The Princeton Laboratory for Artificial Intelligence recently hosted its inaugural AI Collaboration Symposium, a gathering dedicated to identifying new research settings where AI can have a transformative impact and to helping faculty find collaborators in pursuing these projects.

“Normally when we speak as academics, we talk about problems we know how to solve,” AI Lab Director Tom Griffiths said during opening remarks for the event, held in Robertson Hall on Jan. 22. “Our intent here is to talk about problems we don’t know how to solve.” 

The symposium is part of the AI Lab’s mission to support and expand the scope of AI research on campus, said Griffiths, who is also a professor in the psychology and computer science departments. In addition to hosting such events, the AI Lab provides resources such as administrative support, computing resources, and funding. The AI Lab’s work is supported by three major research initiatives: Princeton Language and Intelligence (PLI), AI for Accelerating Innovation (AI^2), and Natural and Artificial Minds (NAM). 

The symposium opened with an overview of the initiatives to help attendees understand the scope of AI research happening on campus, as well as information about how these ideas are having an impact around the state through the New Jersey AI Hub.

“We’ve been embracing a leadership role because we do think a healthy regional ecosystem is an important part of supporting our research, teaching and service missions,” said University Provost Jennifer Rexford. “Our goal here is really to support innovation.”

Faculty members and scholars presented research in a range of disciplines, including anthropology, English, environmental studies, mathematics and the visual arts, as well as fields more commonly associated with AI, such as computer science, neuroscience, psychology, and electrical and computer engineering.

A series of moonshot talks outlined ambitious visions to use AI to solve substantial research problems. Chesney Snow, lecturer in theater at the Lewis Center for the Arts, is looking to build a team to take on the challenge of developing virtual reality training for law enforcement.

“Can generative AI create dynamic, real-time training scenarios tailored to individual participants?” Snow asked the audience.  “What are the most effective methods for integrating sentiment analysis and biometrics to detect emotional state while maintaining ethical standards?”

The technology would be integrated into Snow’s existing “Performing the Peace” project, which uses therapeutic practices based in the performing arts to foster empathy and build community trust with law enforcement.

“We’re looking for someone who can really help us on the ground level of understanding how we can integrate AI into the VR world,” he said. 

Other faculty members presented talks with the goal of finding collaborators whose expertise could help bring AI techniques they’re already using to the cutting edge.

“We use image analysis and computer vision tools regularly in our research, but we are not trained in AI. We are not experts in this field,” said Mary “Cassie” Stoddard, professor of ecology and evolutionary biology. “I really think adopting the latest AI techniques could boost our research program, but we’re not sure about next steps.”

Her research could benefit from AI in several areas, she said. For example, her lab uses time-lapse cameras to study how hummingbirds react to climate change. Although they use a computer vision tool to analyze the images, the results must be double checked for accuracy.

“Using improved AI, I think we can make this process more accurate and efficient,” she said. “We’d also love to design AI tools that could distinguish between male and female hummingbirds, and maybe even identify unique individuals in these images.”

Another talk explored AI as a tool to improve student learning on campus. Vincent Chanethom, lecturer in French and Italian, is looking for collaborators to help harness AI to improve students’ pronunciation as they learn a second language. Pronunciation is critical in language acquisition, he said, but limitations in a classroom setting mean that students can’t get the help they need.

“If we have an AI system that can help the students receive feedback directly from that AI agent, it will save a lot of time during class,” Chanethom said. “It also can promote learner autonomy for the student.”

Existing technology can identify that a pronunciation error has occurred but, as far as he knows, not correct it, he said. 

“As a learner of a language, I want to know where the error lies, and how do I improve that?” Chanethom said. “That’s where AI is falling short.”

The symposium’s third presentation category was lightning talks from faculty who received funding from the AI Lab’s Seed Grant Program. The program, which recently announced a new round of funding, provides funds to projects that seek to leverage artificial intelligence to enhance interdisciplinary research.

One such project is nearing the beta testing stage for an AI-based tool that would help scholars at Princeton whose research is based on deciphering handwritten historical texts.

“At Princeton, we have been observing a really increased uptick of interest across humanities departments and social sciences departments in handwritten text recognition technology,” said Christine Roughan, postdoctoral research associate in Near Eastern studies and collaborator on the project. “Our project has been building a research infrastructure here at Princeton to serve the university community, whether you are a student, faculty or staff.”

The event concluded with a reception, where attendees continued discussions of the possibilities of AI and potential research collaborations. 

“We’re thrilled with the success of our first symposium,” said AI Lab Associate Director Olga Russakovsky, who is also an associate professor of computer science. “We are grateful to all the faculty who participated, and look forward to hearing more about potential collaborations that develop.”