Mark Zuckerberg & Priscilla Chan: How AI Will Cure All Disease

This episode reveals how AI is being systematically applied to build foundational, open-source models of human biology, creating a new computational layer for drug discovery and precision medicine.

The Audacious Mission to Cure All Disease

The conversation begins with the origin of the Chan Zuckerberg Initiative's (CZI) mission: to cure, prevent, or manage all diseases by the end of the century. Priscilla Chan, drawing from her experience as a pediatrician, explains that the goal was born from the frustration of treating children with conditions that lacked clear scientific understanding. She describes the "pipeline of hope" that basic science provides, which is often blocked by a lack of fundamental knowledge.

Mark Zuckerberg frames the strategy not as CZI curing diseases itself, but as building the tools to accelerate the entire scientific community. He notes that most major scientific breakthroughs, like the discovery of bacteria after the invention of the microscope, are preceded by new tools for observation. CZI aims to fill a specific niche that traditional government funding often misses: developing long-term, expensive (hundred-million to billion-dollar) foundational tools.

A Strategy of Building Foundational Tools

When CZI first announced its goal, the reaction from the scientific community was disbelief. Priscilla Chan recalls, "Honestly most scientists couldn't look at us with a straight face." When pressed on why the goal seemed impossible, scientists pointed to a lack of shared tools and large-scale, standardized datasets. This feedback directly shaped CZI's strategy to focus on building that missing infrastructure.

Interestingly, the AI community had the opposite reaction, viewing the goal as an inevitable outcome of technological progress. Mark Zuckerberg highlights this gap, noting that CZI's work sits at the intersection of these two worlds, pairing "frontier biology with frontier AI." The goal is to move beyond using AI on pre-existing public data (like the decades-old dataset used for AlphaFold) and instead generate novel, high-quality datasets specifically for training advanced biological AI models.

The Biohub Grand Challenges: A 15-Year Horizon

CZI structures its research around 10-15 year "grand challenges" hosted at its Biohubs in San Francisco, Chicago, and New York. This timeframe is ambitious enough to tackle significant problems but concrete enough to provide a clear path forward. Each Biohub has a distinct focus designed to generate unique data:

• New York: Cellular engineering to create cells that can act as biological sensors.
• Chicago: Studying tissue-level cell communication to understand inflammation.
• San Francisco: Deep imaging and single-cell transcriptomics—a technique for analyzing gene expression in individual cells—to map cellular states.

The recent arrival of large language models (LLMs) has been a catalyst, providing the computational framework to finally make sense of the massive datasets these Biohubs were designed to create.

Redefining Precision Medicine for All

Priscilla Chan outlines a future where most diseases are treated as rare diseases, acknowledging that individual biology is unique. Currently, treatments for common conditions like hypertension or depression rely on trial and error. The goal is to enable truly precise medicine by understanding how a genetic variant impacts downstream cellular functions and protein expression.

By building models that can connect a mutation to its biological consequences, researchers can identify highly specific drug targets and predict off-target effects. CZI's role is to build the open-source models and datasets that empower startups and pharmaceutical companies to develop these next-generation diagnostics and therapeutics.

Cell by Gene: The Accidental Network Effect

One of CZI's most impactful projects, the Cell Atlas, was inspired by the fact that biology lacks a foundational map akin to the periodic table of elements. The project's key tool, Cell by Gene, was initially built simply to help researchers annotate single-cell data faster, as this was a major bottleneck.

Because the tool was effective and open, it became a standard. This created a powerful network effect: researchers began using the same format and contributing their data back to the platform. Today, the atlas contains millions of cells, with 75% of the data contributed by the broader scientific community—a testament to the power of building valuable, open infrastructure.

The Next Frontier: Building Virtual Cells

The conversation pivots to one of CZI's most ambitious goals: building virtual cell models. These are complex computational simulations intended to become a foundational tool for biological research, much like a new type of model organism. The purpose is to allow scientists to generate and test hypotheses in silico (computationally) before undertaking expensive and slow wet lab experiments.

Actionable Insight: The development of virtual cells represents a paradigm shift toward AI-driven biological experimentation. Investors should track startups leveraging these emerging open-source models to de-risk and accelerate drug discovery pipelines.

Mark Zuckerberg details several models that form the building blocks of this effort:

• `VariantFormer`: Predicts the cellular outcome of a specific genetic edit using CRISPR.
• Diffusion Models: Generate synthetic models of rare or specific cell types for testing.
• Cryo-EM Models: Provide spatial understanding of cellular structures at a near-atomic level.
• Reasoning Models: An early-stage but critical effort to build models that can reason about biological cause-and-effect, moving beyond simple correlation.

Unifying for the Future: A New, AI-Driven Structure

CZI announced a major strategic shift: unifying its disparate science initiatives into a single, integrated Biohub. This new organization will be led by Alex Rives of Evolutionary Scale, an AI researcher with a deep understanding of biology.

Strategic Implication: Placing an AI leader at the helm of the entire scientific program signals that AI is no longer just a tool for analysis but is now driving the biological research agenda. This integration creates a tight feedback loop, or "flywheel," where the needs of the AI models directly inform the collection of new biological data.

This unified structure is designed to solve the communication gap between AI engineers and biologists, allowing them to work "shoulder-to-shoulder" to build increasingly accurate models of human biology.

The New Lab: Compute Clusters Over Physical Space

In a trend familiar to the AI industry, CZI's expansion is focused less on physical lab space and more on computational power. They were one of the first in the biological field to build a large-scale GPU cluster (1,000 GPUs) and have plans to expand to 10,000.

Actionable Insight: Access to large-scale, specialized compute is becoming the primary bottleneck and key differentiator in biological AI. This creates opportunities for infrastructure providers and highlights the immense capital requirements for staying at the frontier, mirroring the dynamics of the LLM market.

CZI also provides access to these resources for external scientists, seeding collaborations and enabling research that would be impossible for individual academic labs.

Conclusion

This conversation reveals a strategic shift where AI is not just analyzing biology but actively shaping its future. For Crypto AI investors and researchers, the key takeaway is the emergence of open-source, foundational biological models. These "virtual cells" and their underlying datasets represent a new, programmable layer of biology, creating fertile ground for AI-native therapeutic startups and decentralized science (DeSci) platforms.