Jennifer Chase on Graphons, Phase Transitions and the AI Revolution

Overview

In this episode, Flora Lichtman chats with Jennifer Chase about the mathematics of networks, the concept of graphons, phase transitions in information systems, and how these ideas underpin modern AI. Chase shares her journey from early mathematical intuition to building interdisciplinary labs at Microsoft and now leading UC Berkeley’s College of Computing, Data Science and Society.

Key insights

• Graphons and phase transitions describe how large networks behave, offering a forest view of complex systems instead of tracking every node.
• Recent shifts in large language models illustrate information network phase transitions, where knowledge becomes accessible across remote areas of the network.
• Interdisciplinary labs, academic freedom, and collaboration across math, CS, and social sciences are essential to advancing AI responsibly and with societal impact.
• Chase emphasizes risk taking, mentorship, and making AI tools accessible while inoculating against harms through education and governance.

Introduction and guest background

The podcast features Flora Lichtman in conversation with Jennifer Chase, a pioneering figure in the math of networks and an influential leader in AI research ecosystems. Chase discusses how esoteric mathematical theories such as phase transitions, graph limits and graphons inform the behavior of massive networks that underlie social platforms, scientific discovery, and the frontier of AI. Her career spans twenty years building interdisciplinary Microsoft research labs and now guiding UC Berkeley as the dean of the College of Computing, Data Science and Society. The discussion covers her early life and the way math became her mother tongue, her views on the AI revolution, and the role of education in shaping a safer, more equitable AI future.

Big ideas: graphons, phase transitions and networks

A central theme in the conversation is the idea that networks can be understood through their limits. Chase explains that graphons provide a way to describe the limiting behavior of large networks when you cannot or should not follow every pairwise interaction. This leads to characterizations of global properties, such as how information propagates through a social network or how communities cohere as connections become denser or sparser. She uses the concept of phase transitions to illustrate qualitative changes in network behavior as parameters cross critical thresholds. The comparison to water boiling or magnets aligning helps lay readers grasp the idea that large networks can suddenly acquire order or exhibit new capabilities as their connectivity changes.

The discussion then moves to information networks in AI, notably the knowledge embedded in frontier models. Chase argues that a similar type of phase transition can occur in knowledge networks when a critical level of connectivity and training data enables information to spread across previously disconnected regions of a model. This conceptual framing complements more traditional computational views and helps explain why large language models exhibit emergent capabilities beyond local, near-term correlations.

Graphons as a practical lens for complex systems

Chase describes graphons as a mathematically meaningful way to take a limit of networks with many nodes. By moving to a limiting object, researchers can discuss “forest properties” without needing to map every tree. This perspective helps researchers compare networks that are too large to inspect in full, whether they are social networks, biological interaction graphs, or economic systems. She notes that several meaningful limit concepts can converge to the same graphon, underscoring a robustness that is valuable for theory and applications alike. The existence of a convergent, unified object enables better predictions of collective behavior in diverse systems, including how learning algorithms scale and how information propagates in real-world networks.

Interdisciplinary labs and the AI revolution

A substantial portion of the podcast centers on Chase’s experience co-founding interdisciplinary research labs at Microsoft and her current tenure at UC Berkeley. The conversation explores the practicalities and benefits of bringing together mathematicians, computer scientists, social scientists, and domain experts to tackle big problems. Chase emphasizes the importance of scientific freedom in enabling bold, high-risk projects, and she reflects on how interdisciplinary collaboration catalyzes breakthroughs that single-discipline work cannot achieve. She also discusses how modern AI can be both a great equalizer, through personalized tutoring and accessible tools, and a source of risk if not guided by rigorous safety and governance frameworks.

AI safety, governance, and the future of work

Addressing the dual impulse of excitement and concern that surrounds generative AI, Chase outlines practical steps for building safer AI systems. She describes a proactive stance toward verification, demanding references and traceable sources from AI agents. She also highlights the role of education in broadening access to AI literacy and preparing a workforce that can work effectively in AI-enabled teams. With UC Berkeley, she envisions scalable training for two million California community college students, aiming to democratize AI fluency and enable people to collaborate with AI as partners rather than bystanders. The MOF and materials science example illustrates how AI can accelerate discovery when paired with experiments, guiding next-step decisions and optimizing research loops.

Personal journey and life philosophy

Chase’s story—from dropping out of high school in New York City to becoming an influential university leader—serves as a through line about risk-taking, resilience, and the value of embracing novice status. The host and guest discuss overcoming barriers for women and minoritized groups in STEM and the importance of mentorship and inclusive practices. The conversation ends with Chase’s life advice: seize opportunities even when they arise at inopportune moments, as these moments often define future paths.

Conclusion and takeaway

The podcast offers a portrait of a scientist who blends deep mathematical theory with practical, real-world applications and institutional leadership. Chase argues for a future in which AI is deployed thoughtfully, safely, and accessibly, with strong governance and education to close gaps rather than widen them. The episode invites listeners to see AI as a collaboration between human and machine intelligence that can accelerate science and improve lives if guided by curiosity, rigor, and inclusive practices.