomi, oppo/vivo, Huawei, 360, Meizu and other manufacturers fought for many years, and finally formed an industry pattern dominated by apple, Huawei, Samsung, Xiaomi and oppo/vivo. Another example is the trend of sharing bicycles. In the early stage, dozens of sharing bicycles rushed forward, and friends once commented that "there are too many bike brands and colors are not enough", but now the sharing bike market is facing a reshuffle and is about to enter a stable period

taking history as a mirror, we can know the rise and fall. Although AI, intelligence and shared bicycle belong to different fields and technical forms, the economic law will not change. After the capital frenzy in 2017, artificial intelligence gradually entered a "stable period" in 2018. It is a year dominated by the implementation of technology, a year of territorial expansion, and an increasingly clear industry pattern

AI has two main tracks: CV track and voice track. The polarization of CV track is very obvious, basically forming a "four dragons" pattern led by "Shangtang, Kuangshi, Yuncong and Yitu". Since 2018, we have rarely seen major breakthroughs in financing and Technology (including top competitions such as ECCV and megaface) by other AI companies, but the number of these CV companies is no less than 100

why is Matthew effect so obvious? AI BusinessWeek summed up three reasons:

AI business model is heavy. Compared with the mode of Internet entrepreneurship, the business mode of artificial intelligence entrepreneurship is very heavy, including computing power, algorithm talents, and the cost of obtaining high-quality data, which is much higher than that of Internet in the past. It is a field of heavy capital investment. In addition, a careful analysis of several major AI start-ups found that they all developed from the integration of production, learning and research, creating a closed-loop ecosystem from academic research, laboratory to industrial application. Admittedly, it is difficult for small companies to play with such a heavy business form

talent monopoly. According to element AI, according to LinkedIn data, there are about 22000 AI practitioners and researchers with doctoral degrees or above worldwide, and only about 600 in China. In China, it is almost monopolized by several leading enterprises. For example, Shangtang has more than 150 AI doctors, accounting for 1/4 of the total number of AI doctors in China, as well as master figures such as Tang Xiaoou and Wang Xiaogang. Other AI unicorns are not inferior in talent. There are talented teenagers such as Yinqi and Sun Jian, and academician Yao Qizhi is the chief consultant; Dr. Zhou Xi, the founder of Yuncong technology, once studied from Professor MAS S. Huang xutao, an academician of the Fourth Academy of Sciences and the father of computer vision, who also recruited executives from apple and Facebook in May this year; Yitu founder Zhu long is a favorite student of Yann Lecun, the ancestor of deep learning, and co-founder Lin Chenxi was the first technical director of Alibaba cloud

when leading enterprises monopolize talents, small and medium-sized start-ups seem to be cut off from the water source, although there is no way to return to heaven. As capital is not so fond of small and medium-sized start-ups, plus this year's financial leverage, it is extremely difficult to get investors' money

founder thinking. Most of the founders of the AI track came from technology, either originally engaged in academic research, or originally were the technology leaders of large enterprises. If you can't completely jump out of the habit of academic thinking, you will suffer a loss in commercialization. There are already some examples in the industry. The technology is very strong, and the company started very early, but it falters in the commercial landing. Ma Yun once said, "the most difficult thing for the company is the realm of the founder"

in the strong pole of "polarization", there is a pattern of one super strong and many strong. Shangtang currently has a total financing of more than $1.6 billion, ranking first among the "four dragons", followed by Kuangshi, with more than $600million, less than one third of Shangtang. Investors are the smartest group of people. They will make a very in-depth investigation before deciding to invest in a company. Behind Shangtang's high financing, it has not only the technical advantages, but also the business model

Shangtang can be rated as the new generation of artificial intelligence open innovation platform of "intelligent vision", and its visual algorithm is certainly leading the industry. It is worth noting that Shangtang is the only self built Supercomputing Center, with a peak of 1billion times, and the light of Taihu Lake in China is only 2billion times. The self built Supercomputing Center will accelerate the iteration speed of the algorithm, which will show a stronger explosive force in the later development. In terms of commercial implementation, Shangtang has been fully opened in security, mutual entertainment, education and other industries, with a total of more than 700 cooperative enterprises, and its revenue is expected to increase several times in 2018

the pattern of the voice track is quite similar to that of the CV track. IFLYTEK, yunzhisheng, spitch and go out to ask four enterprises monopolize the voice market, while other companies have little voice, and the market as a whole presents a polarized form. In the strong pole of "polarization", there is a pattern of one superpower and many strong ones. IFLYTEK has fully opened up in many industries such as education, robotics, real estate, etc., while several other companies have taken a differentiated mode in their respective sub circuits

upgrading of talent competition: from recruiting to cultivating people

the competition for entrepreneurship is ultimately the competition for talents. This phenomenon is more and more obvious in the field of artificial intelligence. The data shows that the gap of artificial intelligence talents in China is more than 5million

since 2017, giant companies including bat have begun to vigorously attract talents, and many industry leaders have joined the entrepreneurial wave. AI BusinessWeek counted the turnover profile of top talents in the industry in, as follows:

in the past two years, the turnover of top talents in AI was as high as 25. 2017 was the craziest year for talent recruitment, and 2018 was relatively reduced. Because top AI talents are very scarce, the total number of talents that can be tapped remains the same. "The scarcity of high-level AI talents is a worldwide problem. Many of our enterprises are now digging people with heavy money, but in fact, digging people can't bring increment." Professor zhouzhihua said at the "2018 jd.com AI Innovation Summit"

since this year, the talent competition in the industry has also been "transforming and upgrading": from recruiting people to cultivating people. Zhou Zhihua said: we should start from the source and cultivate high-level AI talents for the development of the country, society and industry

not only colleges and universities make great efforts to cultivate talents, but the business community also holds competitions to cultivate people through in-depth cooperation with colleges and universities. For example, at the 2018 Huawei connect conference, Huawei cloud launched the "fertile soil AI Developer Program", invested 1billion yuan to cultivate AI talents in universities and research institutes, and reached AI cooperation with Tsinghua University, the Chinese Academy of Sciences, the University of science and technology of China, Zhejiang University, etc; In addition, Huawei also provides each AI developer with 20 hours of free online training, 21 days of AI training camp, irregular AI developer competitions, etc

in August 2018, Shangtang technology, together with the Chinese University of Hong Kong, Amazon, Nanyang Technological University and the University of Sydney, jointly organized the first wide face and pedian challenge 2018 (referred to as: wide (for example: rope sample challenge with a diameter less than 1mm), including the Institute of computing of the Chinese Academy of Sciences, Microsoft Research Asia, Peking University, Carnegie Mellon University, the University of Hong Kong, Yahoo, jd.com, Kuangshi, iFLYTEK Didi and other technology companies participated

in addition, the AI Challenger global AI challenge, hosted by Innovation workshop, Sogou, meituan review, Meitu company, and co organized by Beijing frontier international AI research institute, Peking University, deepdrive, Berkeley, California, has more than 20000 participating teams from 81 countries

behind the artificial intelligence competitions held by the industry, there is a true reflection of the thirst for talent. The cost of recruiting people through headhunting is getting higher and higher. It is better to explore potential talents through competitions or cultivate talents by yourself. This trend of "cultivating people" will continue in 2019

policy sinking: local government checkpoint AI

in the past decade of Internet entrepreneurship, there has been an influx of new technologies, but we have not seen the government pay such attention to AI. 2018 is the first year of AI strongly supported by the government, especially local government checkpoint AI

since 2017, provinces and cities have promulgated AI development plans. According to the report of the prospective industry research institute, as of March 2018, 15 of the 31 provinces and cities across the country have issued AI plans, of which 12 have formulated specific industrial scale development goals. The scale target of these 12 provinces and cities in 2020 will reach 429billion yuan, while the scale of Beijing's artificial intelligence core industry will be 50billion yuan, accounting for only 11.7%

from the above data, we can see that AI is far from being dominated by first tier cities, and the second, third and fourth tier cities show great market potential. For example, the scale of AI core industry in Sichuan is twice that of Beijing. Artificial intelligence of local governments will greatly promote the landing of artificial intelligence industry, including smart city, education and many other scenarios

before the opening of Tencent global partners conference in 2018, Ma Huateng stressed in an open letter that the first half of mobile Internet is nearing the end, and the prologue of the second half is beginning. With the digitalization process, the main battlefield of mobile Internet is developing from consumer Internet in the first half to industrial Internet in the second half

interconnection is a technology wave driven by C-end users. If there is a strong capital drive, it is easier to form a monopoly. However, industrial interconnection is a to-b end, which needs to provide personalized services, and it is difficult to quickly replicate large-scale. It is a comprehensive competition of technology, products, services and capital. The same is true of the competition between cities. In the Internet era, most leading enterprises have gathered in the north, Shanghai and Shenzhen, but in the era of industrial interconnection, it may be that cities rely on their own characteristic industries to form a future pattern of differentiation and checks and balances

it is difficult to guarantee the product quality in AI 2.0 generation: international competition

when startups develop to a certain stage, they will face international challenges, such as didi and Xiaomi are empowering internationalization. Although the voice of "internationalization" of AI companies is not high in 2018, many enterprises have made preparations for internationalization and may make full efforts in 2019

Kuangshi set up a research institute in Seattle very early. In January, 2018, eatO technology announced the establishment of its first international office in Singapore. Shangtang is more nationalized and has set up branches in Kyoto, Tokyo, Singapore, Silicon Valley and New Jersey

in his speech at the main forum of the 2018 AI Conference on September 17, Liu duo, President of the China Academy of information and communications, said that the current trend of AI globalization is obvious, but different regions also show corresponding characteristics. The United States is still one of the core birthplaces of AI, with basic theory, core platform, application technology and advantageous enterprises. Canada with technology