THE HARD DATA PROBLEM AT THE HEART OF LONGEVITY AI
Living forever will require superintelligence, but first the field needs to solve a boring problem: getting good patient data.
by editor5 min readcomments soon
Alphabet's Calico has been quietly burning through Google money for a decade without a home-run breakthrough. Altos Labs raised $3 billion before publishing a single paper. Tech executive Bryan Johnson livestreamed his blood swaps. Twenty years ago, longevity was a niche field. Today, it is a circus of cash, hype, and earnest science trying to figure out which of the competing theories of ageing might actually work.
The problem is that nobody has a good scientific model for longevity. There are hypotheses, correlations, and a growing pile of interventions that seem to do something. GLP-1 drugs, originally developed for diabetes, also reduce weight, inflammation, and heart attacks. Parkinson's patients who did weight lifting had 10 times fewer tremors. Solitude is now understood as a biological state that can shorten both lifespan and health span if left untreated. These are real effects, but they are not a unified theory. They are pieces of a puzzle that no one has seen the full picture of.
THE MISSING PIECE
Unravelling that complexity will require something the field does not have yet: artificial superintelligence for biology. The human brain cannot hold the combinatorial explosion of factors that govern ageing: gene expression, epigenetics, metabolism, immune function, environmental exposure, social connection, and the feedback loops among them. A system smarter than any human researcher could ask the right questions and test them at scale.
Owkin, a startup working with Mount Sinai, is building AI to identify smart questions about longevity. Why do some people with telomerase mutations remain resilient while others do not? Why does GLP-1 therapy induce muscle loss in some patients? Those are the kind of precise, causal queries that move the field forward. But Owkin's effort, and every other AI-driven longevity project, hits the same wall.
THE DATA WALL
The biggest problem for AI for longevity is a lack of data. Not a shortage of spreadsheets, but a shortage of the kind of deep, longitudinal, multimodal patient data that could train models to separate cause from correlation. Most longevity data today is correlational: people who exercise live longer, people who take statins live longer, people who live near parks live longer. Those are signals, but they are not mechanisms. AI needs to get better at predicting causal mechanisms, and it cannot do that on thin datasets.
The solution is larger patient cohorts that track thousands of individuals over decades, collecting not just blood panels and genomic sequences but also lifestyle, environment, and psychosocial data. That is expensive, slow, and requires the kind of infrastructure that most academic labs and startups cannot fund. Calico, an Alphabet life sciences startup, has been quietly burning through Google money for a decade, yet the field still lacks the kind of longitudinal dataset needed.
THE MARKER PROBLEM
Even if the data arrives, researchers need a way to measure progress faster than waiting for people to die. Surrogate markers like biological-age clocks (epigenetic clocks, proteomic clocks, metabolomic clocks) are the obvious answer, but the field has not aligned on one yet. Every lab uses a different clock, and none of them has been validated against hard outcomes across diverse populations. The result is a literature full of promising interventions that lower one clock score in a small trial but have no detectable effect on another clock or on actual mortality.
Precision longevity is the goal: everyone will have different needs based on their genetics, history, and environment. A single pill or protocol for all humans is not realistic. But personalisation requires even more data, not less. It requires knowing not just that a treatment works on average, but why it works for one person and fails for another. That is the level of understanding that only causal models built on rich datasets can provide.
WHERE THE MONEY GOES
None of this has stopped the money from flowing. Altos Labs raised $3 billion before publishing a paper. Tech executive Bryan Johnson livestreamed his blood swaps. Venture capital is pouring into longevity biotechs. The field is not starving for attention or capital.
What it is starving for is high-quality patient data and the AI systems capable of making sense of it. The claim that is not marketing. It is a statement of the bottleneck. Superintelligence for biology does not exist yet, and even if it did, the training data is not ready.
SIGNAL VS NOISE
The next few years will separate the signal from the noise. Companies that are honest about the data problem and build the infrastructure to solve it (large, diverse cohorts, open standards for clocks, causal inference methods) will be the ones that matter in a decade. The longevity field has the energy and the money. What it needs now is the humility to admit that the hardest part is not living forever. It is getting the data to figure out how to get started.
what did you make of it?
more from ai
ai
TSMC ADDS $100 BILLION TO ARIZONA CHIP BET, TOTAL HITS $265 BILLION
The additional investment will build at least four more 2nm fabs and advanced packaging, bringing the company's total US commitment to $265 billion.
ai
META WILL ALERT PARENTS IF TEENS DISCUSS SUICIDE WITH META AI
The opt-in feature flags self-harm references in chatbot conversations, with human review before any notification is sent.
ai
ROBLOX'S "BUILD" LETS ANYONE MAKE A GAME FROM THEIR PHONE WITH AI
The new toolset, launching July 28, turns text prompts into playable experiences and puts game creation on iPhone and iPad.
ai
ZOOX REALLS ENTURE ROBOTAXI FLEET OVER SMOKE DETECTION FAILURE
A robotaxi drove into an active fire scene obscured by smoke. NHTSA called emergency scenes not edge cases.