[Netease Intelligence News Dec. 21] On Dec. 20, IDC and Baidu AI Industry Research Center (BACC) jointly released Baidu Brain Leadership White Paper, which predicted the development trend of Chinas artificial intelligence market in 2019, analyzed how AI went from technology to landing through actual cases, and put forward the 100-day AI deployment plan.
IDC predicts that the Chinese AI market will reach 9.84 billion US dollars in 2022. The white paper predicts ten trends of AI in 2019 from four dimensions: technology orientation, landing implementation, application value and market ecology.
Prospect 1: Machine learning/in-depth learning begins to enter traditional enterprises. Machine learning/in-depth learning will enter the enterprise and provide decision-making-centered services for the enterprise. At the same time, in-depth learning will continue to be widely used in image, audio, text and other unstructured data processing. Especially for large and medium-sized enterprises in traditional industries, the development of artificial intelligence applications using machine learning platform will gradually become the mainstream. IDC expects that 15% of enterprises in the industry will adopt machine learning by 2020.
Prospect 2: Multi-modal computing, which integrates vision, speech and semantics, begins to land. Machine intelligence, which can only see and hear clearly, can no longer meet the needs of human beings. Multimodal computing, which integrates vision, speech, semantics and emotion, has become an urgent need to achieve real intelligence. It is expected that multi-modal computation will begin to land in practical application in the next three years.
Prospect 3: Multi-model databases begin to enter the market. With the investment of the Internet of Things and the process of enterprise digital transformation, various unstructured data in enterprises are growing rapidly, which makes multi-model database which can support multi-format data management become an urgent need. IDC expects multi-model databases to account for 30% of NoSQL database expenditure by 2023.
Prospect 4: Low code development platform reduces the threshold of AI technology. Low code volume/codeless development platform promotes AI deployment automation, reduces the threshold of technology use, and enables SMEs to use AI equally to achieve GSP AI. Users can upload original data such as pictures, audio and text, and the system can automatically train the appropriate model. Typical cases - Google AutoML, Baidu EasyDL.
Prospect 5: AI expands from cloud deployment to edge computing. Infrastructure is shifting to edge locations near data sources and end-to-end devices, and AI will be the first application to benefit from edge computing. Edge devices will include AI algorithms and drive the delivery of computing power. IDC expects that by 2022, 25% of devices on the Internet of Things will be running AI algorithm models.
Prospect 6: Business process intelligence, automation level to a new height. Machine learning-driven artificial intelligence will promote a new wave of business process reengineering, and many applications will be highly simplified. Typical cases such as financial process automation, Verification Automation and many other process automation levels will reach a new level. IDC expects that by 2023, AI will replace 50% of IT business workload and save more than 20% of operating costs.
Prospect 7: Human-computer interaction interface tends to be intelligent. On the one hand, voice empowerment programs are becoming more and more popular - voice conversation capabilities will be embedded in hardware and application software. On the other hand, AI, which integrates voice, image, video and semantic comprehension capabilities, will become the mainstream mode of human interaction with applications. IDC expects to replace 50% of current screen-based B2B and B2C applications with AI-enabled human-computer interaction interfaces by 2023.
Prospect 8: Six industries adopt AI in an all-round way. The government industry, financial industry and Internet industry will expand the application of AI in an all-round way after the application practice in recent years. New retail, new manufacturing and medical fields will also become new growth points of AI market. IDC expects the composite growth rate of AI application in these six industries to exceed 30% in the next three years.
Prospect 9: Software and application lead infrastructure. Software definition computing has become one of the important strategies of chip manufacturers. The phase of software and application-driven AI dedicated chips will also come. In the future, the popularity of machine learning applications, whether machine learning always requires a large number of data sets, and the evolution of deep learning neural networks will affect the development of accelerated computing hardware. Evolution of machine learning technology and trend of AI application are becoming more and more important to infrastructure providers.
Prospect 10: Integration of ecological resources is the key to success. Artificial intelligence technology is penetrating into end-to-side intelligence. Successful application can not be separated from the high degree of adaptation of hardware and software, which makes the integration of technology-based manufacturers with sensor, camera, module and other sub-industries more and more important. The key to success is to integrate all kinds of ecological elements in the solution and build a network platform of partners.
So how can technology be combined with practical applications? IDC tracked nearly 70 application scenarios and found that, with the opening of the market more and more technological capabilities, application scenarios tend to be widespread. To apply AI technology to enterprises and play its effectiveness, three points need to be focused on: first, easy-to-use, simple technology stack, second, model tuning based on vertical scenarios, and third, from the end of the data center. Software and hardware adaptation.
The white paper puts forward a preliminary framework for AI application effectiveness evaluation, which divides the departments affected by AI into four categories: product service, production mode, operation mode and decision-making mode, and evaluates them from various dimensions. For example, in manufacturing industry, AI will give priority to bringing high efficiency to product service, production mode and operation mode; in financial industry, AI application efficiency is best reflected in product service and operation intelligence, followed by production intelligence and decision intelligence. AI can bring different efficiencies to different departments of enterprises, but almost all the efficiencies brought by AI use cases are reflected in the saving of production resources such as time and manpower, cost reduction, productivity improvement, income growth and so on.
The white paper also points out that AI ecology is experiencing changes from highly centralized to self-governing to performing their respective duties, and that AI ecology with high integration and clear division of labor is coming. Under this trend, in order to give full play to the application efficiency, industry participants should not stop simply using these technologies, but actively build and implement platforms and services that help to promote cooperation among all parties, so as to enhance the efficiency of the whole ecosystem.
The white paper emphasizes that with the increasing penetration of AI in various industries, qualified enterprises need to formulate AI deployment plans as soon as possible. IDC recommends that enterprises plan a 100-day AI deployment plan and gradually complete planning, scenario selection, team formation, testing, implementation and replication, so as to continuously adjust the cycle to meet broader business needs. (Dingxi)
Baidu Brain Leadership White Paper: https://ai.bdstatic.com/file/37FB772329554 EAFAE798FE0E111D1001