NITRD Boasts Progress in AI Research

Published: December 11, 2019

Artificial Intelligence/Machine LearningGovernment PerformanceResearch and Development

NITRD released a progress report outlining federal efforts that have taken place since 2016 to help advance AI R&D, particularly as it relates to the National AI R&D Strategic Plan.

The federal government has made strides in the Research and Development (R&D) of Artificial Intelligence (AI) and folks in the Networking and Information Technology Research and Development (NITRD) program are boasting about it.

In 2016, NITRD published an AI R&D strategic plan to help navigate federal investment in “a transformative technology that holds promise for tremendous societal and economic benefit.” The 2016 plan included seven key strategies to boosting the AI research space. Earlier this year, NITRD published an update to the 2016 strategy, which primarily added an eighth strategy to highlight the importance of public-private partnerships in the acceleration of AI.

Last month, NITRD delivered the 2016-2019 Progress Report: Advancing Artificial Intelligence R&D. The report illustrates the important progress agencies have made to deliver federal AI R&D. Filled with a wide-ranging set of programs and activities across the federal government, the report shows a depth of investment in AI. Organized first by category, the progress report provides updates in AI by topical area: science and engineering, defense, the economy, health, and justice and security. Subsequently, the report describes individual agency programs that have made impacts in each category and topical area.

Moreover, the report provides additional agency AI R&D breakthroughs that have boosted the nation’s status in AI.  The following is a sampling of the breakthroughs listed in the progress report and organized by strategy.

R&D Strategy

Agency Breakthrough

1. Make long-term investments in AI research

Researchers at the NIH National Library of Medicine have created an AI algorithm, called DeepSeeNet, to help the screening of age-related eye disease. The deep learning method saves hours of manual performance by doctors with the potential to assist in early disease detection.

2. Develop effective methods for human-AI collaboration

The DAPRA Explainable AI (XAI) program aims to solidify human trust and collaboration with machines by designing systems to explain their reasoning in understandable terms. This function also helps users to spot and correct errors made by the machine, ultimately helping the system become more accurate.

3. Understand and address the ethical, legal, and societal implications of AI

NSF performed various studies on the implication of predictive systems exhibiting discriminatory behavior. This research used AI systems with built-in quantitative measures of fairness. Long-term studies revealed that fairness criteria could still occasionally make certain decisions less fair, elevating the quality of analysis in ethical, legal and social implications of AI.

4. Ensure the safety and security of AI systems

Using large quantities of real and simulated sensor data, coupled with deep reinforcement learning to train an AI system to detect cyber threats, DOE-funded researchers found that systems built with such standards succeeded in rapidly detecting threats with few false alarms. These efforts show a promising future in using AI for cyber defense to secure many digital systems.

5. Develop shared public datasets and environments for AI training and testing

The NASA Earth eKchange (NEX) is a large public database with earth science data, code and publications. NEX-AI is an extension of the database that makes predictions from the data and uses deep learning models to analyze earth images in detail (DeepSAT) and pretrain deep learning models for different satellites (SATnet). The NEX-AI team is working to solve complex problems in land-climate-atmosphere by combining different kinds of machine learning algorithms.

6. Measure and evaluate AI technologies through standards and benchmarks

NIST supported the Materials Genome Initiative by developing a reference set of data and AI techniques to help create advanced materials for manufacturing. The techniques have also helped advance other areas of simulation and robotics for complex factory environments.  

7. Better understand the national AI R&D workforce needs

NSF funded the development of a new Advanced Placement exam in Computer Science Principles, along with a curriculum and associated teacher professional development. The exam has resulted in an increase in diversity in computer science interest and a firm grounding of computer science education at pre-K – 12 levels.

8. Expand public-private partnerships to accelerate advances in AI

Collaborating with a U.S.-International health technology company, the VA implemented an AI Tech Sprint in order to link companies to federal data and create cutting-edge data tools. In exchange for small VA datasets, non-Federal entities can develop tools based on the data that can lead to contracts using larger, richer VA datasets, accelerating development and deployment of AI tools to help veteran health programs.  

Taken as a whole, the report finds that federal AI investments in the last three years reveal the government’s intent on improving in AI through R&D to ensure “prosperity, safety, security, and quality of life for the American people for decades to come.” Collectively, the report identifies three key messages based on its content: 1.) The federal government is substantially investing in the breadth and depth of innovative AI concepts, 2.) The U.S. is benefitting greatly by the diversity of investments made by federal agencies, and 3.) Federal investments in AI R&D have resulted in breakthroughs that are “revolutionizing our society for the better.”

Moving forward, NITRD intends to update the progress report periodically to reveal how the government is meeting the strategies in its AI R&D plans.