1 Plan your four weeks with ChatGPT: research first, then decide

AIM-5 · AIM-9 · AIM-X · about one hour

What you’ll learn

By the end of this chapter you will have:

Set up ChatGPT for the course, with a project that keeps your chats together.
Asked ChatGPT to research what medical educators think doctors in training should know about AI, and checked two of its sources yourself.
Worked with ChatGPT to turn that research, and your own goals and background, into a plan for the four weeks.
Saved the plan in a form you will use: a document, a spreadsheet, or slides, plus your journal template.
Decided whether to share your plan with your classmates.
Learned how to come back to the plan each week and change it.

Time. About 10 minutes of reading, 10 minutes of setup, and 40 minutes of doing. The training video is optional. This chapter is the Thursday hands-on session in week 1; if you miss the session, do it on your own. It is about you and your learning, so nothing in it involves a patient.

1.1 The list you cannot finish

It is Thursday of week 1. You have sat through three talks. You have an interview next Tuesday and a sub-internship evaluation you are still waiting for. The course offers a core chapter and a menu of others each week, guest talks in cardiology, pediatrics, and psychiatry, and a set of open labs. You have about eight hours a week. You cannot do all of it, and nobody has told you which parts matter for you.

You will have this problem again on the first day of every rotation in residency. There is a reading list nobody finishes, and a senior who says “read what matters.” The residents who handle it well do two things in order. First they find out what the field expects: the rotation’s goals, what the attendings ask about, what the specialty’s own guidance says. Then they compare that with what they already know and want, and choose.

That is research before deciding. A chatbot can do much of the first step in a few minutes, if you ask it to show its sources. The second step is yours. This chapter does both, and at the end you have a plan for the next four weeks that started with evidence and fits you.

1.2 Why this matters

This is your first real task with ChatGPT in the course, and it is not patient care on purpose. The stakes are low, the topic is you, and you will know at once whether the answer fits. That makes it a good way to learn what the tool can do: search the web, summarize, interview you, and make a document or a spreadsheet. You will use the same features later on harder problems.

The habit is the bigger point. When you bring any problem to a chatbot, a general question gets a general answer. “What should I learn about AI?” gets the same list for everyone. If you first ask what is already known, with sources, and then add your own situation, the answer is about your problem. This works for a learning plan, a quality-improvement project, or a talk you have to give. It is the same order you use with a patient: gather the history and the evidence before you decide.

1.3 How it works

1.3.1 Ask for research, not an answer

ChatGPT can search the web while it answers. When it does, it shows links to the pages it used. It usually searches on its own when a question needs current information; if an answer has no links, ask it to search.

Links are not proof. A chatbot can attach a real link to a claim the page does not make, or describe a page that does not exist. So you open a few of its sources and read what they actually say. The foundations chapter explains why this happens; here you only need the habit.

1.3.2 You decide

The research tells you what educators and professional groups think matters. It cannot tell you that you are going into psychiatry, that you learn best by building something, or that you are worried about what to say when a patient asks whether AI wrote your note. Only you know that. A good plan uses both: the field’s view, so you do not miss something important, and your own, so you actually follow it.

1.3.3 A plan that changes

The plan is a first draft. In week 2 a talk will interest you more than you expected, or a chapter will turn out to matter less. When that happens, bring the plan back to ChatGPT and change it. Each week’s learning journal asks what you chose to learn and why, so the plan and the journal work together.

1.4 Watch or try

OpenAI’s training site has a guided practice, Plan something from start to finish (about 20 to 25 minutes). It runs in your browser with sample data, so you do not need an account. You plan a team dinner in six steps. This chapter’s Do uses the same six steps for your four weeks:

Table 1.1: The six steps of the training practice, and what each one becomes here.

Step in the practice	What you do in this chapter
Prepare: gather your sources	Ask ChatGPT to research what doctors in training should know about AI
Summarize	Group the research into themes and match them to the course
Decide	Let ChatGPT interview you, then choose your goals
Create	Make your plan and your journal template
Share	Share a goal or your plan, if you want to
Follow through	Come back to the plan each week and change it

The practice uses ChatGPT Work, which can connect to email and files. Your university account may not have Work turned on. Everything in this chapter’s Do works in an ordinary chat, so you do not need it.

1.5 Set up ChatGPT

Use your university account. At CU Anschutz that is ChatGPT Edu; the Canvas page “Get your ChatGPT Edu account” explains how to get it. If your account is not ready, use any free chatbot for today and move your work over later.

Go to chatgpt.com and sign in with your university email.
Check the bottom left of the screen. At CU Anschutz it should say “CU - Anschutz (Edu)”. If it does not, you are in a personal account. Say so in the session, or ask at office hours.
Make a project for this course. Find Projects in the left sidebar, make a new one, and name it “AI in Medicine”. A project keeps the chats for this course together, so you can come back to them.
Start a new chat inside the project.

1.6 Do: build your plan

Copy each prompt into the same chat, in order. Fill in the parts in brackets first. ChatGPT remembers what came earlier in the chat, so each step builds on the last.

1.6.1 1. Prepare: gather what the field says (10 minutes)

I'm a fourth-year medical student applying in
[your specialty, or "undecided"]. I'm starting a
four-week course on AI in medicine.

Search the web for what medical educators and
professional groups say doctors in training should
know about AI. Look for curriculum frameworks,
competency lists, and statements from groups such as
the AAMC or specialty societies, from the last three
years. Also look for what AI is changing in
[your specialty].

List 6 to 10 sources. For each one, give who wrote it,
the year, the link, and one sentence on what it says.

Now open two of the links. Does the page exist? Does it say what ChatGPT said it says? Write down what you found. That is your check for question 2 of this week’s learning journal.

1.6.2 2. Summarize: find the themes (5 minutes)

Group what these sources say into 4 to 6 themes: the
things doctors in training are expected to know or be
able to do about AI. For each theme, name the sources
that support it. Mark any theme that only one source
supports.

Then give ChatGPT the course. It cannot read this book’s website, so the prompt below includes the list of chapters. Copy all of it.

Here is what my course offers. For each theme, which
parts of the course cover it? Which themes does the
course not cover?

Week 1: How these systems work (how AI models learn,
  and why chatbots make things up). Thursday: this
  planning session.
Week 2: Prompting lab (getting good answers, and
  knowing their limits). Also: Critical appraisal
  (checking AI tools, their answers, and the evidence);
  Bias and equity; Grounding AI in real sources
  (tools that answer from documents you give them).
Week 3: Talking to patients about AI. Also: Agentic
  coding (build a small clinical tool, then test it);
  an ambient scribe lab (draft a note from a recorded
  visit, then check it).
Week 4: Ethics: what you owe when you sign a note
  drafted by AI. Also: Will AI take your job?; What AI
  costs the planet.
Every week: talks, guest talks from clinicians in
  several specialties, and a weekly learning journal.

The preface has the same list with links to each chapter.

1.6.3 3. Decide: let it interview you (10 minutes)

Now interview me, one question at a time, about what I
want from this course: my background, what I have
tried with AI so far, what worries me about it, how I
learn best, and where AI might help my work as a
resident.

After about six questions, suggest 3 to 5 learning
goals in my own words. Link each goal to a theme from
the research, or say that it is mine alone. Then ask
me what I want to change.

Answer each question in a sentence or two. There are no wrong answers. If you do not know what you want yet, say that; it is a useful answer.

When ChatGPT suggests your goals, make them yours. Change anything that does not sound like you. Ask for a goal to be smaller, bigger, or more about your specialty. Keep going until you would be happy to show the goals to a colleague. A goal that is yours alone, with no source behind it, is fine.

1.6.4 4. Create: make the plan and the journal template (10 minutes)

Turn my goals into a plan for the four weeks. For each
week, list the chapters, talks, or labs I will use,
one thing I will try with AI, and how I will know it
worked. Keep it to about 8 to 10 hours a week,
including live sessions. Leave room to change it.

Ask for the plan in the form you will use: a table, a Word document, a spreadsheet, or a few slides. ChatGPT can make a file you download. If that does not work, copy the table into Word or Google Docs.

Then make your journal template in the same chat:

Using my goals, make a weekly learning journal
template I can fill in within 30 minutes, as a short
markdown template.

It must include these five course questions:
1. What I chose to learn, and why
2. How AI helped, and where it got in the way
   (including one time I felt I needed to check
   the AI: how I checked, and what I found)
3. What surprised me
4. What changed for me
5. What I used, and whether it was worth my time

Add one or two headings of your own that connect each
entry to my goals.

Move, rename, or cut the extra headings until the template suits how you think. The five course questions stay.

Save both in a document you will keep, named “AI in Medicine plan”. Your chat stays in the project, so you can come back to it.

1.6.6 6. Follow through: come back each week

The plan is meant to change. Each week:

Open the “AI in Medicine” project and start a new chat.
Paste in your plan and your last journal entry.
Ask:

Here is my learning plan and my last journal entry.
What did I do, what did I skip, and what should I
change for next week? Ask me before you change
anything.

Start a new chat each week rather than one long one. Very long chats can lose track of what was said earlier. Your document carries what matters from week to week.

1.7 Check

Look at your plan and ask:

Table 1.2: What a good plan looks like.

Check	A good plan
The research is in it	You can say which theme or source led to each part, or that the part is yours alone
You are in it	It mentions your specialty, how you learn, or what worries you, not only the course’s topics
You checked something	You opened two sources and know whether they said what ChatGPT said
It fits	About 8 to 10 hours a week, with room to spare
It can change	You know when you will look at it again, and how

1.8 Standing question

Go back to Thursday morning and the list you could not finish. You still cannot do all of it. But you now know what the field thinks matters, which of those things the course covers, and which ones you care about.

In your journal this week, answer two questions. What did the research show you that you would not have picked on your own? And what did you keep in your plan that no source told you to?

1.9 Three rules, every session

No patient information in a chatbot unless your hospital has approved that tool for it. See patient information and AI tools.
Make it yours. Read what it gives you and change what does not fit.
You sign what you submit. Whatever helped you write it, it is yours.

1.10 Go deeper

OpenAI, ChatGPT Training: short walkthroughs on projects, documents and slides, and voice. Several use ChatGPT Work, which your account may not have.
OpenAI, Projects and chats: when to use a project, and what it keeps.
What the field expects: the AAMC’s work and the published AI competency frameworks for medical students. Use it to check what your research in step 1 found, and what it missed.