Mark Zuckerberg & Priscilla Chan: How AI Will Cure All Disease...
a16z PodcastFull Title
Mark Zuckerberg & Priscilla Chan: How AI Will Cure All Disease
Summary
Mark Zuckerberg and Priscilla Chan discuss their decade-long initiative to cure and prevent disease, focusing on building infrastructure and tools through AI and frontier biology research.
They highlight the development of open-source data sets, virtual cell models, and a unified Biohub approach to accelerate biological discovery and enable precision medicine.
Key Points
- Chan and Zuckerberg's ambitious goal to cure all disease by the end of the century was initially met with skepticism, prompting a shift from simply increasing funding to developing fundamental new tools for scientific research.
- The core strategy is to accelerate basic science by creating new tools and infrastructure, recognizing that traditional grant funding often supports incremental, near-term research rather than the longer-term, expensive development of groundbreaking tools.
- AI is viewed as a critical enabler, bridging the gap between ambitious biological goals and current capabilities, allowing for the creation of tools like virtual cell models that can test hypotheses in silico before extensive wet lab work.
- The Cell Atlas project, a massive effort to catalog millions of cells, emerged somewhat serendipitously from the need for standardized annotation tools, leading to an open-source data standard for biology.
- Virtual cell models are being developed to simulate biological processes, allowing scientists to test riskier hypotheses computationally, thus reducing the cost and time associated with traditional lab experiments.
- The Biohub model integrates frontier biology with frontier AI, aiming to create a cohesive ecosystem where researchers can generate novel data sets and build increasingly generalizable AI models for understanding human biology.
- Success is defined not just by therapeutic development, but by enabling a broad community to build diagnostics and therapeutics, and by fostering a new wave of precision medicine that accounts for individual biological differences.
- The initiative is moving towards a more unified "Biohub" operating model under Alex Reeves' leadership, consolidating efforts to advance biology at the intersection of AI and biology.
- The importance of collaboration and interdisciplinary work is emphasized, with a focus on lowering barriers to entry for scientists from different fields to contribute to complex biological challenges.
- The development of advanced AI models, from protein structure to virtual immune systems, follows a hierarchical approach to build a comprehensive understanding of cellular function.
- The long-term vision includes democratizing access to powerful computational resources like GPUs to enable a wider range of scientific inquiry.
Conclusion
The integration of AI and frontier biology is accelerating the path towards curing and preventing diseases, with the Biohub playing a central role in building the necessary tools and data infrastructure.
By creating accessible tools and fostering collaboration, the initiative aims to empower a broader scientific community to make significant breakthroughs in understanding and treating diseases.
The long-term vision requires patience, a willingness to embrace ambiguity, and a focus on building a comprehensive understanding of human biology through advanced AI models and data.
Discussion Topics
- How can AI be best leveraged to bridge the gap between fundamental biological research and clinical applications for disease treatment?
- What are the biggest challenges in creating comprehensive and accurate virtual models of human cells, and how can these models be validated?
- Beyond technological advancement, what are the key organizational and collaborative strategies needed to accelerate progress in complex scientific endeavors like disease eradication?
Key Terms
- Transcriptomics
- The study of the complete set of RNA transcripts produced by an organism, which provides insights into gene expression.
- In silico
- Performed on computer or via computer simulation; "in silicon" as opposed to "in vivo" (in a living organism) or "in vitro" (in a laboratory dish).
- Cell engineering
- The modification or manipulation of cells to achieve specific functions or characteristics, often for therapeutic or research purposes.
- Cryo-EM (Cryo-electron microscopy)
- A technique used for determining the molecular structure of biological samples at near-atomic resolution.
- Large Language Models (LLMs)
- A type of AI model designed to understand and generate human language, capable of performing various natural language processing tasks.
- Virtual cell models
- Computational simulations that represent the structure and function of a cell, used for hypothesis testing and understanding biological processes.
- Wet lab work
- Experimental work conducted in a laboratory using physical equipment and biological samples, as opposed to computational analysis.
- Variant of unknown significance (VUS)
- A genetic alteration for which it is unclear whether it is benign or pathogenic.
- Philanthropy
- The donation of money to good causes.
- Interdisciplinary
- Involving or relating to two or more fields of study.
Timeline
Chan and Zuckerberg explain their mission to cure and prevent all disease, driven by a desire to make a difference and the realization of limitations in current medical knowledge for children with rare diseases.
They articulate their strategy of accelerating basic science through the invention of new tools, comparing it to the impact of microscopes or telescopes in scientific history.
The limitations of traditional government grant funding for long-term tool development are discussed, contrasting it with the need for larger, multi-year investments in new technologies.
The philanthropic approach emphasizes enabling scientists and startups, with the goal of accelerating progress without seeking personal credit.
The initial skepticism towards their ambitious goal is recounted, highlighting the shift from "bite-sized" science to "century-scale" science, which led to identifying barriers like the lack of shared tools and data.
The differing perspectives of biology and AI communities on the potential of AI are discussed, identifying a gap that needs bridging.
The Biohub's strategy of pairing frontier biology with frontier AI is explained as a way to accelerate discovery.
AlphaFold is cited as an example of AI built upon publicly available historical data, highlighting the opportunity to create specific data sets for training AI models.
The focus on the Biohub as the main philanthropic effort is mentioned, stemming from the significant returns observed in science research.
The 10-15 year time horizon for grand scientific challenges is discussed, emphasizing the need for a credible pathway and a degree of ambiguity for risk-taking.
The three Biohubs in New York (cell engineering), Chicago (tissues and cell communication), and San Francisco (imaging and transcriptomics) are described, along with their strategic locations.
The integration of AI and large language models is seen as a way to make sense of the vast amounts of biological data being collected.
Success in the therapeutic realm is viewed through enabling new medicines and deploying precision medicine tailored to individual biology.
The challenge of understanding individual biology, often lumped into broad disease categories, is highlighted, with potential for AI to analyze variants and gene expression for targeted treatments.
The value of their open-source data sets and tools for the startup and pharma communities is acknowledged, with an example of their cell hygiene atlases informing drug targets.
The development of the Cell Atlas project is explained as an effort to create a "periodic table of elements equivalent for biology" by standardizing data formats and annotation.
The promise of virtual cell models is discussed, enabling hypothesis generation and testing for various scientific endeavors, including precision medicine.
The addition of Evolutionary Scale researchers to the Biohub is noted, emphasizing a new leadership structure that leans towards AI expertise in biology.
The hierarchical approach to building virtual models, from proteins to virtual immune systems, is outlined as a strategy to achieve a comprehensive understanding of the human cell.
The concept of taking on riskier ideas through virtual biology, circumventing the cost and time of wet lab work, is presented as a key advantage.
The accuracy of virtual cell models is discussed, with the understanding that they don't need to be 100% accurate to be useful for de-risking ideas and identifying directional signals.
Specific AI models like "VariantFormer" for predicting genetic edit outcomes and diffusion models for generating synthetic cell simulations are mentioned.
The development of a reasoning model over biology is highlighted as an early but potentially important direction for AI in this field.
The hypothesis is that better world models and pre-trained models are needed for effective biological reasoning, with a hierarchical approach from proteins to cellular models.
The unification of past decentralized efforts into a single Biohub operating philanthropy under Alex Reeves' leadership is announced.
While doubling down on the Biohub, they will continue work in education and supporting local communities, but the Biohub will be the main philanthropic focus.
Advances in AI are expected to accelerate the timeline for achieving their disease cure and prevention goals.
The unique value proposition of the Biohub lies in its ability to bridge frontier AI and frontier biology, shaping and forming data sets to complement existing knowledge and complete a feedback loop.
The surprise in the AI industry is the effectiveness of domain-specific models, countering the initial narrative of general AI superiority, and the importance of problem specification.
The trend of creating open-source data access and the importance of domain-specific knowledge in curating and annotating this data are discussed.
The intentional design of user interfaces, like CellBitGene, to be accessible to a broad range of users, including those without deep computational or biological backgrounds, is emphasized.
The interdisciplinary nature of biology is highlighted, with immunology being linked to neurodegeneration, necessitating tools that allow cross-disciplinary understanding.
The Biohub will likely grow through both employing more people and expanding its network model with more sites and community-driven data sets.
Science is viewed as a portfolio of societal endeavors, with philanthropy aiming to support underrepresented areas and encourage collaboration.
The success of Biohubs in fostering collaboration, both within academic institutions and across disciplines (biologists and engineers), is highlighted as a key organizational strategy.
While Biohub is centralized, it relies on external labs for frontier work and is expanding compute resources, particularly GPUs, as a new form of "lab space."
The critical role of AI in creating a comprehensive biological pipeline, from basic science acceleration to therapy development and public health, is emphasized.
A 10-year reflection reveals that the initiative has delivered more than expected, providing a signal to continue doubling down, balancing long-term vision with patience.
The success is measured by the adoption and use of their tools by scientists for important publications, filling a crucial void in the research landscape.
The unique position of their work and the question of whether a problem would exist if they didn't are key indicators of founder-market fit.
Episode Details
- Podcast
- a16z Podcast
- Episode
- Mark Zuckerberg & Priscilla Chan: How AI Will Cure All Disease
- Official Link
- https://a16z.com/podcasts/a16z-podcast/
- Published
- November 6, 2025