• High-stakes AI

In medicine, a plausible wrong answer is a category of harm

Every argument for grounded, citable AI applies more forcefully in medicine than anywhere else, for three compounding reasons: the material is largely arbitrary, so you cannot reason your way to noticing an error; it is safety-critical; and guidelines differ by country, so a *correct* answer can still be the wrong one for your exam and your patients.

5 min readAI and learning

You cannot sanity-check arbitrary facts

In physics, a wrong answer often *looks* wrong — the units don't work, the magnitude is absurd. Medicine offers no such protection. A confidently stated wrong dose, wrong contraindication or wrong first-line agent looks exactly like a right one, because there is nothing to derive it from. It is arbitrary, and arbitrary facts are unfalsifiable to a novice.

And you are a novice — that is why you are studying it. The asymmetry is at its worst here.

Guidelines are local, and models are not

First-line agents, thresholds and protocols differ between countries and between guideline bodies, and they change. A model trained on the whole internet has absorbed all of them at once and has no particular reason to give you *yours* — so the answer may be entirely correct somewhere, and wrong for your exam and your hospital.

This is the strongest argument there is for grounding in your faculty's own material: it is the only source that is definitionally the one you are examined on. And it is why a downloaded community deck deserves more suspicion than it usually gets.

And spaced repetition compounds the error

A wrong pharmacology card doesn't just sit there. It gets drilled into you, at optimal intervals, for months, by a system explicitly engineered to defeat forgetting. Of all the places to memorise something false, this is the worst.

Common questions

Is it safe to use AI to study medicine?

Only if every answer cites a page in material you trust. Medical facts are largely arbitrary, so a wrong answer looks exactly like a right one — and you're studying precisely because you can't yet tell.

Why do AI answers about medicine differ from my course?

Guidelines and first-line agents differ by country and body, and a model has absorbed all of them at once. An answer can be entirely correct elsewhere and wrong for your exam.

Can I use AI as a clinical reference?

No. A study tool that reads your course material has no clinical authority, and nothing it produces should inform the care of a real patient.

Try it on your own material

Upload your notes, slides or lecture recordings and get a tutor that answers only from them — and says so when they don't cover it.

$12/month, or $8/month billed yearly · cancel in two clicks