Learn it with AI

The main activity of the course: use AI to teach yourself something you care about · 2 to 4 hours a week

It is your job to learn where AI helps your work. Nobody can tell you in advance. The tools change every few months, and your work is not anyone else’s. The way to find out is to try it, often, on things that matter to you. You will learn where it fails the same way: by trying.

So each week, pick something you want to learn and use AI as your tutor and research assistant. The chapters give you a framework. This is where most of your learning happens.

Pick a topic

Choose something tied to your own path. Some starting points:

  • the AI tools already used in the specialty you are applying to
  • a clinical topic you want to be sharp on for residency
  • a paper, guideline, or debate in your field you keep meaning to understand
  • a skill: reading a statistics section, writing a research question, preparing for interviews

One topic can last all four weeks, or you can change each week.

Use your university account

Sign in to your university’s chatbot with your university credentials. Many universities create the account only after you finish a required AI training. At CU Anschutz, for example, you use ChatGPT Edu or Microsoft Copilot. Both are free to you, and the ChatGPT Edu account is created after a short Canvas course. The Canvas announcement tells you how to join it.

A personal ChatGPT account is not covered by the university’s agreement. Do not use it for university information.

This work is learning, not patient care, so use made-up cases, never real patient information. That is the only line; everything else is yours to try. See the patient information page.

If you would rather talk than type, ChatGPT Edu has a voice mode where your institution turns it on.

A starter kit

Copy a prompt, change the parts in brackets, and see what happens. Then change it again. Every prompt below is a starting point, not a recipe.

Make me a plan

I am a fourth-year medical student going into [specialty]. I have about
four hours this week to learn about [topic]. Make me a study plan in
short steps. For each step, say what I should be able to do at the end.
Ask me two questions first about what I already know.

Explain it three ways

Explain [concept] three ways: first as you would to a patient, then to a
medical student, then to an attending in [specialty]. After each, ask me
one question to check I understood.

Quiz me, and make me explain

Quiz me on [topic], one question at a time. Do not give the answer until
I have tried. When I am wrong, ask me a follow-up question instead of
correcting me straight away. Keep score and stop after ten questions.

Teach it back

I am going to explain [topic] to you as if you were a new intern. Point
out anything I got wrong, anything important I left out, and anything I
said that was more confident than the evidence allows.

Find me the sources

I want to understand the evidence on [question]. Tell me what kinds of
studies would answer it, what to search for in PubMed, and the names of
any landmark papers you know of. Mark every specific paper as
"to be verified", because you may be wrong about titles and authors.

Then look the papers up yourself. This is where a chatbot is most likely to invent something, and checking takes a minute: search the title in PubMed.

Argue the other side

Here is what I currently think about [topic or claim]: [your view].
Make the strongest honest case against it. Then tell me which of your
points is weakest.

Practise the interview question

You are a residency interviewer in [specialty]. Ask me, "What do you
think about AI in medicine?" and then two follow-up questions based on
my answers. Afterwards, tell me what was strong, what was vague, and
one thing a skeptical faculty member would push back on.

Make something I will keep

Turn what we covered into a one-page reference I can keep: the five
things to remember, two common mistakes, and three questions I should
still ask an expert.

One check a week

Each week, take one thing the AI told you that mattered and check it against a source: a paper, a guideline, a textbook, an expert. Note what you found in your learning journal. Sometimes it will be right. Sometimes it will be confidently wrong. Both are worth knowing, and the habit is the point.

When you get stuck

  • The answer is too vague. Say who you are and what you need it for. Ask for an example.
  • It is too long. Ask for five bullet points, or “answer in three sentences”.
  • It sounds sure but you are not. Ask, “How confident are you, and what would change your mind?” Then check it.
  • It will not help. Try the other tool. ChatGPT Edu and Copilot behave differently, and that difference is worth noticing too.

Bring what you are stuck on to Thursday’s office hours. Bring what worked, and what did not, to Friday’s small group.