cool hit counter Science and Technology Innovation 2030 - "New Generation Artificial Intelligence" Major Project 2018 Call for Submission of Project Guidelines_Intefrankly

Science and Technology Innovation 2030 - "New Generation Artificial Intelligence" Major Project 2018 Call for Submission of Project Guidelines

Source: Financial resources request

Abstract: The layout and task deployment of the major project of "New Generation Artificial Intelligence" have been consulted by the Special Invitation Committee for Strategic Consultation and Comprehensive Review and submitted to the State Council for approval.

According to the "Several Opinions of the State Council on Improving and Strengthening the Management of Central Financial Research Projects and Funds" (Guo Fa [2014] No. 11), the "Notice of the State Council on the Program for Deepening the Management Reform of Central Financial Science and Technology Programs (Special Projects, Funds, etc.)" (Guo Fa [2014] No. 64), the "Notice of the Ministry of Science and Technology on the Issuance of the Implementation Plan for Major Science and Technology Projects of New Generation Artificial Intelligence" (Guo Ke Fa Gao [2017] No. 344) and other documents, we are now seeking comments and suggestions from the society on the 2018 project declaration guidelines for the major projects of Science and Technology Innovation 2030 - "New Generation Artificial Intelligence" (draft for comments, see annex). The consultation period is from September 5, 2018 to September 11, 2018.

The layout and deployment of the major projects of the "new generation of artificial intelligence" have been reviewed by the Special Invitation Committee for Strategic Consultation and Comprehensive Review and submitted to the State Council for approval and implementation. The 2018 guidelines of this project focus on the deployment of three directions: basic theories of new-generation artificial intelligence, core key technologies for major needs, and intelligent chips and systems. At the same time, considering the broad scope of the AI field and the openness of the research, and in order to encourage more research and application teams to propose targeted tasks and objectives around the research content, this batch of guidelines does not set specific assessment indicators for the time being. This consultation focuses on the rationality, scientificity and advancement of the above proposed deployment directions to listen to the opinions and suggestions of all parties. The Ministry of Science and Technology will carefully study the comments and suggestions received and revise and improve the project declaration guidelines for major projects. Comments and suggestions collected will not be fed back or responded to.

Department of High and New Technology, Ministry of Science and Technology

Annex: Draft Guidelines for the Application of Major Projects of Science and Technology Innovation 2030 - "New Generation Artificial Intelligence" for 2018

In order to implement the "New Generation Artificial Intelligence Development Plan", the major project of Science and Technology Innovation 2030 - "New Generation Artificial Intelligence" was launched. In accordance with the overall objectives and 2020 milestones set out in the Implementation Plan for the New Generation Artificial Intelligence Major Science and Technology Projects, the proposed project guidelines for 2018 are presented. The 2018 annual project guide deploys and implements three directions: basic theory of new-generation AI, core key technologies for major needs, and intelligent chips and systems, with a three-year implementation cycle (2018-2020).

Applicants should propose specific assessment indicators and clear task objectives according to the characteristics of the projects to be applied for, in accordance with the principles of demand-oriented, problem-oriented and target-oriented, as described in the guidelines.

1. Grounded theory for a new generation of artificial intelligence

Focusing on the major scientific frontier issues of AI, with a focus on breaking through the bottleneck of basic mechanisms, models and algorithms of AI, and laying emphasis on the basic theoretical research of a new generation of AI that may trigger a paradigm change in AI, providing a strong scientific reserve for the continuous development and deep application of AI.

There are seven research directions in the basic theory of next-generation artificial intelligence, and each direction will support 1-2 projects. Inclusive teams that already have a good foundation are encouraged to apply.

1.1 New Generation neural network models

Drawing on neurocognitive mechanisms and machine learning mathematical methods, among others, to carry out neural network Research on new theories and methods such as model nonlinear mapping, automatic evolution of network structure, functional specialization of neurons and modules, small sample learning/weakly labeled/unlabeled sample learning, interpretability, etc., intrinsically enhance the scope and capability of deep neural networks to support the solution of realistic AI problems.

1.2 Adaptive Sensing for Open Environments

In order to address the problem that the performance of intelligent systems may drop sharply due to changes in application scenarios, we develop hierarchical network structures with strong adaptability, machine learning strategies with continuous learning and general performance measurement methods, breakthroughs in unsupervised learning, empirical memory utilization, implicit knowledge discovery and guidance, and attention selection, and promote the formation of universal perceptual intelligence in open environments and changing scenarios.

1.3 Cross-media causal inference

We will study new methods of machine learning based on human common sense knowledge formation in cross-media, and perform bottom-up deep abstraction and induction on cross-media data supported by common sense knowledge, effectively manage uncertainty in top-down deduction and inference, and establish new models and methods to coordinate and complement logical inference, inductive inference and intuitive epiphany, so as to realize the leap from intelligent correlation analysis to causal inference supported by common sense knowledge in cross-media.

1.4 Game Decision Making under Imperfect Information

For human economic activity、 Characteristics of games under conditions of imperfect information such as human-machine confrontation, Combining machine learning、 cybernetics、 Progress in areas such as game theory, Study of dynamical mechanisms and optimal decision making in game confrontation under uncertain and complex environments models, Integrating adversarial learning and reinforcement learning with dynamic game theory, Towards a foundation for task-oriented general intelligence in non-complete information environments models and dynamic game decision theory。

1.5 Mechanisms and computational methods for group intelligence emergence

Research on the organization model and incentive mechanism of large-scale group collaboration in open, dynamic and complex environments, establish expressible, computable and adjustable compound incentive algorithms, explore the emergence mechanism and evolution law of individual contributions into group intelligence, break through the evolution method of group intelligence for global goals and time-sensitive group intelligence synergy, and realize predictable, steerable and sustainable emergence of group intelligence.

1.6 Human-in-the-loop hybrid augmented intelligence

Research on task modeling, environment modeling and human behavior modeling under uncertainty, vulnerability and openness, develop human-in-the-loop machine learning methods and hybrid augmented intelligence evaluation methods, closely couple the advanced cognitive mechanisms of human analysis and response to complex problems with machine intelligence systems, effectively avoid the risk of decision making and system loss of control due to the limitations of artificial intelligence technology, and achieve two-way human-machine collaboration and convergence of solutions to complex problems.

1.7 Human-Machine Object Cooperative Control Methods in Complex Manufacturing Environments

The research is aimed at the complex multi-dimensional human-machine and object collaboration problems in discrete manufacturing and process industries, researching on cross-layer and cross-domain distributed networked collaborative control methods, breaking through the theory of human-machine and object ternary collaborative decision making and optimization, realizing the virtual-real integration and dynamic scheduling of human-machine and object, exploring the reconfiguration of unmanned processing lines and human-machine co-integration and intelligent interaction, and providing theoretical and methodological support for the exploration of smart factory development models and the establishment of standard systems.

2. Key common technologies for major needs

Focusing on the urgent need to enhance the international competitiveness of China's artificial intelligence, we will break through the key common technologies of a new generation of artificial intelligence for major needs, with algorithms as the core and data and hardware as the basis, and comprehensively improve the capabilities of perception and recognition, knowledge computing, cognitive reasoning, collaborative control and operation, and human-computer interaction to form an open and compatible, stable and mature technology system.

There are 7 research directions for key common technologies for major needs, and 1-2 projects are to be supported in each direction. Teams with a clear application background and a basis for technological breakthroughs are encouraged to apply.

2.1 Generalizable Domain Knowledge Learning and Computation Engine

To address the needs of new cross-border integration and knowledge innovation services, and to overcome the key technologies required for the establishment of large-scale and comprehensive knowledge centers. We will break through the core technologies of knowledge processing, deep search and visual interaction, and form the capabilities of concept recognition, entity discovery, attribute prediction, knowledge evolution and relationship mining to achieve the automated acquisition of continuous growth of knowledge, and form the ability of autonomous induction and learning from data to knowledge and from knowledge to service. Service validation in 1-2 knowledge-intensive areas to meet or exceed the average Q&A service level of domain experts.

2.2 Technical Systems for Cross-Media Analytical Reasoning

To meet the major needs of cross-media content regulation, situation analysis and cross-modal medical analysis, we study the theory, model and acquisition method of unified representation of multi-media knowledge, build a billion-plus level knowledge map and analysis and reasoning technology to adapt to the evolution of cross-media content, and establish a generalization mechanism from directed reasoning to general reasoning. Implement traceable and interpretable cross-media intelligent reasoning in 1-2 typical application scenarios, with accuracy rates exceeding that of intermediate experts in the field.

2.3 Active scene perception techniques under cognitive tasks

Research on active visual perception, 3D modeling and localization of natural scenes for cognitive tasks such as target search, scene analysis and interpretation in complex environments; research on acoustic environment detection and active speech perception based on auditory feedback mechanisms in noisy scenes; research on cognitive techniques for active discovery of new targets and their attribute knowledge from natural scenes with audiovisual synergy. Build experimental platform for typical scenarios and perform functional verification.

2.4 A Study on the Convergence of Group Intelligence Stimulation for Crowd-Based Software Development

To study the collaboration and evolution of group intelligence communities and the decomposition and adaptation of group intelligence tasks for large-scale complex group intelligence innovation activities such as groupware development; Research on the analysis and evaluation, quality control and reuse fusion of innovative products of group intelligence; Research techniques for code annotation, test verification, and defect fixing for Groupware software artifacts. Research on the convergence mechanism and technology of group intelligence stimulation in open source communities, promote the formation of a million-scale group intelligence innovation and talent cultivation ecology for specific fields, and contribute to the establishment of AI technology and application ecology.

2.5 Research on human-machine collaboration hardware and software technologies

The research is aimed at the application scenarios of human-machine collaboration such as intelligent manufacturing and autonomous driving, and the research constructs a human-machine collaboration technology platform with integrated software and hardware. Research on models and methods for adapting real-world context understanding and collaborative decision making; Research on novel learning methods that mix human intuition, experience, and behavior from human-computer collaboration; Research on novel hybrid computing architectures and intelligent hardware and software that can naturally understand environments and situations and process large-scale knowledge, etc.

2.6 Autonomous intelligent precision sensing and manipulation of unmanned systems

Research on multi-sensor information fusion based cooperative perception methods in unconstrained environments, and research on semantic modeling and understanding methods for large scale scenes to achieve map construction, thorough perception and dynamic cognition of complex environments; research on fast and accurate segmentation, detection, localization, tracking and identification methods for multi-source heterogeneous perception objects in complex scenes. Establish or use existing autonomous intelligent systems for technology validation to achieve natural, accurate and safe interaction and precise control in autonomous intelligent unmanned systems.

2.7 Dexterous and precise operational learning of autonomous intelligences

To study the needs for autonomous operations in complex unmanned production systems, we study intelligent human-machine interaction-based skill transfer and efficient demonstration of complex dexterous and precise operations; we study machine learning and skill generation for complex skills such as grasping, alignment, convergence and loading; we study dexterous operation motion planning and coordination control of autonomous intelligent bodies to realize motion mapping from skills to dexterous operations; we study multi-level operation skill representation methods to realize knowledge-based representation of complex skills; we conduct technical verification of dexterous operation skill learning around typical scenarios such as precision assembly.

3. Smart Chips and Systems

Focusing on the key aspects of AI industrial development and application ecological infrastructure, the research focuses on new perceptual devices and systems, key technical standards for artificial neural networks and open source AI platforms from the perspective of AI innovation platforms and basic support.

There are three research directions for smart chips and systems, and each direction will support 1-2 projects. Industry-academia-research teams that already have a good industrialization base are encouraged to participate in the application.

3.1 New sensing devices and chips

Research on the signal processing and information processing mechanisms that can simulate the perceptual channels of biological sight, hearing, touch and smell; develop new perceptual devices, chips and corresponding neural network algorithm models for representing, processing, analyzing and recognizing perceptual information; develop perceptual systems with functions similar to and beyond those of living beings and realize functional verification.

3.2 Neural Network Processor Key Standards and Verification Chips

The design of neural network computation instruction sets to support training and inference, the development of neural network representation and compression standards, the development of efficient basic algorithm libraries and development interface standards on this basis, the realization of supporting development tool chains, the establishment of open chip platform standards that do not depend on the specific chip implementation, and the unification of hardware and software system interfaces. Implement verification chips and example applications that support the above instruction sets, algorithm libraries, standards and development interfaces.

3.3 Artificial Intelligence Open Source Open Infrastructure Platform and Intelligent Operating System Prototype

Research on intelligent hardware resource management technologies such as intelligent sensor devices, intelligent processing chips and intelligent controllers, and develop an open source foundation platform for artificial intelligence that supports a variety of heterogeneous hardware. Research on intelligent algorithms, knowledge bases, and other intelligent software and data resource management technologies, and develop a common open source algorithm library for artificial intelligence, a model library, and a basic software platform for human-computer interaction. Support the distributed distribution and scheduling of large-scale intelligent tasks, establish an open source AI ecology that stimulates innovation, organic integration and rapid application, and support the development of intelligent operating systems and other basic software and core hardware.

1、Google Tencent Facbooks latest musthave neural network translation NMT
2、Resident Evil 7 for VR Final Word ah Im really getting set up after all
3、Video Developing GraphQL Services with ASPNETCore
4、The essence of the data Jiangtang index is sorting
5、Dry McKinseys way of thinking and communication rules recommended collection

    已推荐到看一看 和朋友分享想法
    最多200字,当前共 发送