Top Ten Most Growing Technologies in the New Generation of AI from 2018 to 2019

Source: Network Big Data

ABSTRACT: At the main forum of the 2018 World Robot Congress, the Chinese Society of Electronics released "Ten Most Growing Technological Prospects in the Field of New Generation Artificial Intelligence (2018-2019)".

At present, the whole world is experiencing a new round of transformation with high coupling and deep superposition of science and technology and industry. The formation of big data, the innovation of theory and algorithm, the improvement of computing ability and the evolution of network facilities drive AI to enter a new round of peak period of innovation and development. New technologies continue to make breakthroughs, presenting in-depth learning, cross-border integration, man-machine collaboration, open-mindedness and self-reliance. New features of application-oriented, such as master control, etc. Strengthen the forecasting of the new generation of artificial intelligence technology, accurately grasp the global technological innovation trends and development trends, will provide intellectual support with decision-making reference value and practical guidance significance for the healthy development of the industry, orderly entry of funds, policy planning, emerging market development, etc.

Based on a series of policy plans, such as the National Development Plan of New Generation Artificial Intelligence and the Three-Year Plan of Action for Promoting the Development of New Generation Artificial Intelligence Industry (2018-2020), the Chinese Electronics Society has investigated and visited a number of leading enterprises with leading level and characteristics in the new generation of artificial intelligence technology and industry, and organized visits from well-known universities. Experts and scholars in artificial intelligence related fields of research institutes have systematically combed the latest reports of authoritative think tanks and well-known strategic consulting companies, selected ten most distinctive growth technologies, and reached the following conclusions:

I. antagonistic neural networks

It refers to a neural network architecture consisting of a neural network module which continuously generates data and a neural network module which continuously discriminates whether the generated data is real or not, and it is a technology that creates approximately real original image, sound and text data. This technology is expected to greatly improve the accuracy and accuracy of machine translation, face recognition and information retrieval. With the improvement of data sequence capability of three-dimensional model, it will have considerable application value in the fields of automatic driving, security monitoring and so on.

Fig. 1 Data generation and discriminant data continue in adversarial neural networks

Source: MIT Science and Technology Review

Capsule network

It refers to the technology of building multi-layer neuron module in deep neural network to discover and store the detailed spatial position and attitude information of objects. This technology enables the machine to quickly recognize the same object in different situations with less sample data. It has broad application prospects in face recognition, image recognition, character recognition and other fields.

Figure 2 Capsule network algorithm can identify the same object from different angles

Source: International Conference on Learning Representations,

Matrix capsules with EM routing

III. Cloud-end Artificial Intelligence

Cloud computing is a technology that integrates the operation mode of cloud computing and artificial intelligence, and centralizes the use and sharing of machine learning tools in the cloud. This technology transfers huge AI operation cost to cloud platform, which can effectively reduce the threshold of using AI technology for terminal equipment, and is conducive to expanding user groups. It will be widely used in medical, manufacturing, energy, education and other industries and fields in the future.

Figure 3: Major cloud computing companies launching AI services

Source: China Electronics Society

IV. Deep Enhanced Learning

It is a technology that combines deep neural network with reinforcement learning with decision-making ability and integrates perception, decision-making or perception decision-making through end-to-end learning. This technology has the characteristics of no prior knowledge, less complexity of network structure and less requirement of hardware resources. It can significantly improve the efficiency and robustness of machine intelligence to adapt to complex environment. It will have broad prospects in the fields of intelligent manufacturing, intelligent medical treatment, intelligent education and intelligent driving.

Fig. 4. Deep reinforcement learning has good structural characteristics.

5. Intelligent Brain-Computer Interaction

It refers to the technology of information exchange and function integration between nervous system and external equipment by establishing direct connection pathway between human brain nerve and external equipment with high biocompatibility. This technology uses the brain-computer interface controlled by artificial intelligence to analyze the working state of human brain accurately, so as to promote the effect of brain-computer intelligent fusion, make the way of human communication more diverse and efficient, and will be widely used in clinical rehabilitation, autopilot, aerospace and other fields in the future.

Fig. 5 Intelligent Brain-Computer Interaction Makes Human Communication Efficient

6. Dialogue Artificial Intelligence Platform

It is an open platform technology that integrates speech recognition, semantic understanding, natural language processing, speech synthesis and other solutions to provide developers with recognition, understanding and feedback capabilities. This technology can realize the natural interaction between machine and human in the scene of dialog service. It is expected to be widely used in many fields such as smart wearable devices, smart home, smart car and so on in the future.

Figure 6 Dialogue Artificial Intelligence Platform Architecture

Source: IDC, China Electronics Society

7. Emotional Intelligence

Artificial intelligence is used to simulate human-like emotional responses such as expressions, moods and emotions, so as to create a virtual image with emotional attributes. This technology can give the machine and equipment better ability to recognize, understand and guide human emotions, and bring more efficient and humane interaction for users.

Please read the Chinese version for details.