DARPA's research and development budget for fiscal year 2020 saw a sharp increase in investment in AI applied research

DARPA's research and development budget for fiscal year 2020 saw a sharp increase in investment in AI applied research

tenco 2019-04-12

In March 2019, the trump administration released its fiscal 2020 budget request.Under the budget bill, the total national security budget for fiscal 2020 would increase by $34 billion to $750 billion, up 5% from the previous year.The defense department received $718 billion in funding to implement national security and defense strategies, and DARPA's research and development funding for fiscal 2020 grew steadily, totaling $3.556 billion, up 3.77 percent from fiscal 2019 and 15 percent from fiscal 2018.


Analysis of DARPA's fiscal 2020 r&d budget


1. Basic research, about 486 million us dollars, including national defense research science and basic combat medicine research science, accounting for 13.7% of the total budget, an increase of 3.7% over the previous year.


2. Applied research, about us $1.469 billion, including biomedical technology, information and communication technology, biological warfare and defense, tactical technology, materials and biotechnology, and electronic technology, accounting for 41.3% of the total budget, an increase of 4.3% over the previous year.


3. Advanced technology development, about $1.519 billion, including advanced aerospace systems, space programs and technologies, advanced electronic technology, accusation and communication systems, and network-centric warfare technology, accounting for about 42% of the total budget, an increase of 3.3% over the previous year.


4. Management support, about $81.71 million, including task support, small business innovation research, etc., which accounted for the least proportion and increase.Horizontal comparison, we can see that the advanced technology development still occupies the majority of the budget;Longitudinal analysis, compared with the previous year, the four parts have different degrees of growth, of which applied research has the largest increase.


In terms of the project budget, artificial intelligence, machine learning, multi-domain war-related technologies, biotechnology, unmanned aerial vehicles (uavs), swarm computing and human-computer collaboration are still the major concerns of DARPA. Among them, the investment in artificial intelligence and machine learning projects has increased by the most, totaling $409 million.The projects DARPA plans to invest in in fiscal 2020 will be in the areas of "information and communication technology" for applied research in artificial intelligence and "mathematics and computer science" for basic research.


DARPA's key ai investment program for fy2020


From basic research to applied research to the integration and application of advanced technology development, artificial intelligence has become a top priority for DARPA's research and development efforts.According to the budget allocation of the project, the top ai projects funded by DARPA in fiscal year 2020 are:


1. "explainable artificial intelligence" (XAI) project ($26.05 million).The project is developing a new generation of machine learning techniques to form a basic theory to explain the conclusions reached by ai.In fy2019, the project team will evaluate the performance of the initial prototype system, improve the interpretable machine learning method, improve the computational model of artificial intelligence interpretive theory, and focus on optimizing the interpretable machine learning user interface and integrating it into the prototype system in fy2020.


2. "reliable autonomy" project ($25.55 million).The project aims to ensure that autonomous systems can operate safely as expected, improve the reliability of autonomous machine technologies and accelerate their application.In fy2020, the project team plans to develop an extensible method to verify the security attributes of the autonomous system.


3. AIDA ($25 million).This project will carry out the research on the important data screening of fuzzy multi-source information flow and develop a multi-hypothesis "semantic engine", which will generate explicit interpretations of real-world events, situations and trends based on data obtained from various sources.To solve the complex data, contradictions and potential cheating problems in today's data environment.In fy2020, the project team plans to use real-world data to evaluate the validity and completeness of the generated assumptions by developing an intuitive interface that allows users to modify the extracted semantic elements and generated assumptions at any stage of the analysis.


4. "automatic knowledge extraction" (AKA) project (new project, $24.1 million).AKA will develop technologies to automate the integration of disparate data and information sources into a single entity, taking advantage of advances in semantic technology and machine learning to enable machines to perform entire data integration functions without human intervention.AKA technology is mainly used to help battlefield warfighters automatically establish and maintain a wide range of military, political, economic, social and cultural knowledge of the target area.


5. "accelerating artificial intelligence" (AAI) project (new project, $24.1 million).The AAI program seeks to move beyond commercially driven AI development, improve defense processes, and increase the speed at which the department of defense can rapidly adapt and deploy new technologies and capabilities to meet significant national security challenges.AAI application areas include: machine enabling technology;Automated methods for certification of military software systems;And motor and sensory techniques for patients with impaired central nervous system.


6. "ensuring AI anti-spoofing reliability" (GARD) project ($17.24 million).The project aims to develop a new generation of defense technology against adversarial spoofing attacks against machine learning models.In fy2019, the project team will identify the cause of the vulnerability and set indicators for the robustness of machine learning algorithms.Fy2020 plans to develop ways to improve the robustness of machine learning systems against spoofing data and hostile attacks, and to establish a machine learning risk assessment testbed through challenging questions, attack simulations, and open competitions.


7. "basic artificial intelligence science" (new project, $16.5 million).The project will lay the foundation of basic scientific research for mastering and quantifying the performance expectations and limitations of AI technology.The project is scheduled to launch in fiscal year 2020 to identify and develop artificial intelligence architectures that make full use of observation and experimental data, simulated data and prior knowledge, and to design and test physics-based initial machine learning architectures, algorithms and other works.