In This Story
AI is everywhere. It emphatically wends its way and creeps into daily life, from parking cars to vending machines. In some cases, the use of AI needs to be closely monitored to ensure all safety and precautionary measures are taken into consideration.
Thanhvu H. Nguyen, one of the newest professors in the Department of Computer Science, arrived at George Mason University in 2021 and knew of Mason’s diversity and proximity to one of the largest tech hubs in the nation.
Nguyen recently won the National Science Foundation (NSF) CAREER Award "NeuralSAT: A Constraint-Solving Framework for Verifying Deep Neural Networks.” Nguyen will work with his student Hai Duong, and a select team of students to develop technology that ensures AI machine learning is robust, safe, and unbiased.
“In certain scenarios, where AI is used for things like controlling airplanes or autonomous vehicles, any mistake could prove fatal,” says Nguyen. “This grant work will look at the deep neural networks (DNN) embedded in AI/machine learning technology and develop technology to prevent errors and ensure safety.”
According to the project’s abstract, the use of DNNs - a layered network that processes complex data - have emerged as an effective approach to tackling real-world problems. However, just like traditional software, DNNs can have ‘bugs’ and be attacked.
This can be especially concerning, when it comes to the use of AI in driving cars, for example. Although it could eliminate some human error, Nguyen says any faulty DNNs could throw an AI program – and the car – off the road when it comes to something unexpected, like bad weather and icy roads.
“We all depend on AI, whether we realize it or not,” says Nguyen. “It’s a part of our lives. Whatever I can do to make it safer, is a benefit to my family and the society.”
Since coming to Mason, he has helped to bring in around $1.73M in external funding. Nguyen is excited to work with his students and the group of computer science faculty, noting how everyone brings a fresh perspective to the table.
The project runs for five years with a total of $510,509 in anticipated funding. Nguyen sees the project as helping society as a whole by improving the reliability of systems embedding DNNs. The research conducted throughout the project will allow AI and machine learning researchers and users to improve their DNNs and deploy them with confidence.