Energy Department: Help Us Design the Next A.I. Grand Challenge

The U.S. Department of Energy is looking for input from artificial intelligence researchers on how to design a series of open competitions that spur innovation and bolster U.S. technological superiority in A.I.

The DOE’s Artificial Intelligence and Technology Office (AITO) is exploring launching one or more Grand Challenges aimed to “push the frontiers of AI technologies, creating the next generation architectures, infrastructure, tools and approaches,” according to a request for information released on June 10.

The AITO is seeking input from academia, the private sector, and national laboratories on how to design competitions that promote technological breakthroughs in A.I. fields, such as computer vision, natural language processing, distributed intelligence (“swarms”), and edge computing among many others, according to the announcement. The deadline to submit responses is July 10.

Photo: Getty Images

The office’s top priorities include research into:

• Novel methods of learning, including continuous learning models, working with unlabeled or unstructured data, limited data, and the development of A.I. with common sense;
• Development and validation of trustworthy A.I., including secure, tamper-proof algorithms capable of explaining their reasoning and mitigating unintended bias;
• Applications for and defenses against adversarial A.I.; and
• Tools to enable better data management and data sharing.

“AITO requests responses that relate one or more of the listed breakthrough AI technology areas to aid in tackling our Nation’s largest issues, such as curing cancer, preventing or mitigating epidemics, and ensuring grid resiliency,” according to the RFI.

The Trump administration directed federal agencies to “drive technological breakthroughs in AI” with a February 2019 executive order to “promote scientific discovery, economic competitiveness, and national security.” Soon after, Energy Secretary Rick Perry established the AITO in September 2019 as a hub for A.I.-related research across the department’s many offices and national laboratories.

DOE will invest an estimated $162 million on A.I.-related research and development in fiscal 2020, per a supplement to the president’s budget.

The original Grand Challenge was a 2004 contest sponsored by the Defense Advanced Research Projects Agency (DARPA) to design an autonomous vehicle capable of navigating a 142-mile course through the California and Nevada deserts. Although no team was able to complete the first Grand Challenge, and the $1 million prize went unclaimed, the project arguably set the scene for advances in the technology and the emergence of a commercial market for self-driving cars.

DARPA continues to employ the model of its first challenge to push technological boundaries in fields such as cybersecurity, robotics, and advanced radio frequency communications.

Although Grand Challenge prizes tend to be relatively small, compared to most government contracts, innovative proposals are often rewarded with higher-value work with the government. For instance, the winner of the 2016 DARPA Cyber Grand Challenge, ForAllSecure Inc., subsequently received $6.3 million in other transaction agreements with the Defense Innovation Unit. The company then won a $45 million follow-on contract to supply the Defense Department with its automated software vulnerability testing tools.

To contact the analyst: Chris Cornillie in Washington at ccornillie@bgov.com

To contact the editor responsible: Daniel Snyder at dsnyder@bgov.com