I gave ChatGPT my medical data. It was much less scary—and extra helpful—than anticipated

For apparent causes, permitting ChatGPT to present medical recommendation is fraught with controversy. Is it actually a good suggestion to belief a big language mannequin with our well being? What if it hallucinates, or misdiagnoses clear warning indicators (as a just-filed lawsuit alleges)? And may we belief AI chatbots with our personal medical knowledge?

These are all good causes for concern, however they’re not stopping OpenAI from rolling out ChatGPT’s new health-oriented options to all U.S. customers. First launched earlier this yr, ChatGPT Well being can — along with your permission — faucet into a wide range of medical knowledge sources, starting from telemetry from the Apple Well being app to data out of your physician, and (as OpenAI fastidiously phrases it) allow you to “perceive your info in context.”

Now, it simply so occurs that hours earlier than OpenAI introduced its broader ChatGPT Well being rollout, I had my annual bodily. It was a routine go to (no main issues, phew!), and I’m fortunate to have a common practitioner who’s pleasant, all the way down to earth, and clear about his findings. Amongst his directives: watch my weight-reduction plan, train extra, and in the reduction of on the cocktails. (Oh, c’mon, Doc!)

Usually, I wouldn’t really feel the necessity to feed my checkup particulars to an AI chatbot; usually talking, I solely like understanding what I must find out about my numerous well being circumstances. However this is PCWorld, and I cowl AI for a residing, so I took a deep breath and agreed to let ChatGPT verify my chart.

Simply to be clear, proper up entrance: This isn’t a full assessment, and I’m in no manner suggesting you strive ChatGPT Well being your self except you actually need to. Whereas the service makes use of a safe portal for linking medical data to ChatGPT and permissions controls abound, I imagine there are nonetheless loads of privateness considerations, and certainly not ought to ChatGPT’s musings be confused with the prognosis of a physician (as OpenAI is fast to say). The one motive I’m wading into ChatGPT Well being is so you may sneak a peek at my expertise.

With that prolonged preamble, I’ll begin with the nuts and bolts. Navigating to the “Well being” menu merchandise within the ChatGPT sidebar (iOS and net just for now, and once more, only for U.S. customers), a connection wizard guided me by way of the choices, together with connecting Apple Well being and my healthcare supplier.

For every well being supply you join, you will need to resolve exactly which knowledge you’d wish to share, and the choices are fairly granular. With Apple Well being, for instance, you possibly can select to share your each day steps however not your stand minutes, or decide your coronary heart fee however exclude coronary heart fee variability. 

The identical goes along with your well being supplier; mine, for example, had particular person check-off sections for allergy symptoms, care plans and scientific notes, and demographic info, for starters. You can even select how lengthy you need ChatGPT to have entry to your medical knowledge: anyplace from a day to a yr.

Subsequent, ChatGPT Well being will observe up with extra choices, together with including drugs and household historical past particulars, in addition to any present medical circumstances you’re experiencing.

With the up to date ChatGPT Well being, you don’t have to go to the Well being part within the ChatGPT app to debate your vitals (though you continue to can if you wish to). As an alternative, ChatGPT can entry your well being knowledge from any chat. Ask ChatGPT a well being query, and a permissions pop-up will seem, with choices for “Enable As soon as,” “All the time Enable,” or “Deny.” When you select “Enable As soon as” (as I did), ChatGPT will ping you for permission for subsequent healthcare-related chats.

Ben Patterson/Foundry

In my first conversations with ChatGPT Well being, I requested some common questions based mostly on my Apple Well being knowledge, together with whether or not it noticed any traits over the previous 12 months. We had a number of back-and-forths in regards to the uncooked knowledge ChatGPT had entry to and the doable sorts of studies it may compile; certainly, ChatGPT appeared virtually timid right here, and I needed to nudge it (“OK, are you able to do the annual well being assessment?”) to get its evaluation.

When it lastly did come, ChatGPT’s abstract was attention-grabbing if (to me) not all that stunning: “Your yr seems secure total, with a number of modest enhancements towards the top.” It singled out my decrease resting coronary heart fee, larger HRV (a sign for total cardiovascular well being), my sooner strolling pace, and extra frequent train over the previous yr. 

However I additionally bought dinged for my sleep patterns: not sufficient, in a nutshell, though my oxygen saturation has been trending up.

Subsequent, I moved on to my latest bodily. What did all these check outcomes popping in MyChart imply, anyway (and what’s up with the one within the yellow vary)? What’s a lymphocyte depend, and why is mine a tad low? And the way am I doing total?

In fact, I already knew the reply to most of those questions, courtesy of my flesh-and-blood physician (nothing to fret about with this, let’s control that). However I do need to say, ChatGPT did a pleasant job of filling in a few of the particulars and reminding me of others, like the truth that I want a pneumonia vaccine, or that my LDL ranges (for “dangerous” ldl cholesterol) are barely elevated. “Eat with coronary heart well being in thoughts,” ChatGPT stated, mirroring my physician’s recommendation.

Ben Patterson/Foundry

Certainly, all of ChatGPT’s observations have been grounded in my physician’s written evaluation plan — which, I’m positive, is the entire thought behind ChatGPT Well being. Nothing I noticed in my ChatGPT Well being chats was stunning or surprising; as a substitute, it simply (as promised) gave me extra context, together with explanations of medical terminology that I heard my physician use however hadn’t thought to ask about.

Now, I ought to level out that whereas I can revoke ChatGPT’s permission to entry my well being knowledge at any time, my chats with ChatGPT about my well being will stay except I delete them manually, which means you’d need to be scrupulously about pruning your previous conversations to wipe your medical particulars fully from the chatbot’s reminiscence. (You would, in fact, limit health-related chats to momentary conversations which are instantly deleted as soon as closed.)

General, I’d name my (transient) expertise with ChatGPT Well being to be attention-grabbing however not all that important. I ought to notice that I didn’t spot something all that doubtful, with most of its judgement calls and observations rapidly adopted by “verify along with your physician” hedging. And ChatGPT did remind me of that pneumonia shot I’m alleged to take, in order that goes within the “plus” column.

However would I flip to ChatGPT if I have been dealing with a severe medical situation? Frankly, I’d be extra snug turning to a doctor first. (Then once more, my physician did use an AI note-taking app throughout my examination, so…)