The BBC predicts that the global artificial intelligence market will reach11900 million yuan in 2020. In addition, the iteration and development scale of AI-driven cloud services, big data analysis, mobile Internet, Internet of Things and other industries is immeasurable.
BAT, which is at the forefront of the internet, will naturally not miss such an opportunity. To this end, the author plans three special reports to comprehensively analyze the layout and latest progress of BAT in the field of artificial intelligence.
Today, let's talk about Tencent Empire, which kidnapped hundreds of millions of users through products such as WeChat and QQ.
In BAT, Tencent's AI layout was relatively late and more business-driven. After that, it gradually increased its investment and blossomed in an all-round way, mainly around three paths:
1. Based on Tencent's core products and technological advantages, different business systems have made breakthroughs in artificial intelligence.
2. Secretly set up research teams and laboratories focusing on the underlying foundation to accumulate long-term technology.
3. Pursue and invest in a number of outstanding companies in the field of artificial intelligence at home and abroad.
First, pay for the enclosure: biased overseas
Most of Tencent's huge investment M&A list is based on pan-entertainment, medical health, games and automobile transportation. For AI start-ups, the space is small, but it is closely related to the main fields. Investment projects are biased towards overseas areas, and the investment rounds are mostly angels and early A rounds.
In the past two years, the investment projects in the AI field disclosed by Tencent are as follows:
1)Diffbot
Diffbot is a data company located in the Bay Area, which was founded on 20 12. It mainly uses artificial intelligence technology to let' machines' capture the key content of web pages and output structured data that can be directly recognized by software. This year, Tencent and Felicis Ventures, a venture capital institution in Silicon Valley, led a round of Diffbot10 million US dollars.
2) Carbon clouds
Carbon Cloud Intelligence, whose Chinese name is Carbon Cloud Intelligence, was founded by Wang Jun, the former managing partner of BGI. Its business field continues BGI health care, focusing on the research of digital life ecosystem created by life big data, Internet and artificial intelligence.
In April this year, Carbon Cloud Intelligent completed the Series A financing of nearly 654.38 billion yuan, with Tencent, Zhongyuan Concord and Tianfu Group as the main investors.
3)CloudMedx
CloudMedX is a medical health data collection and analysis service provider, which collects data of different patients in many hospitals in real time, and provides health prediction and analysis in line with HIPAA (Health Insurance Circulation and Responsibility Act) to medical institutions after processing and analysis. After analyzing the original patient data, CloudMedx can reflect the development trend, occurrence mode, deviation and predicted possible results of some diseases, thus providing very useful reference for clinical treatment and early detection and diagnosis.
2065438+In May 2005, Tencent injected $6.3 million in angel round investment into CloudMedx.
4)Skymind
Skymind was founded by Adam Gibson in June 20 14, and its main business is the open source project DeepLearning4j. Skymind mainly provides support services for companies that use DeepLearning4j, integrates deep learning into Hadoop system, and allows users to store, process and analyze various types of data.
2065438+In April 2005, Tencent injected millions of dollars of angel round investment into Skymind.
5) Scale deviation
Scaled inference was founded by two former Google engineers on 20 14. It is a cloud computing service for all individuals. Develop machine learning and artificial intelligence technologies used by companies such as Google and develop them into cloud computing services that anyone can use.
2065438+July 2004, Tencent invested in large-scale reasoning.
6) sogou
This investment case can be traced back to 20 13. Tencent invested $448 million in sogou, and incorporated its search and QQ input methods into its existing business in sogou. The new sogou will continue to operate independently as a subsidiary of Sohu.
Tencent's investment at that time was more like a continuation of the 3Q war, which was based on the background that its search business did not improve. However, in the development of the following years, sogou gradually developed from a search and input method product company to a voice intelligence and big data company.
In addition, Yao Xing, senior vice president of Tencent, once said that he and Tencent's investment and M&A departments have reached an understanding and launched a large number of machine learning platform startups in the United States. One is that there are not many technology companies like China; Second, the acquisition of such a company can quickly make up for Tencent's shortcomings in the field of algorithms.
Second, basic research: scattered around.
In the field of basic research, Tencent has established different laboratories and business teams in combination with major business advantages and field directions such as WeChat, QQ and finance. While serving its own business, it also launched industry products in related fields.
1) Hong Kong: What laboratory?
Compared with the university resources gathered in Beijing and Hangzhou, young Shenzhen is slightly thin. At the end of last year, Tencent and HKUST crossed the ocean and set up the' WeChat-Hong Kong University of Science and Technology Artificial Intelligence Joint Laboratory' in a low-key manner, which is abbreviated as WHAT LAB in English. Taking artificial intelligence as the main research direction, it aims to improve users' life service experience and expand the boundaries of machine learning with big data.
In addition, Tencent Foundation donated 6,543,800 yuan to promote HKUST's scientific research innovation, and through in-depth cooperation between the two sides, an effective exchange and interaction platform in Industry-University-Research was built to promote the transformation of application results.
2) Shanghai &; Hefei: Youtu Laboratory
Tencent Youtu belongs to Tencent's social network business group, which was gradually established on 20 12 and is located in Shanghai and Hefei. Carry out technical research and development and business landing around the fields of image processing, pattern recognition, machine learning and data mining.
After several years of development, Tencent Youtu has gradually accumulated the world's top face technologies such as face detection, facial features registration, feature extraction and comparison, and live detection. Combining voice, lip language and facial expressions,' Youtu Face Recognition' can provide three-dimensional protection against attacks such as photos, videos and head models.
In June, in the international authoritative face recognition database LFW, Youtu set a new record with a score of 99.65%, surpassing top teams such as Linkface, Face++, Facebook and Google.
The Shanghai team has provided image technical support for more than 50 businesses of Tencent, and the technology mainly serves QQ space, watermark camera, daily P-picture and other products. Hefei team's related technologies mainly serve daily P pictures, charm shots and other products, and are good at picture beautification, portrait beauty, image recognition and various filter technologies.
3) Beijing: WeChat Pattern Recognition Center
There are two core teams in promoting the research and development of new technologies and functions of WeChat. One is the Beijing R&D team from the former Tencent Research Institute. After the dissolution of Tencent Research Institute, some people were assigned to the WeChat department. The perception of the outside world is that this is a team doing' pattern recognition' research. It was first established in September of 20 10, and now it is called WeChat Beijing R&D Center.
The research direction of the team includes speech recognition, image analysis, semantic understanding, WeChat big data mining and so on. Its research and development results are widely used in Tencent products such as WeChat. External services include WeChat voice recognition, image analysis, audio fingerprinting, WeChat semantic customer service robot, etc. The daily service volume reaches tens of millions. At present, we are committed to building a WeChat dialogue robot platform and connecting the services of the WeChat public platform more intelligently.
In fact, most WeChat users have used what this team has done, such as voice-to-text in WeChat, voiceprint recognition, and Tik Tok in TV programs.
A former Tencent employee said that although these new functions seem simple, the technology behind them will take a long time to settle down. The underlying technology will be silent for several years, and the implementation of the application layer will be much shorter. '
4) Intelligent Computing and Search Laboratory
Tencent Intelligent Computing and Search Lab was established on 20 15. According to reports, it focuses on four research fields: search technology, natural language processing, data mining and artificial intelligence.
The laboratory mainly organizes employees with academic ambitions in technology and engineering enterprises to participate in research work, and cooperates with universities such as Harbin Institute of Technology and China University of Science and Technology. The research results have been successfully applied to WeChat, QQ, QQ music, QQ video and other products. The search department mainly focuses on natural language recognition. Through Tencent Cloud, the two major products of the laboratory, Yunsou and Wenzhi, have been opened to the industry. However, there are few projects and progress in the laboratory recently, so official website can't open it.
In addition, according to informed sources, Tencent Artificial Intelligence Research Institute will be established in the near future.
Third, landing business: home products and industry users
BAT's artificial intelligence technology research and development is more driven by business needs, and Tencent is no exception. This method avoids aimless research and unnecessary failure to a certain extent, but it is difficult to achieve great innovation and breakthrough in its technology and products.
1) WeChat: Before blocking' Xiao Bing', we launched' Xiao Wei'.
Wechat business group mainly focuses on the expansion of the field of human-computer interaction, and also implants technologies such as image and speech recognition.
A year ago, on WeChat, the chat bot of Microsoft Xiao Bing was banned on the grounds of "revealing privacy, inducing users to attract groups and registering junk accounts". Microsoft has written many open letters to express its strong dissatisfaction. The two sides were at daggers drawn. At one time, there was a rumor that' Microsoft wanted to ban QQ on the platform', but in the end it went away.
A year later, Microsoft Xiao Bing returned to the WeChat platform. In addition, WeChat also launched the intelligent robot' Xiao Wei', where users can communicate in natural language to overcome the shortcomings of the intelligent mechanical response of voice assistants in the past.
According to the official introduction, compared with ordinary voice assistants, the communication between Xiao Wei and users is no longer a question-and-answer form. With nearly 600 million daily active users' data on WeChat platform,' Xiao Wei' can combine users' personal characteristics, social friends and daily habits to give results closer to users.
In addition, Xiao Wei is also "hand in hand" with WeChat payment function, trying to link all services such as shopping, payment and taxi on WeChat platform to form an O2O closed-loop business system.
Of course, the realization of voice-to-text translation and translation functions in WeChat is also supported by voice intelligence technology.
2) Yunsao and Wenzhi
Yunsou and Wenzhi are the two main B-side products of Tencent Intelligent Computing and Search Lab mentioned above.
Cloud search can highly abstract and integrate the major vertical search services within the company, and provide users with visual data customization services. At the same time, it also has the functions of association word recommendation, advanced error correction, manual intervention, domain search, personalized word segmentation and other attachment construction.
The Chinese semantic platform of Wenzhi has powerful Chinese semantic analysis function. Based on its API, users can conduct semantic analysis applications such as search, recommendation, public opinion and mining, and can also customize semantic analysis solutions with product characteristics.
3) Other businesses
In addition, UTU face recognition has been applied to finance, security, identity recognition and other fields. Among them, the application of Tencent Credit Information, Weizhong Bank and Tenpay allows it to evaluate users' credit in combination with consumption and social situation, and strengthen the security line of identity verification. This technology has even been officially recognized, and Youtu has joined hands with the National Citizen ID Number Enquiry Service Center under the Ministry of Public Security to make the next layout.
Fourth, hardware platform: cooperation is the mainstay.
In terms of hardware product layout, Tencent mainly attracts cooperation through its own platform users and traffic advantages, mainly including QQ IOT, WeChat intelligent hardware and TOS+ strategy.
1)QQ IOT
If products such as WeChat and QQ are designed to connect people, then QQ IOT based on the device side is Tencent's efforts and layout in connecting' things'.
QQ IOT is a hardware and software development kit. Through SDK or chip module, the rapid networking of devices can be realized. In 20 15, QQ IOT launched the "Easy Link Plan", which will invest 2 billion resources to support partners in terms of capital, traffic, service, training and marketing. Using Tencent's 20 national entrepreneurial bases and mature open entrepreneurial service system, through free access, free promotion, free sharing, cloud fee reduction and exemption, QQ IOT hard-created investment funds, etc.
Previously, QQ IOT had cooperated with Intel to jointly release a smart home-based pension platform, which changed the traditional pension medical model. At the beginning of this year, we launched children's Bluetooth positioning running shoes with Li Ning and Daphne.
2) WeChat hardware platform
In fact, as early as 20 14, WeChat was trying to access wearable devices. 20/kloc-in August, 2005, wechat officially launched' wechat hardware industry solutions' in eight industries, including air conditioning, toys, routers, home, TV, recharging, health and wearing.
Hardware manufacturers can connect devices through WeChat scan code, and users can manage devices through WeChat service number. At present, the WeChat hardware platform has been connected to more than 2,400 hardware manufacturers, and the amount of device activation has reached 25 million.
3)TOS+ strategy
TOS+ strategy was put forward by Tencent on 20 15, which is divided into' TOS' and'+'. TOS is actually a mobile phone ROM based on Android and Tencent.
TOS is not news. As early as 20 15, 1, Tencent has invited users to this ROM. It is said that the March version is still based on Android 4.3.
As for the history of Tencent mobile phone ROM, it can be traced back to 20 10. After several twists and turns, it has experienced the rise of MIUI and the fall of players such as dim sum OS. In the words of a netizen, the reason why Tencent does TOS again is only four words' I won't give up'.
From 20 10 to 20 15, Tencent successively launched QQ service mobile phone and TITA mobile phone system, but it has never been done.
"+"in tos++ is an extended operating system, which connects the mobile phone with the underlying intelligent hardware. Its areas of concern include smart watches, micro-game consoles and virtual reality products.
The ambition of TOS++ is more like being a' connector', and some long-tail products such as sphygmomanometer, body fat meter and smart home have been connected by two platforms, namely' WeChat intelligent hardware' and' QQ IOT'. As long as the standards are open, Tencent does not need to define solution rules at all.
On the one hand, TOS+ and two * * * are isomorphic, forming the layout of Tencent in the field of intelligent hardware. On the other hand, it will change the situation that different intelligent hardware devices are independent of each other due to the lack of underlying connections. QQ IOT and WeChat hardware will also be integrated and take the road of parallel development.
Verb (abbreviation of verb) conclusion
Judging from the published data, WeChat currently has more than 600 million active users, and still generates a large amount of data every day, which is transformed into users of search and e-commerce, while matching user behavior. The huge user base and data make Tencent an AI' gold mine' with great demand and development space.
By continuously developing its own business, carrying out the bottom R&D reserve to promote the integration of learning and research, technology and data can be quickly verified and generate economic value. At the same time, by accelerating the acquisition and cooperation of excellent companies, it can make up for its shortcomings and quickly enter the track.
Compared with Baidu and Ali, Tencent's layout and technical advantages in artificial intelligence still lack a high moat. Based on the lifeline of social and communication, data mining and technology landing on WeChat may be a favorable opportunity for Tencent.
So in general, what Tencent has done in artificial intelligence is to assist current products and services, such as adding artificial intelligence elements to WeChat, mobile payment and other functions; There is no such typical product based on artificial intelligence itself. For Tencent, artificial intelligence technology is the evolutionary thrust of product experience.