The Existential Question You Probably Don’t Want to Ask AI
Chatbots may get pretty good at predicting how long you will live.
My wife, Mary, and I both turned 65 this year, and have developed divergent views on the necessity of discussing aging, senescence, infirmity, cognitive decline, and the inevitable journey to the undiscovered country from whose bourn no traveler returns.
“You talked about death all night!” she chided me after a dinner party in which, to set the record straight, I talked about getting old and dying only about half of the time.
I concede that death, as an idea, does not lend itself to amusing and enjoyable conversation. It’s very possible that I’ve become a Mortality Bore. (In my defense, there are worse ways to be boring: My brother, who has become a Wellness Bore, can talk until he’s blue in the face about his techniques for preparing his oatmeal.) Mary claims that my new interest in mortality comes from having a back injury, followed by the overdue revelation that I’ve become just another old guy with a ready supply of ice packs and painkillers. This is hard to deny. But what really forced me to stare into the darkness was the guidance from my professional helpers: my financial adviser, my lawyer, my doctor, and my AI, Claude.
“How long do you expect to live?” my financial adviser asked me eight years ago, when I was still in my 50s. Life expectancy is very important for financial planning because you have to decide, for example, whether you’ll live long enough to “break even” if you delay taking Social Security. He put together a plan showing me dying at 80, which seemed at the time a prudent, conservative choice. But in recent years that “80” has been making me uneasy. This summer, I asked the financial planner to let me live a bit longer on paper.
About the same time, I finally got around to hiring a lawyer and putting together an estate plan with Mary. The process was somewhat sobering, which should not have been surprising given the premise of the endeavor. Among other things, we had to decide about end-of-life care. Apparently, between life and death there could be some sketchy interludes. The medical directives asked us to decide what we would do in each of the following scenarios:
If my doctors certify that I am in an end-stage condition—that is, an incurable condition that will continue in its course until death and that has already resulted in loss of capacity and complete physical dependency.
If my doctors certify that my death from a terminal condition is imminent, even if life-sustaining procedures are used.
If my doctors certify that I am in a persistent vegetative state, that is, if I am not conscious and am not aware of myself or my environment or able to interact with others, and there is no reasonable expectation that I will ever regain consciousness.
“Which comes first?” Mary asked the night before we signed all of the documents.
“I think ‘terminal condition is imminent,’” I said, hesitantly, wondering whether some of these dire scenarios can overlap, like a Venn diagram. Right at the center, there might be an imminent end-stage terminal vegetative state.
When is the right time to prepare for the endgame? More tools than ever seem to exist to give us a clue. Recently I’ve come under the care of a very young and excellent doctor, Dr. M. Younger doctors, in my experience, are more likely to be technologically and pharmacologically aggressive. They will cram a lot of data onto their online portals. If I were to describe Dr. M’s style, it would be “search and destroy.” She ordered up new tests, put me on new medicines. I now have many more diagnoses and potential diagnoses and things to keep an eye on.
The good news is that I am on the right track, in great hands, nicely medicated. The bad news is that my online patient portal does not exactly scream “healthy young person.” This is why I’ve turned to Anthropic’s chatbot, Claude, for additional counseling.
There’s a saying: Age is just a number. That’s a feel-good statement that makes no sense on its face. Age is actually a lot of numbers, starting with, yes, your birth date, and includes such things as your resting heart rate, blood pressure, LDL, HDL, HbA1c, ALT, AST, BUN, TSH, eGFR, hs-CRP, ApoB, Lp(a), PCSK9, and so on. With modern medicine throwing more numbers at us every year, it’s tempting to turn those numbers over to AI to help make sense of them.
Me: What does this mean, in my doctor’s patient profile notes: 2026 Abnormal echocardiogram TTE 10/2025: LVEF 65-70%, mild concentric LVH, grade I diastolic dysfunction, trace MR, trace TR, trace PR.
Claude: This echo shows a heart that pumps normally and strongly, with mild, common, age-related changes: slightly thickened muscle and slightly less flexible relaxation, most likely connected to blood pressure over time, plus incidental trace valve findings that are essentially noise. None of these findings individually or together suggest heart failure, significant valve disease, or dysfunction requiring urgent treatment.
This was reassuring! So long as you don’t look too closely and obsess over phrases such as slightly less flexible relaxation. And you definitely don’t want to ask a question for which you are not prepared to accept the answer. Stupidly, I asked Claude whether a 65-year-old is still “middle-aged.” Nope!
I can, of course, seek guidance on these numbers from an actual human being, and I’ve certainly done that. Dr. M is remarkably responsive and generous with her counsel, even at night and on weekends. But I feel guilty when I barrage her with questions, some of which have a whiff of hypochondria. I feel only a modest amount of embarrassment when wasting Claude’s time. Whereas once Google fed our medical anxieties, now the bots are always standing by.
Having access in our pockets to a superintelligent, eager-to-please, but sometimes-delusional polymath is, at best, a mixed blessing. It’s great to have Hippocrates so near at hand, but the relationship could become obsessive. (The writer pauses here to double-check the potentially dire significance of ringing in the ear.) Chatbots don’t always agree. Claude told my author friend Robert Wright to take his relative to the ER immediately. But when Wright checked Gemini, it said to chill out—no big deal. Gemini turned out to be right, and ChatGPT (entering stage left) helped explain why: The lab report indeed contained an alarming number, but that number had to be a mistake, ChatGPT said, given the other numbers in the report. My big fear is that I’ll give Claude all of my medical numbers, and in just two seconds it will write back: These data appear to come from a cadaver.
AI, in theory, does have the power to give you an estimate of how long you are likely to live. The Social Security Administration will also provide an estimate, but based on just two factors: your sex and birth date. (A man with my birth date will, on average, live another 18.7 years and die at the age of 84.5, which for me would be right around July 2045.) Other, more advanced online calculators exist that account for a greater variety of demographic and health data. But AI opens the door to even more tailor-made analysis. AI-powered death-prediction tools have already attracted some headlines. Just this week, a study was published about a model that can predict a cancer patient’s survival based on one photograph. Feed an advanced model enough of your own information and have it compared against a large-enough data set, and you could find yourself staring at a number that feels more ominously predictive, more personal—more dead certain, if you will—than a mere statistics-based national average.
I’ve been tempted to command Claude to pull together everything it has on me, hs-CRPs and all, and give it to me straight. But I won’t. For one thing, it seems creepy and maybe even unwise to let a bot get that deep into my personal life. How secure is that data? Some people may rightly worry that an insurance company or employer would find this info actionable. But in my case, I mostly just prefer the hazy future. There’s a concept in medicine that has been around for several decades: the right not to know. I want to be oblivious about the date of my oblivion. A fuzzy future should be considered a fundamental human right, especially for those of us prone to fretting (and counting how many days until July, 2045).
In this respect, I hesitate to admit that my brother, the Wellness Bore, may be an inspiration. He is 67, and when I asked him recently whether he has made plans for later in life, he proudly declared that he has not. He doesn’t think about this stuff at all. His wellness kick has left him feeling young again. “I’m just going to go with the flow,” he told me. “You live for the day. You don’t have regrets. Everything’s cool.”
Including the oatmeal! I discovered that he boils it a day ahead and cools it overnight to be included in his big breakfast extravaganza. His recipe, for any of you who might be curious: “I have the plain, nonfat, no-sugar Greek yogurt with organic blueberries with some maple syrup added, and I add some walnuts, and then I’ll add the organic oatmeal from Whole Foods and top it off with chunks of cantaloupe and some peach chunks.”
You know, he might be onto something.