A University of Texas at Arlington researcher will build a safety net for scientific experiments guided by artificial intelligence.\u00a0<\/p>\n
The U.S. Department of Energy announced Wednesday that UTA will lead a $750,000 project\u00a0 developing safeguards for AI used at particle accelerators and other advanced research facilities.\u00a0<\/p>\n
UTA will receive $500,000, with the remaining $250,000 supporting its partners, said Habeeb Olufowobi, assistant professor of computer science and engineering, in an email to the Fort Worth Report. The nine-month project will run from Sept. 2006 through May 2027.\u00a0<\/p>\n
Olufowobi, the project\u2019s leader, said negotiations with the DOE are complete, and the budget has been approved.\u00a0<\/p>\n
Many DOE research facilities rely on an open-source software called the Experimental Physics and Industrial Control System. The software serves as a digital control room, allowing scientists to monitor and operate complex equipment.\u00a0<\/p>\n
Scientists at those DOE facilities increasingly use AI to interpret data and recommend adjustments in real time. But an AI model could fail, be configured incorrectly or face a cyber attack, which might allow unreliable recommendations to influence how scientists conduct or interpret an experiment.\u00a0<\/p>\n
\u201cThis project addresses a national priority by helping ensure that AI can be deployed safely and reliably in some of the world\u2019s most advanced scientific facilities,\u201d said Hong Jiang, chair of UTA\u2019s department of computer science and engineering, in the press release.<\/p>\n
Olufowobi\u2019s team plans to build a checkpoint between AI models and scientific operations. The \u201ctrust layer\u201d would monitor an AI model\u2019s behavior, assess the risk of its recommendations and prevent suspicious outputs from affecting an experiment.<\/p>\n
The framework preserves human oversight by showing scientists why an AI recommendation was flagged and allowing them to decide what happens next.<\/p>\n
Researchers from Los Alamos National Laboratory, the University of Colorado Colorado Springs, Washington State University and Metro State University will work with UTA. Two UTA graduate students will also participate.\u00a0<\/p>\n
“This award reflects UT Arlington’s role in solving problems of national importance,” UTA President Jennifer Cowley said, adding that the project will give students \u201chands-on experience at the forefront of AI and cybersecurity research.\u201d\u00a0<\/p>\n
The Phase I project is one of 12 to be led by Texas organizations<\/a> among 278 selected for the\u00a0 Genesis Mission. Of those, 168 are led by universities nationwide. The federal initiative aims to transform scientific research<\/a> by combining AI, supercomputing, scientific data and advanced instruments.<\/p>\n \u201cThe extraordinary response to this Genesis Mission application process demonstrates that America\u2019s scientific community is ready to reimagine how discovery happens,\u201d Dar\u00edo Gil, DOE Under Secretary for Science, said in a press release.\u00a0<\/p>\n Olufowobi said the project\u2019s primary measure of success will be whether it advances to Phase II. Only a small fraction of Phase I teams will do so. The DOE envisions Phase II budgets running for three years and being three to five times the size of Phase I annual awards<\/a>.<\/p>\n Dang Le is the higher education reporter for the Fort Worth Report. Contact him at dang.le@fortworthreport.org<\/a>.\u00a0<\/p>\n The Fort Worth Report partners with Open Campus on higher education coverage.<\/p>\n The Report\u2019s higher education coverage is supported in part by major higher education institutions in Tarrant County, including Tarleton State University, Tarrant County College, Texas A&M-Fort Worth, Texas Christian University, Texas Wesleyan University, the University of Texas at Arlington and UNT Health Fort Worth.\u00a0<\/p>\n