The recipients are Stefan Lee and Somayeh Pasebani of the College of Engineering and Jeff Hazboun of the College of Science.
Lee, an assistant professor of computer science, will use his five-year, $565,000 award to explore ways to enable artificial intelligence systems to learn complex, interactive tasks by reading instruction manuals written in natural language.
Typically, AI and machine learning systems figure out how to behave in complex, interactive environments through reinforcement learning, which involves significant trial and error. For tasks like driving a car, operating new appliances, or using new software, the process is inefficient and potentially dangerous.
“As a society, we’ve put a great deal of effort into developing learning materials to help others understand how to interact with new
Giving AI systems the ability to comprehend and predict complex behaviors described in everyday language could significantly expand their capacity to perform digital and physical labor and reduce the cost to develop AI that can quickly learn how to conduct new tasks. For example, if a home-assistant robot is directed to operate a new dishwasher and uses trial and error to achieve its goal, the results could be unpredictable and unsatisfying.
“Do you want the robot to press each button on the dishwasher over and over until it figures out how to use it, or would you rather it just read the user manual, which contains all of the important information, and quickly learn how to operate the appliance?” Lee said.
He added that meeting future demands for both digital and physical work may be challenging, so the easier it is to train AI systems to
“Developing AI systems that can quickly gain new skills and adapt to the needs of an aging population will be of great value to our society,” Lee said.
Pasebani, an associate professor of advanced manufacturing, will use her five-year, $756,000 grant to explore new strategies for the fabrication of multimetal components via laser-based additive manufacturing processes, or AM. Additive manufacturing is also known widely as 3D printing.
Multimaterial AM can be particularly challenging, according to Pasebani, because of the dynamic metallurgy of laser-matter interactions.
“There’s often a mismatch of mechanical, physical, chemical, and thermal properties between the alloys, and the greater the mismatch, the greater the probability of failure at the interface between them,” she said. “We want to develop an interface in which the mismatch is as small as possible.”
To do that, Pasebani intends to combine computational modeling with experimental processing to better understand the thermal and fluid-flow behavior within the interface between melted dissimilar alloys. The work will enable the composition and microstructure of 3D-printed multimaterial components to be spatially varied and built with great precision, thereby enabling the fabrication of parts with tailored properties.
“With the increasing demand for higher-performance metallic components, we would like to improve their properties and push the envelope of what’s possible with additive manufacturing,” Pasebani said. “We intend to establish a practical methodology for understanding joining and failure mechanisms in multimetal additive manufacturing that will enable us to simultaneously achieve design freedom and effectively dictate the properties exactly where they’re needed in a component.”
Pasebani’s Metal Additive and Gradient Microstructure Alloy Lab will also use AM processes to design high-performance novel alloys. Customized, complex, high-performance alloy components are used in a variety of industry sectors, including aerospace, defense, health, manufacturing, and energy.
The project also aims to integrate research with teaching, mentoring, and training students at different levels, especially women and underrepresented minorities. With the understanding that participation in the fine arts activates creative thinking, the project offers summer programs involving hands-on activities based on art and STEM integration to encourage K-12 students to pursue science and engineering fields.
Hazboun, assistant professor of physics, will use his $400,000 award to continue his work with gravitational waves. In June 2023 he was part of a collaboration that made worldwide news with the announcement that low-frequency gravitational waves are permeating the universe.
“The announcement by the group I’m part of, the North American Nanohertz Observatory for Gravitational Waves, and pulsar timing array collaborations around the world demonstrates that the arrays are moving into an unprecedented sensitivity regime for gravitational wave signals in the nanohertz frequency range,” Hazboun said. “Pulsar timing arrays can see further back in time than the cosmic microwave background radiation and allow us to survey the population of supermassive binary black holes as a stochastic gravitational wave background. Detection of the background heralds the coming of individual source detections.”
Hazboun’s project includes curriculum development and the implementation of a summer workshop in astrophysics-themed data analysis designed to foster inspired teaching, stimulate excitement in PTA research, facilitate the learning goals of undergraduate and graduate students, and support the community college students’ transition into four-year schools.
The NSF CAREER program supports “early-career faculty who have the potential to serve as academic role models in research and education and to lead advances in the mission of their department or organization,” according to the NSF website.
~Steve Frandzel, College of Engineering