Thematic Semester - Autonomous LLM Agents: Risks and Scientific Challenges

At a time when large language models are already transforming our daily lives, a new revolution is brewing in research labs and in the industrial world: autonomous agents. IVADO, a centre of excellence in AI, is organizing a thematic semester dedicated to this emerging field that raises as many hopes as it does questions.

From August to December 2025, the Université de Montréal campus will become the global epicentre for reflection on these artificial entities capable of representing us and acting independently in the world. This major evolution has implications that extend far beyond the technological framework given their potential impacts on society.

While large language models have already demonstrated impressive capabilities in generating text or code, the transition to autonomy represents a considerable qualitative leap. Emerging autonomous agents—such as those being developed by tools like Operator and Claude—can already perform tasks like navigating the web to purchase items, posting content on social media, executing complex workflows by interacting with software interfaces, or calling tools like calendars, email, or banking apps. These examples hint at the profound potential of agents to operate independently across digital and physical environments. With numerous technical challenges still to be mastered, it is essential to anticipate and address the technical, ethical, economic and societal issues raised by these technologies.

How can we create agents capable of operating in complex environments without constant human supervision?
How can we guarantee their reliability and safety when they interact with the real world?

This thematic semester will bring together researchers from across disciplines - natural language processing, computer vision, robotics, ethics - to collaborate closely in the face of this major scientific challenge.

KEY DATES

August 11, 2025: Opening Event
August 12-15, 2025: Bootcamp - Current State of Agents (Siva Reddy, Alane Suhr, Victor Zhong)
October 3-6, 2025: 1st Workshop - Evaluating Capabilities and Research Challenges for Agents (Yu Su, Alane Suhr, Victor Zhong, Chris Pal)
November 17-21, 2025: 2nd Workshop - Safety Evaluation, Risk Mitigation, and Creating Safe Agents (Siva Reddy, Dan Song, Yoshua Bengio)

Organizers

Siva Reddy

IVADO, McGill, Mila

Siva Reddy is an Assistant Professor in the School of Computer Science and Linguistics at McGill University, a Canada CIFAR AI Chair, a core faculty member of Mila Quebec AI Institute, and a research scientist at ServiceNow Research. He co-leads the McGill NLP Group. Previously, he was a postdoctoral researcher at Stanford University and a Google PhD fellow at the University of Edinburgh. His research interests are in representation learning for language with a specific focus on reasoning, grounding, conversational modeling, and safety.
Dawn Song

UC Berkeley

Dawn Song is a Professor in Computer Science at UC Berkeley and Co-Director of Berkeley Center for Responsible Decentralized Intelligence. Her research interest lies in AI and deep learning, security and privacy, and decentralization technology. She is the recipient of various awards including the MacArthur Fellowship, the Guggenheim Fellowship, the NSF CAREER Award, the Alfred P. Sloan Research Fellowship, the MIT Technology Review TR-35 Award, ACM SIGSAC Outstanding Innovation Award, and more than 10 Test-of-Time Awards and Best Paper Awards from top conferences in Computer Security and Deep Learning. She has been recognized as Most Influential Scholar (AMiner Award), for being the most cited scholar in computer security. She is an ACM Fellow and an IEEE Fellow. She obtained her Ph.D. degree from UC Berkeley. She is also a serial entrepreneur and has been named on the Female Founder 100 List by Inc. and Wired25 List of Innovators.
Yoshua Bengio

IVADO, Université de Montréal, Mila

Yoshua Bengiois a full professor in the Department of Computer Science and Operations Research at Université de Montréal, as well as the Special Advisor and Founding Scientific Director of IVADO and Scientific Advisor and Founder of Mila . He holds a Canada CIFAR AI chair and is the recipient of the 2018 A.M. Turing Award, considered the “Nobel Prize of computing.”

He is a fellow of both the U.K.’s Royal Society and the Royal Society of Canada, an officer of the Order of Canada, a knight of the Legion of Honor of France, and a member of the U.N.’s Scientific Advisory Board for Independent Advice on Breakthroughs in Science and Technology.
Alane Surh

UC Berkeley

Alane Suhr is an Assistant Professor of Computer Science at UC Berkeley, affiliated with the Berkeley AI Research (BAIR) Lab. Alane’s research focuses on language use and learning in interaction, especially situated and collaborative interactions; agents that use language and language-conditioned agentic systems; and language grounding to perception and action.
Christopher Pal

IVADO, Polytechnique Montréal, Université de Montréal, Mila

Chris Pal is a world renowned expert in deep learning (DL), probabilistic machine learning and applied AI. He received a Google focused faculty Research Award twice (in 2017 and in 2009). He brings a unique multidisciplinary expertise to the team with his strong theoretical background (DL, graphical models, causality, etc.) as well as his key expertise in Natural Language Processing and Computer Vision applied to a wide range of applications domain (medical imaging in health care, satellite imaging for biodiversity and climate change, autonomous driving related to transportation).
Yu Su

Ohio State University

Yu Su is a Distinguished Assistant Professor of Computer Science and Engineering at The Ohio State University, where he co-directs the NLP group. He has broad interests in artificial intelligence, with a primary interest in the role of language as a vehicle for reasoning and communication. His recent work covers many facets of AI agents including memory, planning, grounding, and safety. He is a 2025 Sloan Research Fellow and has received multiple paper awards from CVPR and ACL.
Victor Zhong

University of Waterloo, Vector Institute

Victor Zhong is an Assistant Professor in the David R. Cheriton School of Computer Science at the University of Waterloo, and a Canada CIFAR AI Chair at the Vector Institute. His research is on efficient machine learning methods that interpret language to generalize to new problems. His group, the Reading to Learn Lab, works on grounded language agents, test-time adaptation, and learning from language feedback. Victor received a PhD in Computer Science from the University of Washington, a Master’s in Computer Science from Stanford University, and a Bachelor’s in Computer Engineering from the University of Toronto.