Bio:
Ryan Wright co-directs the University of Virginia’s AI Research Initiative and serves as Special Advisor to the Provost for AI. He is the Rolls-Royce Commonwealth Professor of Commerce at UVA’s McIntire School of Commerce and leads the Virginia AI coordination hub, a consortium spanning eight of Virginia’s public universities, 4 state agencies, 23 of the community colleges, and several industry partners.
His research examines how people and AI systems work together, which includes the behavioral foundations of cybersecurity, how people decide whether to trust algorithmic advice, and the design of autonomous agents that keep humans meaningfully in the loop.
At McIntire, Wright teaches a graduate course on building with agentic AI and previously served as Senior Associate Dean for Faculty and Research, Area Coordinator and Associate Dean for Strategic Initiatives. Across all of his roles, his throughline has been the same: helping organizations and the people in them adopt new technology in ways that are both ambitious and safe.
Talk Title: AI Agents, Guardrails, and the Pace of Scientific Discovery
Talk Abstract: AI is moving toward autonomy in science, from chatbots to agents that design experiments and operate instruments. This architecture is emerging across materials science, chemistry, genomics, astronomy, and so on. A planner agent can decompose tasks, specialist agents can execute them, and a critic agent can challenge the results. This talk focuses on two constraints: validation and guardrails. Bioscience and cybersecurity, where agents are often most capable, are also the domains where frontier developers impose the strictest access restrictions, despite the lack of any connection between university research oversight and model access tiers. I will offer three rules for distinguishing capability from marketing and argue that the returns to these systems compound across institutions.