Here’s how we live forever

Everyone is talking about longevity. Twenty years ago this was a niche field, the preserve of calorie-restriction enthusiasts, rapamycin tinkerers, and a handful of academics counting telomeres. But now it’s everywhere. Altos Labs raised 3 billion before publishing a paper. Tech executive Bryan Johnson livestreamed his blood swaps. Calico, an Alphabet life sciences startup, has been quietly burning through Google money for a decade. And that’s fair enough. Who wouldn’t want to grow older slower—or not grow old at all? The problem is we really don’t know where to start. There is no good scientific model for longevity, which means we have no idea what “normal baseline” is for lifespan. Say you live to 103. How can you judge what factors got you there? Were they internal (genes, inflammation, hormones)? External (diet, wealth, a ventilator)? How do all these factors interact? THE POTENTIAL FACTORS So the field is making educated guesses. Among my favorites are: GLP1s: These diabetes drugs turned out to do a bit of everything. Less weight, less inflammation, fewer heart attacks. Not dying early is a good way to increase longevity. Exercise: Pilates, running, and also weight lifting. When I was at Mount Sinai as a student, we learned to recognize the Parkinson’s patients who did weight lifting. They had 10 times fewer less tremors. Love: Solitude is a biological state. Even if we don’t get it. Solitude has to be treated, or it can lead to a shorter lifespan and health span. Love is probably the answer. These could all be true. But don’t expect any of these to be the silver bullet. The reality is that everyone will have different needs when improving longevity. You might take magnesium at a dose that’s totally fine for you, but too low for your auntie, and too high for your next-door neighbor. There will be no generic approach to longevity. Precision longevity will be the key. AI FOR LONGEVITY And that’s where AI comes in. Because unravelling a problem of this complexity will require superhuman intelligence—artificial superintelligence for biology. That’s what we’re building at Owkin and it’s already helping us identify the smart questions to start asking: Why do some individuals with telomerase mutations remain resilient? Why does GLP-1 therapy induce muscle loss (GLP-1–associated sarcopenia), and how can we prevent it? But AI for longevity is not without problems. The biggest is a lack of data for the AI to learn from. We will need to find more large cohorts of patients that can tell us about putative longevity mechanisms. For example, those genetic telomerase mutations that lead to fibrosis or organ malfunction. Then there’s causation. Most longevity data is correlational. How do we untangle the causal signals from everything else happening throughout people’s lives? AI needs to get better at predicting causal mechanisms and at hypothesizing new biology. Also, reaching the endpoint of any experiment takes a long time, hopefully longer and longer. We need surrogate markers we can trust—biological-age clocks, organ-specific aging signals. But right now we haven’t aligned on one. The longevity hypotheses are going to come thick and fast. That’s where the money’s going and that’s where the interest is. And I hope we keep discovering important pieces of the puzzle. But until we can bring superintelligence to bear on the problem, fuelled by high-quality patient data, we’re not going to live forever. Thomas Clozel, MD, is CEO of Owkin.