Can AI restore humanity in healthcare?

Ten years. That's how long it took doctors to find my diagnosis.

Ten years. That's how long it took doctors to find my diagnosis.

Ten seconds. That's how long it took AI.

I fed my symptoms into a language model — just as a test, years after I'd already figured it out myself — and watched it arrive at POTS instantly. The same condition that thirty-three doctors had missed. The condition that cost me a decade of my life.

I sat with that for a long time. Not with anger — though that was there too — but with a specific, clarifying question: what does it mean that a machine could see me when the humans couldn't?

That question is what I took to TEDxMedUniGraz in March.

Here's the thing nobody wants to say out loud about AI in healthcare: we're afraid it will make medicine less human. That algorithms will replace the doctor who holds your hand. That efficiency will replace empathy. That the already-cold system will get colder.

I think we have it exactly backwards.

The system is cold right now. Not because of too much technology. Because doctors are drowning in administrative burden, diagnostic pressure, and a system that measures their performance by throughput. The human capacity for empathy doesn't disappear — it gets buried under everything that isn't care.

AI can take the information load. The pattern recognition. The cross-referencing of a thousand symptoms against a million data points. The paperwork. The coding. The scheduling.

And what gets freed up? The human part. The part that sits with you. The part that understands what "I'm fine" actually means when someone says it while gripping the chair. The part that can hold grief, create meaning, notice that a patient's hope is declining before their vitals are.

Machines handle information. Humans handle transformation.

That's the thesis. That's what I said on stage.

But there's a harder part. The part the talk doesn't get to resolve.

AI is trained on data. Medical data comes from systems that already exclude people. Women whose pain has been dismissed. Patients from under-documented populations. Conditions that don't match the textbook because the textbook was written without them.

Garbage in, garbage out. But worse: biased data in, confidently wrong algorithm out.

The question isn't whether AI should be in healthcare. It should. The question is: who gets protected when it fails? Who gets left behind when the algorithm is confident and wrong?

That's the thing I'm still sitting with. The thing the room needs to hold.

The talk is on YouTube. Watch it. Disagree with it if you want. Tell me where I got it wrong.

[Watch: Can AI restore humanity in healthcare? | TEDxMedUniGraz](https://www.youtube.com/watch?v=lMkt7d0i3fg)

The conversation isn't finished. It's barely started.