Reposted from College of Engineering.

The University of Utah’s John and Marcia Price College of Engineering hosted its first AI Summit for Utah on June 18, bringing artificial intelligence researchers from every department together for an all-day symposium. They connected with more than 400 faculty, students, industry partners and policymakers from across the region. 

Charles Musgrave, dean of the Price College of Engineering, kicked off the sold-out proceedings held at the S. J. Quinney College of Law. As a chemical engineer, Musgrave has been using machine learning techniques for more than a decade in the pursuit of developing new materials. But from the vantage of the dean, and at the precipice of a new technological revolution, the prospects for artificial intelligence are limitless. 

“Those who lead in AI will lead in science, economics, national security and innovation,” Musgrave said. “But if we do it right, we’ll also lead in art, entertainment and personal fulfillment.” 

U trustee Steven Price detailed the ways university research redounded to the public’s benefit over the past decades. “These are the ingredients that fertilized Utah’s growth,” he said. “We’re in a movement and a moment. The movement is AI, and the moment is now. AI is moving fast and we have to move faster.” 

In addition to a series of poster presentations by more than 60 students from across the state, the symposium was organized into four panels, each bookended by a series of one-minute “lightning talks” from a selection of the student poster presenters, as well as Q&A sessions with the audience. 

AI in sensing, seeing and securing the world

Weilu Gao, Department of Electrical & Computer Engineering, “Machine learning with optics”
Ziad Al-Halah, Kahlert School of Computing, “Multimodal Embodied AI”
Guanhong Tao, Kahlert School of Computing, “Towards Safe and Secure Large Language Models”

Moderator Varun Shankar, assistant professor in the Kahlert School of Computing

Panelists engaged with the ways AI systems are increasingly integrated with physical ones. As such, these systems need new ways of sensing features of their environments and making sure they are interacting in ways that won’t harm the people in them. Weilu Gao’s work adds another dimension to this field; he and his colleagues recently published research on an “optical neural engine” that can potentially accelerate the computation involved in these applications. 

Next-Gen AI: From supervision to autonomy

Jacob Hochhalter, Department of Mechanical Engineering, “Reducing Training Costs with Derivative Informed Data from Hypercomplex Automatic Differentiation”
Daniel Brown, Kahlert School of Computing, “Toward Robust, Interactive, and Human-Aligned AI Systems”
Vivek Srikumar, Kahlert School of Computing, “Human language technology: Can we do better than bigger?”

Moderator Tucker Hermans, associate professor in the Kahlert School of Computing

This panel explored the tradeoffs AI systems must make when it comes to the data they’re trained on. Many machine learning techniques rely on massive data sets, but not all applications have access to them, for practical, economic, or even legal reasons. Beyond reducing the upfront costs of training, new strategies for building up these models will have applications in low-resource and off-the-grid environments.  

The nexus of health, humans and machines

Tolga Tasdizen, Department of Electrical & Computer Engineering, “Towards interpretable AI models in radiology”
Neda Netagh, Department of Electrical & Computer Engineering, “Toward less artificial intelligence”
Alan Kuntz, Kahlert School of Computing, “Autonomous Surgical Robots that Learn from Human Surgeons”
Amir Arzani, Department of Mechanical Engineering, “Scientific machine learning: From no data to large data”
Ashley Dalrymple, Department of Biomedical Engineering, “Reinforcement Learning for Predicting Walking-Related Events”

Moderator Laura Hallock, assistant professor in the Department of Mechanical Engineering

This panel continued the theme of AI’s integration into physical systems, in this case, biological ones. Healthcare applications hold some of the highest promise of AI systems, but technical and ethical issues abound.  Panelists discussed automated analysis of medical images, the neuroscience of vision, physics-based models of blood flow and of tumor tissue, and the biomechanics of walking.

“Bridging” AI and infrastructure

Ryan Johnson, Department of Civil & Environmental Engineering, “High-Resolution Snow Mapping with Machine Learning: Pioneering Products to Enhance Season-to-Season Water Supply Forecasting”
Masood Parvania, Department of Electrical & Computer Engineering, “AI for Autonomous Power Grid Operation”
Chenxi (Dylan) Liu, Department of Civil & Environmental Engineering, “Leveraging AI for intelligent transportation systems”
Taylor Sparks, Department of Materials Science Engineering, “The materials of tomorrow, today”

Moderator Cathy Liu, associate professor in the Department of Civil & Environmental Engineering

Participants zoomed out to even larger physical systems: transportation networks and other pieces of civic infrastructure. Here, the multiplicity of inputs and complexity of their interactions are a natural fit for machine learning techniques. Panelists discussed digital twins of the electric grid for training against cyberattacks, the connection between winter snowfall and summer water levels, preventing chain-reaction crashes on highways and how the materials that make up everyday life are now being invented with these technologies.

• Evan Lerner Director of communications, John and Marcia Price College of Engineering
  801-581-5911 evan.lerner@utah.edu