Ghoreishi Receives U.S. Army Grant for Advancing Artificial Intelligence Research
Fatemeh Ghoreishi, assistant professor of civil engineering, received a $60,000 grant from the U.S. Army for “BRITE: Bayesian Inference and Preference Learning for Unknown and Time-Sensitive Environments” to further research on AI systems in unpredictable environments.
Fatemeh Ghoreishi, assistant professor of civil engineering, received $60,000 in funding from the U.S. Army to advance her work in Bayesian statistical learning that will allow AI systems to operate with increasing independence in developing and uncertain environments. The program, “BRITE: Bayesian Inference and Preference Learning for Unknown and Time-Sensitive Environments” will run approximately nine months and support graduate students on her team.
In war zones, armies often carry out disaster response operations and knowing the next safest move is vital. To address this, Ghoreishi is exploring how pre-trained AI systems can be effective in unpredictable military environments with intelligent autonomy.
“The proposed research aims to allow AI systems not only to operate autonomously but also to adapt to dynamic and unpredictable environments in events of uncertainty,” Ghoreishi says.
Bayesian statistical learning implements probabilities to quantify the uncertainty in models and decisions and allows updating, predicting, or estimating the uncertainty according to the latest data. Preference learning is a subfield of machine learning that enables AI agents to understand human or user preferences according to their behavioral data patterns.
“As AI continues to impact different aspects of our lives, autonomy expands across different domains while facing increasingly complex challenges,” reads the BRITE project summary. “This can be seen in autonomous vehicles in our transportation systems or in the integration of drones and AI-based agents/robots for disaster response. In all these scenarios, there is a critical need for robust and resilient autonomy that can adapt to dynamic environments, anticipate risks, and make informed and safe decisions in real-time.”
Ghoreishi says the BRITE project will address practical applications of the BRITE project, such as AI providing a military disaster response team with the exact location of war victims in an area to aid them. Another example would be giving the military a series of the safest moving progressions into an uncertain war zone.
“By quantifying defense sources of uncertainty that we might have with AI that takes into account different scenarios, we aim to have the safest reliance in decision-making,” says Ghoreishi.