The U.S. DOE (Department of Energy) has recently raised $8.5 million in funding for the development of five projects. These projects are aimed at increasing the accessibility and reusability of the AI (artificial intelligence) models and data, to drive AI R&D (research & development).
The key focus of the department is to apply FAIR (Findable, Accessible, Interoperable, and Reusable) Data Principles, which will allow science data to drive AI innovations. Originally, the FAIR principles were endorsed and proposed by an international collaboration of scholarly publishers, funding agencies, industry, and universities in 2016.
As per the statement made by Director of the Office of Science at DOE, Dr. Chris Fall, the effective assessment & analysis of mounting data quantities is considered a major scientific challenge currently impacting various organizations. An effective means of enhancing access to data is the recent FAIR Data Principles. The 5 projects announced by the DOE will reportedly apply these principles and thereby maximize the benefits to science.
Range of topics covered by the selected projects include microbial science, high energy physics, materials science, and high-performance computing. These projects will develop frameworks and make science data FAIR, to systematically study the relation between AI models and data. This development is also aimed at attaining a deeper knowledge of AI functioning as well as its applications.
The projects were selected by the competitive peer-review under the DOE Funding Opportunity Announcement, which was titled ‘FAIR Data & Models for Artificial Intelligence & Machine Learning’, along with a companion announcement for the DOE laboratories, which was sponsored by the Office of ASCR (Advanced Scientific Computing Research) within the Office of Science of DOE.
The total funding of $8.5 million in the FY2020 has been raised for these projects, that are expected to last 1 to 3 years in duration. The list of the projects is currently made available on the ASCR website homepage.