From Robotics to Human Interaction: Embodied AI Takes Center Stage at Stony Brook Workshop

Researchers from leading institutions across the country gathered at Stony Brook University for an invite-only Embodied AI Workshop, a day-long event exploring how artificial intelligence is beginning to move beyond purely digital systems and into the physical world.

Organized by I.V. Ramakrishnan, C.R. Ramakrishnan and Nilanjan Chakraborty, the workshop on February 27 brought together experts from multiple universities and research labs, including MIT, Columbia University, Duke University and Brookhaven National Laboratory. Over the course of the day, speakers explored what it means for AI systems to perceive, reason and act in real-world environments - from robotics and physical design to communication and creativity.

"Embodied AI represents an exciting frontier for the field," said Samir Das, professor and chair of the Department of Computer Science. "Bringing together researchers from different institutions and disciplines helps us explore how intelligent systems will interact with the physical world and with people in increasingly meaningful ways."

Several talks showed how embodied AI is already influencing both physical systems and creative work. Lav Varshney, professor of computer science and director of Stony Brook's AI Innovation Institute (AI3), described how AI-driven methods can help design more sustainable building materials, including new concrete formulations that reduce carbon emissions while improving strength. Because concrete production accounts for a significant share of global emissions, advances like these could have a meaningful environmental impact. Varshney also discussed creative applications of embodied AI, such as AI-assisted choreography, highlighting how intelligent systems might generate movements and designs that interact directly with human environments.

The workshop also examined challenges that arise when AI systems interact with people. In one presentation, Owen Rambow, professor of linguistics and computer science at Stony Brook, discussed how today's large language models still struggle with what researchers call "pragmatic competence" - the ability to track shared context and mutual understanding during conversations. Humans do this naturally when collaborating or referring to objects in shared tasks, but building AI systems with similar capabilities remains an open challenge.

"Our goal was to create a space where researchers could exchange ideas across areas such as robotics, machine learning and human-AI interaction," said Ramakrishnan. "Embodied AI raises fundamental questions about how intelligence operates in the real world."

By bringing together researchers from multiple disciplines, the Embodied AI Workshop offered a snapshot of a rapidly evolving field. As artificial intelligence continues to move beyond screens and into the physical world, conversations like these are helping shape the next generation of research in embodied intelligence.