No Priors

Mark Zuckerberg on Turning Biology Into an Engineering Science

Mark Zuckerberg· CEO of Meta at Meta
·~56 min·English·No Priors
AI SafetyMultimodalOpen SourceAI Company
TL;DR

Zuckerberg, Priscilla Chan, and Alex Rives on fusing frontier AI with frontier biology to turn biology from a discovery science into an engineering science — and cure all disease this century.

01Core Mental Model

From Discovery to Engineering Science

<strong>Biology has been a discovery science — build world models of cells and you turn it into an engineering science</strong> where you simulate before you ever touch a pipette.

What if we could actually understand how biology worked? Um, move it from a discovery based science to an engineering based science where we could systematically understand how living beings, living cells worked

— Priscilla Chan, No Priors
Key Insight
The pair never say 'we will cure disease.' They consistently say 'we will build the tools so the field can.' That framing is load-bearing: it lets a philanthropy claim a century-scale mission without owning the outcome, and it reframes every failure as a tooling gap rather than a scientific dead end.
02Origin Story

Cure All Disease — The Goal Scientists Laughed At

<strong>Nobel laureates laughed at the century goal — the real bottleneck was never ambition, it was shared tools and a knowledge base nobody was building</strong>.

I thought that by the end of the century was a stretch. Now, I think it's like uh too conservative.

— Mark Zuckerberg, No Priors
Key Insight
Notice the inversion: the scientists weren't skeptical of the science, they were skeptical of the sociology — silos, unshared tools, methods that die with the PhD who built them. The Biohub thesis is that AI's real contribution is organizational: it makes sharing data finally worth the effort, because something can actually reason over the pile.
03Method

Build It Up Hierarchically

<strong>You can't jump straight to cells — model proteins, then cells, then systems, layer by layer, with bridging data tying each level to the next</strong>.

a big part of the strategy is this view that you need to build it up hierarchically.

— Mark Zuckerberg, No Priors
Key Insight
This is a quiet argument against the 'just scale one giant model' religion. Zuckerberg is explicitly saying each biological layer is qualitatively different and cannot be skipped — the field's wins will come from the connective data between layers, not from a single monolithic model trained end-to-end. That is also why a wet-lab org has to be physically bolted to the AI org.
04The Binding Constraint

Data Is the Real Constraint — Not Compute

<strong>Unlike language models, you can't download biology off the internet — you have to invent new science to generate the data the models need</strong>.

it's not just like there's some factory somewhere that you can pay to produce the the data like you actually need to invent new novel scientific approaches

— Mark Zuckerberg, No Priors
Key Insight
This is the structural reason the Biohub can't be a pure AI lab and can't be a pure biotech. In language AI, the data precedes the lab; here the lab must precede the data. So the wet-lab effort isn't a cost center attached to a model team — it is on the critical path, and it sets the ceiling on how good any model can get.
05The Vision

Treat the Individual as an Individual

<strong>Today medicine guesses by analogy — trace the genetic-to-protein-to-disease chain and you design a drug bespoke to one person</strong>.

my goal is to be able to treat the individual as an individual, understand the mechanisms and be able to intervene.

— Priscilla Chan, No Priors
Key Insight
Read this against how medicine actually works today: a sick patient Googles PubMed, checks the supplement, hunts the methods, and asks 'am I even represented in this paper?' The vision implicitly concedes that most clinical practice is statistical matching of populations, not understanding of a person — and that AI's rare win in medicine is collapsing that gap from 'analogy to a cohort' down to 'mechanism of one.'
06The Proof

Protein Design as an Emergent Property

<strong>They never built an antibody model — one general protein model, trained on billions of sequences, designs nanomolar binders as a side effect</strong>.

we didn't design a model for antibodies. We didn't design a model to, you know, to be able to bind one particular target. You know, we just designed a model that could understand proteins and you kind of get protein design as an emergent property.

— Alex Rives, No Priors
Key Insight
This is the strongest evidence in the interview for the 'engineering science' thesis in section 01: a model with no therapeutic objective beat task-specific pipelines at the task. The implicit claim — that biology generalizes the way language does once you train on enough of its raw sequence — is exactly the bet that, if it holds up the hierarchy to cells, makes the century goal plausible rather than aspirational.
07Structure

Open Source Over Venture

<strong>Putting tools in everyone's hands beats monetizing each piece — the long tail of rare diseases only gets solved by decentralizing the science</strong>.

We'll have a bigger impact by getting this in more scientist hands quicker by doing it as open source projects instead.

— Mark Zuckerberg, No Priors
Key Insight
The rare-disease cell is the real argument. Priscilla's framing — that rare-disease patients self-organize registries, biobanks, even their own trials, and move in '3 to 5 years rather than decades' — is what makes 'open' a load-bearing structural choice rather than a values flourish: a closed model physically cannot reach the long tail, because no single firm can monetize each niche.
08Philosophy

No Central Superintelligence

<strong>The future isn't one AI that solves all science — it's a tool in every scientist's hands, which makes people more important, not less</strong>.

Our vision is not that there's going to be like some central super intelligence that solves all of science. I think like people are really important and I think we'll be more important in the future

— Mark Zuckerberg, No Priors
Key Insight
This is the through-line that reconciles a Meta CEO with a cure-disease mission: progress historically comes from empowering individuals to try 'things somewhat out of the mainstream,' and AI — done open — is just the latest tool that widens who can try. Read cynically, it is also a convenient defense of a decentralizing, open-source posture against the firms building the one big closed model.