Would you let AI predict the day you'll die?

I was reading a November 3 announcement from Differentia Bio this week. The company is developing what it calls a human digital twins, basically a computer model of how diseases behave and respond to treatments. Sometimes the media extends this to a digital copy of another human, i.e. to simulate responses in clinical trials.

The whole announcement is really about reporting that some mechanisms their models identified a couple of years ago have since received support from clinical findings. So as of right now, just models for drug research, not copies of patients. Overall, a nice update, but what I want to focus on here is the idea of a digital copy of you… and what happens in this future?

For example, your doctor could have a model built from your medical history, data from your smartwatch, and whatever other invasive tech you have to monitor your “health”. In theory, this model can now simulate how you would respond to treatments, but it can also estimate which diseases you’re likely to develop long before you show any symptoms.

Suppose you’re 42 and go in for a regular checkup, everything perfectly fine. Your doctor says everything looks normal, except their model of you estimates that you have an 80% chance of developing Parkinson’s disease before turning 60. Technically, you don’t have Parkinson’s, and maybe you never will! But I suspect you would leave feeling a little like someone who had just received a life changing diagnosis.

What is the real consequences of that? You walked in just to see if you’re ok, and now your whole world has changed. How would you live your life now? I’d be for most, they would notice things that they never paid attention to, maybe a slight hand tremour after too much coffee, or forgetting why they walked into a room, might suddenly seem significant.

There’s a phenomenon called the nocebo effect, where a negative expectation can make people experience real symptoms or make existing ones worse. That doesn’t mean worrying about Parkinson’s causes Parkinson’s… but anxiety can produce physical symptoms, and once you’re looking for signs of illness, ordinary life things suddenly become rather suspicious.

Now, in this AI future, your digital twin keeps collecting information after the “diagnosis”. You begin sleeping badly after the appointment and your smartwatch records changes in your heart rate. Would the model take this as further signs of evidence it’s right? i.e. going from 80% sure to 90%? So when you go back to your doctor, it just seems more convincing that this predicted future is going to be a reality.

So far, this is just a “you” problem, but what happens when you take this news home. Your partner might begin to worry about decisions that once seemed normal, like taking a demanding job or buying a house with a long mortgage. Blowing all your life savings over something that hasn’t happened yet.

Worse, maybe they begin to behave and act as if you’re already someone with Parkinson’s, and further enforced every time you forgot where you placed your car keys. You're still the same person, but someone else is already treating you as if you’re ill. I’m not sure I’d want a statistical version of my future self having that much influence over my marriage.

Once we start predicting diseases that far into the future, I suppose the next logical step is predicting how long we could live. What if, these models could eventually become accurate enough to tell you the date you are going to die… maybe plus or minus a few weeks. How would you feel about that?

Part of me thinks I’d like to know, maybe I could plan my retirement differently, travel while I’m healthy, and stop putting things off. But I might also spend years measuring ordinary experiences against that date. Every birthday is basically a countdown… I mean it kind of already is but now it’s even more real. One more thing to think about here is, what if you apply this to everyone… so not just you who has an expected date, but your family as well, everyone’s got a “date” constantly running in their minds.

I realize this is quite a long way from modeling disease mechanisms in a lab, and I still think these technologies could eventually do tremendous good. If we tone down the death part a bit, knowing about a risk that early could be enormously useful if there’s a treatment that can prevent the disease. For example, maybe a change in diet or lifestyle could add six healthy years to your life. When you see a bit of doom, you would rather spend your evenings doing things you enjoy.

In the end, I would certainly want to know about a medical risk I could do something about. But if we ever reach the point of showing people detailed predictions of their own lives, I wonder whether the hardest part of building a digital twin will be something other than our own biology. The model would also have to predict what happens after the real person finds out what it thinks their future will be.

Eli Lilly also happens to be the most active pharmaceutical company in our database when it comes to AI deals.

Since November 2025, it has signed 10 deals with seven AI companies, including a collaboration with Profluent that could be worth up to $2.25 billion.

Most of these partnerships are fairly familiar by now, using AI to design antibodies, mini-proteins and small molecules, or to automate laboratory work. That’s partly why the SpeechDx agreement stands out. It’s a rather different application of AI, where the information comes from how people speak.

And unlike experiments carried out in a lab, speech is something we produce all the time, often in places where nobody is thinking about medical testing.

Recursion reported encouraging early results for REC-4881, with a 47% median reduction in intestinal polyp burden after 12 weeks. The drug is being tested in people with familial adenomatous polyposis, an inherited condition where patients develop many intestinal polyps and face a much higher risk of colorectal cancer. In this small study, 13 of 15 patients showed a reduction in polyp burden. The results look promising, although we're still talking about a small, early-stage study without a control group, so there's quite a bit left to learn. [Link]

10x Genomics launched Sentira Single Cell, an AI tool that lets scientists analyze single-cell sequencing data by asking questions in everyday language. It's built using Anthropic's Claude and can work out an analysis plan, run it, and produce an editable notebook showing how the results were generated. So rather than writing code for every step, researchers can start with a biological question and let the system handle much of the computational work. The results still need to be checked, of course, but it's an interesting example of AI moving into everyday laboratory research. [Link]

The FDA approved Voceryq, the first plasmid DNA-based immunotherapy in the United States. Developed by Inovio, it's designed to treat recurrent respiratory papillomatosis, a rare condition caused by HPV where patients develop recurring growths in their airways, sometimes requiring repeated surgeries. The treatment works by delivering DNA instructions that help the immune system recognize HPV-related targets. Patients in clinical studies needed fewer surgeries, which supported the FDA's accelerated approval, although further studies are still required to confirm the clinical benefit. [Link]