Huberman LabCuring All Human Diseases & the Future of Health & Technology | Mark Zuckerberg & Dr. Priscilla Chan
CHAPTERS
- 0:00 – 10:40
Intro, CZI Origin, And Audacious Mission
Andrew Huberman introduces Mark Zuckerberg and Dr. Priscilla Chan, outlining their backgrounds and the Chan Zuckerberg Initiative’s mission to cure, prevent, or manage all human diseases by the end of the century. They discuss why CZI was founded in 2015 and how their complementary experiences in medicine, education, and engineering shaped the initiative.
- •CZI launched in 2015 with a long-term goal to help cure, prevent, or manage all disease by 2100.
- •Chan brings physician and educator experience; Zuckerberg brings engineering and scaling expertise.
- •The strategy is explicitly disease-agnostic and rooted in basic science rather than chasing single conditions.
- •CZI focuses on funding scientists, building tools, and running collaborative Biohubs.
- 10:40 – 22:50
Why Tools Precede Breakthroughs: Engineering The Future Of Biology
Zuckerberg frames scientific progress as historically driven by new tools (e.g., telescopes, microscopes, vaccines). CZI aspires to be that tooling layer for biology—especially for measurement, observation, and data handling—over the next decade.
- •Large-scale discoveries often follow the invention of a new measurement or observation tool.
- •CZI’s 10-year focus is on improving observation in human biology, especially at the cellular level.
- •Analogy to software engineering: you can’t debug a codebase you can’t instrument; similarly, biology needs better ‘instrumentation.’
- •Investments include microscopy, imaging institutes, and software to manage and visualize biological data.
- 22:50 – 40:00
Cells, Single-Cell Atlases, And The Road To A Virtual Cell
Chan explains why cells are the fundamental unit for understanding disease and how single-cell RNA sequencing reveals how different cells interpret the same DNA. They describe CZI’s work on the Human Cell Atlas, tools like CellxGene, and the vision of building AI-powered virtual cells for in silico experimentation.
- •There are ~37 trillion cells in the body; disease emerges from how specific cell types interpret genetic information.
- •Single-cell RNA-seq shows cell-type-specific mRNA expression and differences between healthy and diseased states.
- •CZI-funded work helped discover previously unknown cell types, e.g., in cystic fibrosis-affected lungs.
- •CellxGene lets scientists input a gene/mutation and see which cell types express it, supporting cross-organ hypotheses.
- •The long-term goal is to build a ‘virtual cell’ trained on vast datasets to simulate perturbations and accelerate discovery.
- 40:00 – 57:00
LLMs, AI Clusters, And Hypothesis Generation In Biology
Zuckerberg details how transformer-based LLMs can be trained not just on language but on biological datasets (e.g., Human Cell Atlas) to imagine possible cell states and interactions. He stresses that current AI is best used to propose hypotheses, which scientists then validate experimentally.
- •Transformers scale much better with data than previous ML architectures and power modern LLMs.
- •Feeding cell atlas data into these models allows prediction of cell states and interactions, akin to language prediction.
- •CZI is building a large nonprofit AI cluster (~1,000 GPUs) dedicated to life sciences.
- •AI hallucinations are reframed as a feature for exploring possible solutions—if paired with human validation.
- •Comparison to AlphaFold illustrates how similar architectures can transform protein folding and cell modeling.
- 57:00 – 1:11:40
Biohubs: Chicago’s Sensor Tissues And New York’s Cellular Endoscopes
Chan and Zuckerberg describe how Biohubs extend beyond the original SF hub to Chicago and New York, forcing multi-institutional, interdisciplinary teams to tackle grand challenges. Chicago focuses on embedded sensors in engineered tissues; New York aims to engineer immune cells as diagnostic and therapeutic agents inside the body.
- •Biohubs must involve at least three institutions to break down siloed university structures.
- •Chicago Biohub (UIUC, UChicago, Northwestern) builds sensor-embedded tissues (e.g., skin, neuromuscular junction) to study inflammation and aging-related failures.
- •New York Biohub designs engineered immune cells as in vivo ‘endoscopes’ to detect pathologies (e.g., plaques, cancers) and eventually intervene.
- •These projects are 10–15-year efforts that merge biology with hardcore engineering (e.g., mini-phase plates for electron microscopy).
- •CZI’s role is to fund and host such projects but allow startups and pharma to later translate tools into therapies.
- 1:11:40 – 1:28:20
Philanthropy, Collaboration, And Rethinking How Science Gets Done
They contrast CZI’s model with the traditional independent-investigator grant system, arguing that big questions now require durable tools and deep collaboration. CZI runs independent institutes but tightly partners with universities and patient groups, especially in rare diseases, to both drive basic science and prime the translational pipeline.
- •Most of CZI’s science budget goes to tools and collaborative structures that individual labs can’t easily build or maintain.
- •Biohub scientists are independent of universities but networked with faculty who contribute part-time to grand projects.
- •Rare disease programs (Rare As One) empower patient communities to build registries and work with researchers and drug developers.
- •Rare diseases are ‘windows’ into normal biology; single-gene disorders often reveal fundamental pathways.
- •CZI intentionally avoids becoming a drug company; it focuses on pre-competitive knowledge and enabling platforms.
- 1:28:20 – 1:47:00
Optimism, Family History, And The Moral Backbone Of CZI
Huberman asks how personal optimism and having children influenced CZI. Chan shares her family’s refugee story—parents and relatives sent out on small boats from Vietnam as teens, guided only by faith in a better future—while Zuckerberg reflects on long time horizons and the urgency children created to ‘get on’ building CZI.
- •Chan’s grandparents sent pairs of children on unsafe boats as refugees, believing in a better future; all survived.
- •This background makes Chan’s optimism feel almost obligatory: her existence depends on radical risk for improvement.
- •Zuckerberg distinguishes his technological optimism from Chan’s person-centered optimism as a physician and educator.
- •They literally finished editing CZI’s launch letter in the hospital delivery room before their first child’s birth.
- •Zuckerberg cites the aphorism: “Optimists tend to be successful and pessimists tend to be right,” arguing productive work requires optimism.
- 1:47:00 – 2:06:00
Social Media, Mental Health, And Safety For Teens
Transitioning to Meta’s consumer platforms, Zuckerberg outlines when social media can be beneficial (connection, community) versus harmful (passive consumption, negative news, bullying). They discuss tools for safety and self-regulation, especially for teens, and the balance between empowerment and paternalism.
- •Connecting with others via social media can support well-being and longevity; passive doom-scrolling generally does not.
- •Meta provides blocking, restricting, comment controls, teen-private defaults, and parental supervision tools.
- •In-product cues nudge teens away from single-content rabbit holes and surface time-use information.
- •Zuckerberg resists a paternalistic approach: Meta aims to give tools and guardrails but respects user preferences where possible.
- •Time-on-platform is hard to evaluate in the abstract; impact depends heavily on activity type (e.g., group chats vs. negative news binges).
- 2:06:00 – 2:29:00
VR, Mixed Reality, And Embodied Computing
Huberman tries the latest Quest mixed-reality demo and they unpack why embedding VR into the physical environment—rather than replacing it entirely—changes the experience. Zuckerberg frames VR/AR as ‘computers for your whole body’ that can transform exercise, training, work, and education.
- •New Quest headsets offer high-quality passthrough so users see the room and people while interacting with virtual content.
- •Mixed reality enables engaging physical workouts (e.g., boxing, dancing) without the disorientation of older VR systems.
- •Zuckerberg sees AR/VR as extending the physical world, not escaping it: future rooms will mix physical objects and holograms.
- •Education examples include surgery training, enlarged anatomical models, and collaborative holographic workspaces.
- •Technical challenges remain around tracking full-body motion, adding resistance, and enabling realistic grappling or advanced martial arts.
- 2:29:00 – 2:54:00
Ray-Ban Meta Glasses And Ambient AI Assistance
They explore Meta’s Ray-Ban smart glasses, which look like normal eyewear but include camera, audio, and an AI assistant. Huberman is struck by being able to control music and get responses without leaving a conversation, and Zuckerberg explains how this foreshadows full AR glasses with holographic displays.
- •Current Ray-Ban Meta glasses offer photos/video capture, open-ear audio, calls, podcasts, live-streaming, and voice-controlled Meta AI.
- •A bright recording light signals when cameras are active, making covert recording harder than with phones.
- •The long-term vision is supercomputer-level capability in a glasses form factor, with holographic displays overlaying directions, media, and interfaces.
- •Glasses allow AI to ‘see what you see, hear what you hear’ to provide highly contextual assistance, while keeping users present in the physical world.
- •Eye tracking and external sensors create trade-offs in size, battery, privacy, and cost; different models will target different needs and price points.
- 2:54:00 – 3:14:00
AI Personas, Creators, And The Future Of Digital Identity
Zuckerberg discusses AI personas in social apps—both fictional characters (like a Snoop Dogg ‘dungeon master’) and future creator-linked assistants. He emphasizes that AI versions of real people must be controlled by those people and reliable enough in how they represent them.
- •Current Meta AI personas are fictional roles (chef, trainer, travel expert), sometimes voiced/played by celebrities but not impersonating them.
- •These personas live in chats and threads, offering specialized help and entertainment (e.g., interactive storytelling with kids).
- •A next step is creator- and business-specific AIs that can answer community questions at scale in the creator’s style and knowledge domain.
- •Strict impersonation policies mean only the real person should control any AI that bears their identity; platforms must avoid unauthorized clones.
- •LLMs need additional work on controllability so creators can trust AI assistants not to misrepresent them.
- 3:14:00
Closing Thoughts: AI, Health, And An Optimistic Future
The conversation ends with reflections on AI as a neutral force whose impact depends on use and guardrails. Huberman underscores how Meta’s tools and CZI’s infrastructure can substantially improve health and scientific progress; Zuckerberg reiterates a commitment to broad accessibility rather than premium-only tech.
- •AI can exacerbate harms or massively enhance well-being depending on design and governance.
- •Meta aims to build tools (VR, AR, AI, social) that support physical activity, education, and healthier social connection.
- •CZI’s science work and Meta’s product work operate independently but are united by a long-term, optimistic view of human potential.
- •Meta’s business strategy favors large-scale, affordable platforms over high-margin niche hardware.
- •All three participants converge on an optimistic outlook that science and technology can make the future materially better for human health.
