By Rachel Robey
Members of the Wisconsin AI Safety Initiative visited the Rayburn House Office Building to lead presentations and live demos highlighting the dangers associated with artificial intelligence’s growing popularity.
In February, nine members of the Wisconsin AI Safety Initiative (WAISI)—Benjamin Hayum, Elise Fischer, Jackson Kunde, Jeremy Kintana, Mason Baloun, Max Kamachee, Reid Kuenzi, Sam Baumohl, and Will Anderson—visited Washington, DC to participate in the Center for AI Policy (CAIP)’s Advanced AI Expo.
Described by the nonpartisan research organization as a “first-of-its-kind tech exhibition,” the expo gathered artificial intelligence (AI) research groups from 14 universities across the country. In the Rayburn House Office Building on Capitol Hill, researchers presented to Congressional staffers, journalists, and event sponsor Rep. Bill Foster (D-IL-11), a UW–Madison alumnus, on next-generation AI systems and their societal implications.
“Our goal was to educate key decision-makers on the frontier of existing AI security risks,” says Kintana, director of WAISI. “It was a great opportunity to demonstrate our club’s growing role in connecting technical expertise with public policy. Our student leaders not only gained hands-on experience communicating AI risks, but also developed professional connections that will help us with future projects.”
Housed within the Department of Computer Sciences, WAISI is a student organization that hosts regular speaker events with professors and industry professionals, runs semester-long introductory courses on AI safety, and organizes membership groups focused on reading research papers and skill-building.
The two WAISI teams presented on the attack surfaces present in multi-agent language model frameworks and the dangers of voice models that can mimic emotional speech, demonstrating how the proliferation of AI tools may cause security risks to businesses, flood 911 dispatch lines with fake emergency calls, or mislead Congressional representatives through a coordinated campaign of fake constituent calls.
“The exhibition allowed America’s policymakers to experience these advanced technological dangers firsthand,” said Jason Green-Lowe, executive director of CAIP, in a statement. “This exhibition will help motivate Congress to pass legislation protecting the American public against unreliable AI.”
For more information on WAISI membership, visit https://waisi.org/