Two New Mexico State University professors are working on one of the biggest challenges in artificial intelligence – how to build an AI you can trust with life-or-death decisions.
Son Tran, NMSU computer science professor and department head, and Stylianos Vasileiou, NMSU assistant computer science professor, received a $600,000 grant over two years through a special research area under the Defense Advanced Research Projects Agency (DARPA). It’s called CLARA, Compositional Learning-And-Reasoning for Automated Complex Systems Engineering, a fundamental research program is designed to tightly integrate artificial reasoning and machine learning components to create high-assurance machine intelligence.
Their proposal is titled “SPARTA: Splitting-based Polynomially-bounded Automated Reasoning with Tensorized Adaptation.” NMSU is working with two other research groups, one led by Chitta Baral from Arizona State University and the other by Yuliana Lierler and Jorge Fandinno from University of Nebraska at Omaha.
They are among only 10 teams in the country awarded grants to work on this project.
“Integrating these branches of artificial intelligence will provide the speed and flexibility of machine learning with verifiability based on artificial reasoning proofs that have strong logical explainability and computational tractability,” Tran said. “We will focus on developing efficient and scalable approaches to combine machine learning approaches with logic programs. Such systems will provide explanations for their answers, which is a key difference compared to contemporary machine learning systems.”
The team’s goal is to develop high assurance and broadly applicable methodologies for combining Automated Reasoning and Machine Learning.
“Right now, AI generally splits into two camps. One is great at spotting patterns quickly, and the other is great at following strict logic. Combining them gives you the best of both worlds,” Vasileiou said. “You are getting an AI that can think fast on its feet while still showing its work in a way humans can verify and trust.”
Right now, building an AI that is safe enough for critical jobs takes massive amounts of custom engineering. CLARA aims to create one universal foundation that is rock-solid, mathematically proven, and ready to adapt to almost any complex problem.
CLARA aims to build a reliable, "plug-and-play" AI engine that can be reused across dozens of high-stakes industries without needing to be rebuilt from scratch every time.
-30-
CUTLINE: From Left: Son Tran, NMSU computer science professor and department head, and Stylianos Vasileiou, NMSU assistant computer science professor, are working together under a two-year $600,000 grant from the Defense Advanced Research Projects Agency (DARPA) to create more trustworthy AI solutions. (NMSU photo by Azure Mitchell)
CUTLINE From Left: Son Tran, NMSU computer science professor and department head, and Stylianos Vasileiou, NMSU assistant computer science professor, discussing their research to make AI more trustworthy. (NMSU photo by Azure Mitchell )