A few days ago, the "China Artificial Intelligence Application and Ecology Research Report" (hereinafter referred to as the "Report") was released. The "Report" selects visual intelligence, voice intelligence, data intelligence, and enterprise intelligence for the current application scenarios of artificial intelligence and the construction of intelligent platforms. research in this field.
The Report is divided into six categories:
1. The current situation of artificial intelligence applications, platforms and ecology in China;
2. Visual intelligence applications, platforms and ecology;
3. Voice intelligent applications, platforms and ecology;
4. Data intelligence application, platform and ecology;
5. Enterprise service intelligent application, platform and ecology;
6. Prospects of artificial intelligence applications and trends in China;
Some views of the report:
1. Currently in the integration stage of artificial intelligence technology and enterprise business, some enterprises have begun to deploy artificial intelligence applications and have achieved certain results.
2. When selecting artificial intelligence manufacturers, enterprises pay the most attention to the technical strength, brand, business understanding and integration capabilities of the manufacturers.
3. Simultaneous implementation of application landing, platform construction, and ecological cultivation, and competition between manufacturers will be carried out simultaneously in multiple dimensions. The elements of the AI ecology evolve dynamically, and the ecology based on the open platform is gradually taking shape.
4. After computing power, algorithms and data, scenarios become the key to AI applications.
5. Visual intelligence has passed the technological inflection point and entered the stage of application popularization, and will become the field with the most commercial prospects in the next five years. Face recognition has opened up people's cognition, but intelligent security and smart city are the "sea of stars" of visual intelligence.
6. The voice intelligent interactive system has been popularized in various fields, and the B-side and the C-side have different development characteristics: the C-side presents a competitive situation at the operating system level, the B-side market window is still large, and the vertical application scenarios of different industries need to be explored.
7. Data analysis and data mining are the foundation of data intelligence. Local intelligence will transform into global intelligence. Internet giants represented by BATJ have huge advantages. Massive amounts of data will be aggregated in the "cloud", and a high degree of integration between data will be achieved, and "cloud data intelligence" will become more common.
8. Due to the rise of the new retail concept, everyone has gradually paid attention to the value of offline data, and has obtained a large amount of offline data, especially online data, through applications such as AutoNavi Maps, Meituan, Alipay, WeChat, etc., combined with new methods such as face recognition. offline consumption and payment data, and finally realize the integration of online data and offline data, and build a more complete user portrait.
9. At present, enterprise intelligence is still in its infancy, and "artificial intelligence technology + business scenarios + management processes" is an ideal model for enterprise intelligence. The integration of internal data and human-machine collaboration is the next development focus of enterprise intelligence.
10. The low degree of integration of technology and business and lack of data are important factors hindering the application of artificial intelligence. Data monopoly has become a major obstacle to the development of the AI industry. Difficulty in coordinating commercial interests and unclear data boundaries are the key factors hindering data sharing. A possible state in the future is that the government and enterprises are linked to establish a data ecology that is open and shared, and has clear boundaries of interests.
The report argues that:
Visual Visual intelligence is the most promising artificial intelligence application field in 5 years. In 2018, intelligent security will be the development center of visual intelligence manufacturers. According to the "Thirteenth Five-Year Plan (2016-2020) Development Plan for China's Security Industry" issued by the China Security Association, during the "Thirteenth Five-Year Plan" period, the security industry will transform and upgrade to large-scale, automated and intelligent, and by 2020, The total revenue of security enterprises has reached about 800 billion yuan, with an annual growth rate of more than 10%. Just in the field of intelligent security, it is enough to support multiple unicorn companies of tens of billions of dollars and meet the performance expectations of capital for visual intelligent manufacturers. In the field of intelligent security, manufacturers can be divided into two categories according to the business evolution path: "AI + security" manufacturers represented by Megvii Technology, SenseTime and Bostin, and "security protection" manufacturers represented by Hikvision +AI" type manufacturers. Smart cities will be the largest application area of visual intelligence, and the future market size will surely exceed trillions. Visual intelligence + data analysis will become the main body of the smart city system and play the most critical role. Visual intelligence applications have shifted from image recognition and object recognition to computer vision understanding and video understanding, which are more challenging and valuable fields, and their application prospects will be broader in the future. With the development of chip technology, terminal intelligence will be rapidly improved. The data processing method of cloud + terminal has become a trend. In the future, smart terminals will become an important computing carrier for visual intelligence applications, and the cloud will play a key role in data access.
In the field of voice intelligence , on the whole, the C-side voice intelligence market has begun to show competition at the operating system level, and competition is carried out in the form of an open platform and an ecosystem. It is a mobile phone and smart speaker, and gradually expands to smart robots, smart homes and other fields. The focus of competition among manufacturers is on mobile phone manufacturers and intelligent hardware manufacturers; continuously access more services in their own voice interaction systems, and constantly expand what users can do through intelligent voice assistants; build an open platform to gather a large number of software A hardware partner that leverages the power of an ecosystem to strengthen its competitive advantage and market leadership.
At present, the recognition rate of mainstream voice intelligent manufacturers in China is generally higher than 97%, which is a technical threshold for commercial application. In terms of speech recognition accuracy, iFLYTEK, Baidu, and Spichi are leading, and iFLYTEK's high recognition rate in the field of dialects is a major feature. The improvement of far-field noise reduction and far-field recognition capabilities has spawned smart speaker products and further promoted the development of smart hardware. Mainstream manufacturers such as iFLYTEK and Baidu have also reached the international leading level in their machine translation capabilities. Voice intelligence, visual intelligence and unmanned driving have become the three business cards of my country's artificial intelligence industry. It is a major difficulty to form a more natural human-computer interaction based on context understanding and intention understanding based on semantic understanding.
The B-side voice intelligence market has a large window, and vertical application scenarios based on voice interaction need to be explored. The vertical application scenarios based on semantic understanding and voice interaction are further developed, with the ability of context understanding, multiple rounds of dialogue, and the ability to interrupt at any time, which is an important competitiveness of voice intelligent manufacturers. Intelligent customer service, voice assistants embedded in enterprise management software, and medical robots are important segments. Especially for intelligent customer service, the development of scenario-based voice interaction systems for the business attributes of different industries is a blue ocean market that needs to be further explored. In the fields of smart vehicles, smart homes, smart robots, and smart wearables, voice smart manufacturers have followed suit. In enterprise management software systems, voice intelligence has great application potential: on the one hand, in office software, especially mobile office software, some manufacturers have tried to embed voice assistant modules in their products, allowing users to interact with voice Deal with work matters; on the other hand, the business process based on semantic understanding is opened up and the voice interaction is integrated into its business process system, which is also the main development direction of enterprise management software manufacturers.
In the field of data intelligence , data analysis and intelligent decision-making, data visualization, intelligent marketing, user profiling and personalized recommendation, and BI are important subdivisions; in terms of data types, the analysis of mobile data and offline retail data has become a hot spot. With the development of the big data industry, the basic data bureau accumulation and data processing system has been basically formed. In the future, the focus of attention will turn to the integration of multi-channel and various data forms, and will shift from local intelligence to global intelligence. Local intelligence is data intelligence in a segmented field to solve certain types of problems for users. Correspondingly, Global Intelligence will achieve three breakthroughs:
1. Fusion of multi-channel data. Including the integration of PC data, mobile data, IoT data, online data and offline data;
2. Fusion of multiple types of data. Not only can handle structured data, but also unstructured data such as text, voice, image, video, etc., and can realize data access, and provide results and decision-making suggestions based on comprehensive analysis of various types of data;
3. Multi-dimensional data fusion. Including the integration of multi-dimensional data such as e-commerce data, social data, search data, and offline consumption data;
In addition, local intelligence is more focused on analysis, while global intelligence is more focused on decision making and automating some things for users.
In-depth user portraits have become the foundation of the industry, and Internet giants have huge advantages in the field of data intelligence. Many data intelligence applications are based on user portraits, from which personalized recommendations, intelligent marketing, business intelligence, and security situational awareness are derived. Further, in-depth user portraits can further describe user characteristics, and based on this, some higher-value intelligent services can be developed, such as financial risk control, which is widely used in finance, especially in the field of Internet finance.
Internet giants, due to their massive data accumulation and strong comprehensive strength, have great advantages in the field of data intelligence applications, and this advantage will be further strengthened. Among them, Alibaba has the most obvious advantages. Alibaba's e-commerce data, payment data, and logistics data are the types of data with the highest value. Based on the comprehensive characterization of users, rich intelligent applications can be developed. Due to the rise of the new retail concept, everyone has gradually paid attention to the value of offline data, and has obtained a large amount of offline data, especially offline consumption, through applications such as AutoNavi Maps, Meituan, Alipay, WeChat, etc., combined with new methods such as face recognition. And payment data, and finally realize the integration of online data and offline data, and build a more complete user portrait. The maturity of cloud computing has effectively promoted the development of data accumulation and data analysis industry, and also effectively promoted the integration of data. In the future, massive amounts of data will be aggregated in the "cloud", and a high degree of integration between data will be achieved, and "cloud data intelligence" will become more common.
In the field of enterprise intelligence , "artificial intelligence technology + business scenarios + management processes" is an ideal model for enterprise intelligence. The "Report" believes that enterprise intelligence is still in the exploratory period, and the model is not fixed. Intelligent assistants, face recognition punch cards, etc. are only primary applications, and do not cut into the core of enterprise intelligence. A relatively mature enterprise intelligence application must achieve a high degree of integration of artificial intelligence technology with enterprise management processes and business scenarios. In the future, artificial intelligence technology, especially semantic understanding and data mining technology, will be embedded in various fields of enterprise information systems, including ERP systems, CRM systems, HR systems, SCM systems, and financial systems. Through the understanding and value mining of various data information in these systems, combined with the management process and the knowledge map of the integration of business scenarios, a comprehensive and detailed analysis of the operation of the enterprise as a whole and each sub-field is carried out. Further, the intelligent system can automatically process certain items according to a specific process.
The integration of internal data and human-machine collaboration is the next development focus of enterprise intelligence. Through the integration of data and industry and business knowledge to build a knowledge graph for a specific industry, the "expert system" may be reborn and play a greater value. On the one hand, the establishment of the enterprise's internal expert system to assist the enterprise's management decision-making and business development, but also to provide assistance for the training of new employees. On the other hand, as an extension of the enterprise business system, it provides external services, such as the intelligent customer service system, which answers customer questions and performs some business operations based on the knowledge map of the industry and enterprise business. In the next few years, embedded intelligence will be further popularized, and the use of intelligent hardware devices in the field of enterprise services will become a bright spot.
The relationship between humans and artificial intelligence can be divided into three categories: machine-led, human-led, and human-machine collaboration. The future is no longer dominated by a single person or machine, but by efficient collaboration between humans and machines. People ask questions, provide data, intelligent systems make decisions and suggestions based on the information, and realize automated operations within a certain business scope. People are the beneficiaries of intelligent services. Through the application of enterprise intelligence, the organizational state of scientific decision-making, automated operation, and efficient human-machine collaboration can be realized.
In addition, regarding the development strategy of promoting my country's smart industry, the "Report" pointed out:
On the one hand, accelerate the research and release of relevant laws and regulations. Clearly define various data values and data transaction specifications, and clarify the relationship between data sharing, personal privacy, and corporate interests. The government has stepped up efforts to open data, widened the types of open data, and improved the quality of open data. Promote enterprises, especially Internet giants, to form data sharing alliances to promote data sharing between enterprises and between enterprises and governments. Establish and improve data trading centers, promote data trading based on real data value, fully consider personal privacy and corporate interests, and achieve sustainable development of data trading and data sharing. Set up a public service platform for basic resources, and establish a training resource library and standard test data set that are open to the society in various data forms such as images, voice, video, and text.
On the other hand, under the guidance of the government, establish a data alliance between enterprises, clarify the relationship of interests, negotiate the scope of data that can be traded and shared, and standardize the way of trading and sharing on the premise of fully respecting their respective data rights and interests. process. Promote the flow of data among Internet companies, traditional companies, research institutions, and governments, and realize the value of data in circulation. Establish data standards, especially data interface standards, and build data interface APIs to facilitate data calls. Establish an interest coordination mechanism between enterprises, especially Internet enterprises, between Internet enterprises and traditional enterprises, and fundamentally guarantee the effective promotion of data sharing.
Going a step further, implement the "One Body Two Wings" strategy at the national level:
"One":
Government joint enterprises build a public service platform for basic resources, such as a new computing cluster sharing platform, an open platform for algorithms and technologies, and a training database for artificial intelligence models.
"Two Wings":
1. Provide certain tax incentives and subsidies to enterprises implementing artificial intelligence applications to promote the popularization of artificial intelligence applications in traditional enterprises.
2. Establish a super-large-scale national artificial intelligence industry fund, at least on the order of 500 billion or more (not less than the size of SoftBank's "vision" fund). Support start-up artificial intelligence enterprises internally, and conduct global industrial mergers and acquisitions externally. In addition, it will provide policy and capital support for the domestic listing of some "China Concept Stock" companies in the future.
At the enterprise level, while competing, strengthen technical exchanges and technical cooperation, especially technical cooperation in the bottom and frontier fields. Drawing on the artificial intelligence alliance established by five technology companies such as Google, Facebook, Amazon, IBM and Microsoft in the United States, domestic Internet giants represented by BAT and iFLYTEK, as well as emerging artificial intelligence unicorns such as Megvii and SenseTime, should Establish similar alliances as soon as possible to promote technical cooperation and industrial development. Accelerate the construction of a public service platform for artificial intelligence basic resources, and promote the opening of basic data and technology with greater efforts. Leading artificial intelligence companies should take greater responsibility to deepen the frontier fields such as artificial intelligence chips, quantum computing, machine learning, deep learning, and enhanced learning, so as to broaden the development boundaries of China's artificial intelligence industry. Promote the talent flow and joint training mechanism between enterprises, universities and research institutes.
Inquire about the content and data of this report
For details , please contact WeChat: oyg0001
Overview of the report
Chinese software