Overview of Science Friday with Jennifer Chayes
This Science Friday conversation with mathematician and UC Berkeley dean Jennifer Chayes explores how she uses math as a way of “making sense” of the world, how her work helped shape the theory behind modern networks and machine learning, and why she sees generative AI as both deeply promising and potentially risky. Chayes also shares pieces of her unconventional life story—dropping out of high school, living in New York City as a teen, returning to school, and eventually building interdisciplinary research labs at Microsoft and now leading Berkeley’s College of Computing, Data Science, and Society.
Key Themes and Takeaways
Math as a “native tongue”
- Chayes describes math not as abstraction, but as her first language for understanding reality.
- Even as a child, she says she instinctively tried to model how things worked, asking questions and reasoning through patterns before she was formally taught to read.
- For her, mathematics is less about symbols and more about sense-making.
Networks, phase transitions, and graphons
- A major thread in her work is the study of networks: systems where entities interact in pairs, such as social networks, biological systems, or information networks.
- She explains phase transitions as sudden qualitative changes in behavior when a system crosses a threshold—like water boiling or a virus spreading through a population.
- Chayes connects this to modern AI, suggesting that large language models may exhibit a kind of phase transition when they move from local pattern completion to more global, surprising capabilities.
- She also explains graphons, the mathematical framework she co-developed to understand the limits and large-scale behavior of massive networks without tracking every individual connection.
Why graphons mattered
- Before graphons, researchers often had no elegant way to describe the behavior of huge networks as a whole.
- Graphons helped formalize the idea that you can study collective properties without following every node individually—similar to how thermodynamics describes gases without tracking each molecule.
- Chayes is especially gratified that younger researchers have found new uses for graphons, including interpreting video.
Chayes’ Personal Story
A stormy childhood, but math as stability
- Chayes shares that she had a difficult adolescence and dropped out of high school.
- She spent time living on the streets of New York City in the early 1970s, while also returning to school through a dropout program.
- During that period, math remained her “happy place” and a source of calm.
Teaching while still learning
- At the dropout school, she ended up teaching math because no one else was available.
- She credits that period with teaching her resilience, comfort with risk, and a willingness to fail and try again.
Learning to embrace being a novice
- Chayes says one of her strengths is that she is not afraid to enter new fields.
- She moved across disciplines—biology, physics, math, computer science, economics, chemistry—because she finds being a beginner exhilarating rather than intimidating.
Interdisciplinary Work and Building Institutions
Why she left academia for Microsoft
- In the 1980s and 1990s, interdisciplinary research was often difficult to sustain inside traditional universities.
- Chayes joined Microsoft after asking for an unusually ambitious setup: complete scientific freedom and the ability to build interdisciplinary labs.
- Microsoft said yes, and she spent 20 years building research environments that brought together mathematicians, computer scientists, and social scientists.
Lessons from institutional building
- She argues that important discoveries increasingly happen at the boundaries between fields.
- She also notes that the old model of forcing people to choose one discipline too early can limit innovation.
AI: Promise, Risk, and Scientific Discovery
Excitement about generative AI
- Chayes is enthusiastic about generative AI’s potential in:
- biomedicine
- sustainability
- scientific discovery
- education
- She sees AI as a possible great equalizer, especially through personalized tutoring and access to expertise.
Concerns about AI
- She also worries about:
- uneven distribution of benefits
- misinformation
- safety problems
- AI deepening existing inequalities if not developed responsibly
AI for science in practice
- She describes a collaboration with chemist Omar Yaghi on metal-organic frameworks (MOFs), extremely porous materials with enormous surface area and potential uses in capturing molecules like carbon dioxide.
- With AI, they reduced the time needed to synthesize candidate materials from years to weeks.
- Her view: the best use of AI in science is to connect it tightly with experiments, using each result—especially failed ones—to guide the next step.
Advice and Mentorship
Teach people to verify AI
- Chayes says one way to protect against AI harms is to teach people to use it critically.
- Her own habit: always ask chatbots for references, verification, and ways to check answers.
Build new AI-ready institutions
- She believes community colleges, universities, and new academic structures should teach people not just the basics, but how to work alongside AI in hybrid teams.
- She sees this as a source of many future jobs at the intersection of disciplines.
Life advice: “Grab the brass ring”
- Her final advice is to take opportunities when they appear, even at inconvenient times.
- She emphasizes that major opportunities often arrive when life is messy, and the key is to seize them anyway.
Notable Ideas and Quotes
- “Math is a mother tongue.”
- “You can’t see the forest for the trees.” — her explanation of why network limits matter
- “Grab the brass ring.” — her final life advice
- AI should be used as a partner, not a black box to trust blindly
Overall Message
The episode presents Jennifer Chayes as both a pioneering mathematician and a builder of institutions. Her story connects abstract mathematics to everyday life, shows how interdisciplinary work can transform science, and offers a balanced view of AI: powerful enough to reshape discovery and education, but only beneficial if people learn to use it wisely and critically.
