SCHEDULE

Sessions | Absolutely Interdisciplinary

Absolutely Interdisciplinary will take place in person at the Schwartz Reisman Innovation Campus at the University of Toronto.

Morning keynote

May 29, 2025 | 11:20 AM – 12:30 PM

Speaker: Sara Hooker (Virtual)

Moderator: Roger Grosse

Sara Hooker is a leading expert in machine learning (ML) research who specializes in developing more efficient, safe, and grounded AI models, with a strong focus on fundamental research, trustworthy AI, and real-world applications. With experience spanning industry, open research initiatives, and AI ethics, Hooker brings a deep understanding of the challenges and opportunities shaping the field and will deliver a keynote lecture on her recent work. Hooker leads Cohere For AI, a non-profit research lab that seeks to solve complex ML problems and create more points of entry into ML research.

New frontiers in AI governance

May 29, 2025 | 9:30 AM – 11:00 AM

Speakers: Atoosa Kaziradeh, Nitarshan Rajkumar

Moderator: Karina Vold

Governing the most advanced ‘frontier’ AI systems presents distinct challenges that go beyond general AI governance. This session will explore key governance considerations for highly capable AI systems, offering insights into the sociotechnical complexities of AI risk governance and emerging approaches to regulation. Kaziradeh and Rajkumar will examine the political, social, and technical dimensions of frontier AI governance and discuss trade-offs in different strategies that balance innovation and risk mitigation.

Navigating autonomy and accountability in AI agents

May 29, 2025 | 1:30 PM – 3:00 PM

Speakers: Megan Ma (Virtual), Atrisha Sarkar

Moderator: Anna Su

Autonomous agents are shaping the future of AI development, raising legal and societal questions about accountability, governance, and trust. Who is responsible when an AI agent makes a harmful decision? How should laws evolve to regulate AI autonomy? This session will bring together experts in law, ethics, technology, and policy to explore the challenges of governing AI agents, from liability and bias to power dynamics and regulatory gaps. Using real-world scenarios and examples, Ma and Sarkar will examine how to balance innovation with accountability in an era of increasingly independent AI systems.

Afternoon Keynote - Gradual disempowerment: Systemic existential risks from incremental AI development

May 29, 2025 | 3:20 PM – 4:30 PM

Speaker: David Duvenaud

Moderator: Sheila McIlraith

This talk will explore the systemic risks posed by incremental advancements in artificial intelligence, introducing the concept of gradual disempowerment as an alternative to the abrupt takeover scenarios often discussed in AI safety. Drawing on recent research, Duvenaud will examine how even incremental improvements in AI capabilities can erode human influence over critical societal systems, including the economy, culture, and governance.

As AI increasingly supplants human labor and cognition, it may weaken both explicit control mechanisms—such as voting and market choices—and the implicit human alignments embedded in societal structures that historically relied on human participation. Furthermore, if these systems incentivize outcomes misaligned with human values, AI-driven optimization could exacerbate those misalignments. These effects may reinforce one another across domains, as economic power influences cultural narratives and political decision-making, while cultural shifts, in turn, shape economic and political behavior.

This session will explore whether such dynamics could lead to an effectively irreversible decline in human agency over key societal functions, heightening the risk of existential catastrophe through long-term disempowerment. The discussion will highlight the need for both technical research and governance strategies to counteract the incremental erosion of human influence and ensure that AI-driven systems remain aligned with human interests.