The John and Marcia Price College of Engineering held its annual Utah AI Convergence on June 23rd and 24th. The free-to-attend summit brought together experts in multiple facets of artificial intelligence for a series of talks, panel discussions, and workshops. Students from across the state also presented their AI research in a poster competition.
Speakers included faculty members from all departments of the Price College of Engineering, as well as researchers from around the University, studying topics as diverse as cardiovascular medicine, pharmacology, and philosophy. They were joined by representatives from universities around the state including BYU and Utah State University, major tech companies, such as Google, Anthropic, Cisco, and members of national laboratories.
“Artificial intelligence is at the heart of today’s most important conversations, which is why we’re assembling such a wide range of voices. All of us need to come together to understand AI’s challenges and opportunities,” said Sneha Kasera, incoming Director of the Kahlert School of Computing and chair of the Price College’s AI committee that is organizing the summit.
DAY 1
The AI Convergence kicked off with a keynote address by Girija Narlikar, Director of Engineering, Head of AI Safety and Security at Google. Narlikar has led efforts around Generative AI Safety and Security Engineering, focusing on building secure, reliable, and trustworthy AI systems that impact millions of users worldwide. Her talk — “Building AI that Powers and Protects the Digital Workspace” — detailed the way users need protection from AI systems overstepping their bounds, as well as how AIs need protection from malicious users who might try to exploit them.
Curricular Response to AI
Moderator: Vivek Srikumar (Kahlert School of Computing)
Panelists: Stephanie Menhart (CLEAR Program)
John Regehr (Kahlert School of Computing)
Swomitra Mohanty (Chemical Engineering)
Vineet Pandey (Kahlert School of Computing)
The first panel of the Convergence tackled a thorny issue: how AI is disrupting the college classroom. The panelists drew on their own teaching experiences, highlighting the different approaches different subjects require. But whether writing essays or code, the consensus was that curricular responses to AI must emphasize transparency, so that teachers can tell when students are using AI to circumvent assignments rather than enhancing learning.
Workshop: AI + Advanced Manufacturing
Moderator: Ashley Spear (Mechanical Engineering)
Panelists: Victor Walker (Idaho National Labs)
Trevor Shoemaker (Hill Air Force Base)
Benjamin Garcia (Weber State/MARS Center)
Eric Mortenson (Northrup Grumman)
In this lunchtime workshop, moderator Ashely Spear guided a discussion on advanced manufacturing, — techniques that produce otherwise impossible materials, or allow them to be produced at unprecedented scales. These materials are particularly relevant in challenging environments, such as space. With representatives from national labs and leading members of industry, the panel explored how AI is accelerating materials discovery through automation efficiencies. The proximity to these resources and expertise is transforming Utah into an advanced manufacturing hub, with specializations in aerospace/defense, energy, and health/medicine feeding back on one another.
Advanced Manufacturing Academic Panel
Moderator: Ashley Spear (Mechanical Engineering)
Panelists: Long Luo (Chemistry),
Sterling Baird (BYU)
Yongzhi Qu (Mechanical Engineering)
A counterpart to the previous session, Spear also moderated this discussion of advanced manufacturing in academic settings. At the U, artificial intelligence is assisting with materials research in surprising ways, for example, by collating metadata on samples as they are shipped between colleagues, eliminating the need to manually search multiple email threads. AI is assisting in more direct ways, automatically scanning materials for anomalies and using logic to determine if they are indeed defects.
AI in Energy
Moderator: Salah Faroughi (Chemical Engineering)
Panelists: Chris Ritter (Idaho National Labs)
Masood Parvania (Electrical & Computer Engineering)
John Hedengren (BYU)
Kody Powell (Chemical Engineering)
The complexity of energy generation and delivery systems lend themselves to the efficiencies and insights enabled by artificial intelligence. From generation to grid management, the ability to optimize the relationship between supply and demand unlocks new capabilities. The panel ranged from academic research to real-world implementations on industrial and grid-scales. Throughout the panel, a theme emerged: concerns about data centers’ high energy consumption must be balanced with the potential for those AI systems to generate gains in energy efficiency elsewhere.
AI in Health
Moderator: Karli Gillette (Biomedical Engineering)
Panelists: Jake George (Electrical & Computer Engineering)
Yue Lu (Pharmacy)
Ravi Ranjan (Cardiovascular Medicine)
Tucker Hermans (Kahlert School of Computing),
Florian Solzbacher (Electrical and Computer Engineering, Blackrock Neurotech)
The final panel of Day 1 convened experts from multiple healthcare disciplines. Artificial intelligence intersects with their fields in a variety of ways, from revealing hidden information that clinicians miss, to personalizing cancer treatment, to enabling entirely new ways for patients to interact with the world, as with Brain-Computer Interfaces and “neuroprosthetics.” Perhaps counterintuitively, these advances in artificial intelligence have the effect of centering the human experience in health and medicine, throwing into relief what machines still can’t replicate.
DAY 2
Industry Panel Focus
Moderator: Sneha Kasera (Kahlert School of Computing)
Panelists: Pierre Emmanuel Gaillardon (ChipNexus, Electrical & Computer Engineering)
Porter Jenkins (Delicious.AI, Enzy)
David Williams (Lease End)
Royal Hansen (Google)
Jodee Varney (Cisco)
Day 2 of the Convergence began with representatives from the world of business, both those who develop AI products and those that implement them in diverse fields. From computer vision systems that tell managers how shelves are stocked, to decision-trees for helping users make financial decisions, these systems are enabling entirely new business models. The panel also features speakers who are more directly in the business of artificial intelligence, developing models and agents that work together, enhancing security and making better decisions.
Revolutionizing Utah’s AI Infrastructure
Panelists: William Miller (Scientific Computing and Imaging Institute)
Greg Jones (Director of Corporate Engagement for Scientific Computing, Technology, and AI, University of Utah)
As a major research institution, the U has unusually large computing needs — The U’s Center for High Performance Computing’s own supercomputer recently debuted at 159 on the TOP 500 list. The U also has a special relationship with the state and its scientific and technical needs. The U is therefore establishing new computational platforms and integrating them with growing national networks. Their efforts are part of a larger community-building effort, entailing monthly meetings with stakeholders from across the state and beyond.
Workshop: AI-Supported Socratic Activities
Swomitra Mohanty (Chemical Engineering)
Director of Undergraduate Affairs Swomitra “Bobby” Mohanty led this lunchtime workshop, delving deeper into his presentation from the previous day. In Mohanty’s classrooms, students engage with a custom-built “socratic” AI system. Built with guardrails that prevent the system from revealing the answers to problem sets right away, it instead guides students through a series of questions that test their understanding of the material.
AI Systems
Moderator: Kenneth Marino (U of U Kahlert School of Computing)
Panelists: Luis Garcia (Kahlert School of Computing)
Mary Hall (Kahlert School of Computing)
Ganesh Gopalakrishnan (Kahlert School of Computing)
Jacob Austin (Anthropic)
This panel discussed the interplay between AI hardware and software, detailing the ways computer architecture can maximize AI’s abilities. The panelists promoted a cyberphysical view of AI development, in which new autonomous capabilities are retrofitted into real-world workflows in mindful ways. This can be straightforward as ensuring agents are not mixing units when talking to one another. Ultimately, co-design of hardware and software is necessary for both efficiency and security.
Ethical AI
Moderator: Sadia Khan (U of U Responsible AI Initiative)
Presenters: Zachary Boyd (State of Utah AI Policy Office)
Eliane Wiese (Kahlert School of Computing)
Heather Holmes (Chemical Engineering)
Thi Nguyen (Philosophy)
The final panel of the convergence drew on the most diverse array of experts, examining the ethics of using and implementing AI in a variety of contexts. There are multiple potential frameworks in play, some of which may be counterintuitive. On a society scale, policymakers and technologists must both balance the need to prevent harm with the opportunity costs of inaction. On the individual scale, those implementing AI systems have an obligation to provide ethical context to enable better human decision-making.
Student Presentation Winners
Lightning Talks
3rd Prize, presented by XRDNA: “Comparing Reinforcement Learning Models To Explain Human Neural And Behavioral Measures Of Decision-Making Under Risk” (Niloufar Shahdoust)
2nd Prize, presented by Blackrock Neurotech: “Coordinated Flexibility In Grid Interactive Ai Data Centers” (Mahsa Omri)
1st Prize, presented by Amplify|SI: “Computer Vision-Driven Plant Inventory, And Seeding Tray Viability Assessment.” (Omisha Sapra)
Posters
3rd Prize, presented by SummitLine Construction: “TARget: Topology-Aware Fusion-based Radio Frequency Circuit Functional Modeling Using Graph Neural Networks” (Soroosh Noorzad)
2nd Prize, presented by Google: “Crisis-Intel: A Generalizable Multi-Agent System for Crisis Intelligence” (Prachi Aswani)
1st Prize, presented by Cisco: “Learning to Adapt: Continual Reinforcement Learning Under Progressively Weakened Musculoskeletal Modes” (Sonny Jones)
People’s Choice Award, presented by the John and Marcia Price COllege of Engineering: “Data Driven Inverse Modeling of Analysis of Wilderfire Post-Event Reports” (Sarah Khan)