Silicon Valley GirlHow AI Is Breaking the Rules of Biology | Dr. Priscilla Chan, Chan Zuckerberg Initiative
CHAPTERS
AI could rewrite biology within five years: the big promise
The episode opens with a bold claim: science—and especially biology—may look fundamentally different within the next five years. Marina frames the conversation around what that shift would mean for real patients, setting up a focus on applied impact rather than abstract tech.
- •Priscilla predicts a near-term step change in how biology is practiced
- •The interview is framed around patient outcomes, not just research progress
- •AI + biology convergence is positioned as the core driver of acceleration
Miscarriage, loneliness, and why public honesty matters
Marina shares how Priscilla’s public openness about miscarriage helped her through her own multi-year journey to having children. Priscilla reflects on how common miscarriage is, how under-discussed it remains, and how isolating the experience can feel.
- •Marina credits Priscilla’s disclosure as emotional support during infertility/miscarriage
- •Miscarriage is common but often not openly discussed
- •Doctors’ messaging can compound guilt and isolation
- •Shared stories reduce stigma and make people feel less alone
Why Chan and Zuckerberg launched CZI after their first child
Priscilla explains how becoming a parent made the future feel immediate rather than abstract. That urgency helped drive the decision to commit major resources to building a healthier world for the next generation via the Chan Zuckerberg Initiative.
- •Parenthood reframed long-term societal risk into a personal, immediate responsibility
- •CZI is described as “doing our part” to build a healthier future for kids
- •The origin story is grounded in family and generational time horizons
CZI/Biohub mission: cure or prevent all disease—and make every scientist faster
Priscilla lays out the mission: curing or preventing all disease, sooner than earlier timelines suggested. The strategy is not to do it alone, but to increase the speed, efficiency, and risk-taking capacity of the entire scientific ecosystem by building shared tools.
- •Mission: cure or prevent all disease (timeline accelerating due to AI)
- •Key lever is enabling scientists to work faster, better, and with more boldness
- •Tool-building and ecosystem enablement over single-organization heroics
- •Biohubs combine frontier biology with frontier AI to drive impact
From “you’re nuts” to “maybe possible”: reactions then vs. now
Priscilla contrasts early skepticism with today’s more pragmatic optimism, catalyzed by large language models and better datasets. She describes how criticism forced clarity: the blockers were tools, data, and interdisciplinary collaboration—so CZI built them.
- •Initial public reaction: curing all disease sounded impossible
- •Skepticism helped identify concrete bottlenecks (tools, datasets, lab techniques)
- •Biohubs expanded and interdisciplinary teams became central
- •LLMs created a new pathway to extract meaning from complex biological data
The clinic turning point: pediatrics exposed the limits of medicine
Priscilla recounts her experience at UCSF caring for children with conditions lacking names, mechanisms, or treatments. Reading sparse research PDFs highlighted the gap between basic science and actionable care—and convinced her that advancing foundational biology is where hope originates.
- •Many pediatric cases lacked clear diagnoses and treatments
- •The clinic-to-lab translation gap was stark and emotionally intense
- •Basic science progress is positioned as the upstream source of medical breakthroughs
- •This experience motivated her shift toward enabling biology research
Why CZI builds platforms, not disease-specific moonshots
Rather than picking a single disease, CZI prioritizes tools and shared resources that make many disease solutions possible. Priscilla explains the core biological question: how identical DNA yields different cell types, and how errors propagate into disease—best studied directly in human-relevant models.
- •Strategy: empower all scientists instead of targeting one disease at a time
- •Single-cell “cell by gene” mapping to understand cell identity and malfunction
- •Goal: understand healthy vs. diseased cellular states to design precise interventions
- •Virtual human-relevant experimentation could outperform mouse/fly translation limits
Virtual cells explained: what they are and how close we are
The conversation turns to “virtual cells”—computational models that simulate how real cells behave. Priscilla notes differing timelines: AI optimists think a few years, biologists see more complexity, but she expects a meaningful shift in scientific modeling within ~5 years.
- •Virtual cells = computer models that replicate cellular behavior and responses
- •Timelines vary by discipline; biology has many dimensions to capture
- •Expectation: science workflows change substantially within five years
- •Virtual models promise cheaper, faster, more iterative experimentation
What this changes for patients: personalized medicine and better first-try treatments
Priscilla connects virtual-cell capability to care delivery: moving from population averages to individual biology. She argues many “common diseases” are really bundles of subtypes, and that granular modeling could help choose (or design) the right therapy sooner—reducing trial-and-error suffering.
- •Medicine today relies heavily on averages; individuals vary meaningfully
- •Personal genetics and cellular behavior could guide therapy selection
- •Common conditions (hypertension, depression) likely consist of multiple sub-diseases
- •Reduced guesswork could prevent suffering and speed effective treatment
Early wins: immune system leverage, autoimmunity, and engineered immune cells
Asked what diseases could be first to benefit, Priscilla focuses on the immune system as both protector and potential cause of disease when dysregulated. She highlights autoimmune diseases and the possibility of directing or enhancing immune cells to detect and repair issues like arterial plaques.
- •Immune balance is delicate; understanding its “levers” could treat autoimmunity
- •Immune cells already travel throughout the body—useful for monitoring/repair
- •Vision: engineer immune cells to detect plaques and intervene
- •Applications go beyond infections to whole-organ health (e.g., neurodegeneration, MS)
Scaling the cell atlas: from 100 million to 1 billion cells—and why it’s still not enough
Priscilla describes dramatic acceleration in cell mapping: a decade to reach ~100M cells, then months to reach 1B. But she emphasizes transcriptomics is only one dimension—proteins, spatial layout, living dynamics, and context-dependent behavior are still needed for accurate models.
- •Mapping throughput accelerated massively due to improved hardware and focus
- •Single-cell transcriptomics is only one slice of biology
- •Additional layers needed: proteomics, spatial imaging, living-cell dynamics, context
- •The modeling goal requires multi-modal, high-quality, richly annotated data
The AI–wet lab flywheel: co-designing experiments and models
CZI pairs AI labs with wet labs to avoid silos: models reveal blind spots that guide new experiments, while lab constraints inspire AI tools to remove bottlenecks. Priscilla describes this as a virtuous cycle that speeds discovery and improves data quality and interpretability.
- •AI teams identify model gaps; wet labs generate targeted new measurements
- •Wet-lab bottlenecks (e.g., analyzing Cryo-ET tomograms) become AI tool opportunities
- •Metadata and experimental context are treated as critical for modeling
- •The core thesis: frontier AI + frontier biology creates compounding acceleration
Virtual immune system and immune “communication sensors”
Priscilla introduces the “virtual immune system” as a higher-order model capturing how diverse immune cells coordinate via signals across the body. She highlights a tiny wearable-like sensor developed at the Chicago Biohub that reads immune-cell communication in living systems—enabling dynamic modeling and intervention testing.
- •Virtual immune system models multi-cell coordination and signaling, not just single cells
- •Immune cells communicate across distances; understanding on/off switching is key
- •Chicago Biohub sensor measures immune signaling in vivo, akin to a CGM concept
- •Sensor data can feed models to simulate and manipulate immune parameters
A 2040 scenario: early-warning patches, flare prediction, and custom drugs
In a speculative but grounded future, Priscilla imagines wearable patches detecting early molecular signals of diseases like lupus before symptoms appear. Virtual-cell models could then help identify the misbehaving pathway and support designing tailored interventions—shifting healthcare from reactive to preventive.
- •Wearables could detect the earliest molecular signatures of disease flares
- •Example: lupus risk + rising trigger molecule → early intervention before organ damage
- •Virtual models could pinpoint faulty proteins/pathways in an individual
- •Preventive monitoring + custom therapeutics = healthier trajectories
What worries her, advice to future clinicians/scientists, and the new physician role with AI
Priscilla says speed and removing barriers are what keep her up at night—her role is to enable scientists to move efficiently. She encourages aspiring scientists/doctors to enter the field now, and describes physicians as crucial translators: asking the right questions of AI while remaining human “healers” alongside patients.
- •Top concern: moving quickly by eliminating barriers to research progress
- •Career advice: it’s an unusually exciting moment to enter biology/medicine
- •Physician-scientists are “magic” because they link patient reality to research questions
- •AI will outperform humans on some tasks; physicians must guide, contextualize, and care
When it becomes personal: reproductive biology gaps, rare diseases, and what “success” means
Priscilla shares personal motivation points: studying reproductive biology (including the still-mysterious trigger of labor) and supporting rare-disease communities through “Rare As One.” She argues many diseases can be cured in our lifetime as molecular understanding improves, but she doesn’t define a single finish line—there’s always more work.
- •Reproductive biology remains under-understood; labor initiation is still not fully known
- •Rare As One supports patient groups to participate meaningfully in research
- •Even “just” getting a diagnosis can reduce isolation and powerlessness
- •Cures are likelier where genetic/molecular mechanisms are clear; the goal is expanding that clarity
- •Mission fulfillment isn’t a single milestone; impact and continued progress drive her
Balancing leadership and motherhood: strict boundaries
Closing on a practical note, Priscilla describes how she manages parenting alongside demanding work through disciplined scheduling. She keeps dedicated blocks for children and for work, accepting trade-offs in social time until later life stages.
- •Strict time-blocking: dedicated kids time and dedicated work time
- •Avoids mixing roles to stay present in each domain
- •Accepts reduced social/leisure time as a temporary trade-off
