Skip to content

By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.

Silicon Valley GirlSilicon Valley Girl

Stanford Doctor: Never Tell AI What You Think Is Wrong With You

Dr. Jonathan Chen is a Stanford physician and computer scientist who teaches doctors how to use AI. His team found that GPT-4 alone outperformed doctors using it on diagnostic reasoning tasks.

Dr. Jonathan ChenguestMarina Mogilkohost
Sep 25, 202648mWatch on YouTube

Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.

Great conversations run long. Their best ideas don't have to.

uListen is free on iPhone, iPad, Android, Chrome, and Claude.

See how

At a glance

WHAT IT’S REALLY ABOUT

Use AI for medical clarity—never feed it your diagnosis guess

  1. Stanford’s Dr. Jonathan Chen argues that modern AI may outperform many doctors on knowledge-heavy diagnosis tasks, but trust, accountability, and clinical judgment remain essential for real healthcare decisions.
  2. He explains why people (and even clinicians) can do worse with AI when they ask leading questions, lack domain context, or get anchored—causing the model to confidently reinforce the wrong path.
  3. Chen recommends using AI to translate medical jargon, summarize records, generate appointment questions, and explore alternatives through follow-ups, while avoiding framing prompts around your suspected diagnosis.
  4. He warns that long-record summarization can fail through omissions, date/number confusion, and reinforcement of erroneous “chart lore,” making verification and careful context management necessary.
  5. Looking ahead, he expects more autonomous AI-driven clinical workflows (including limited prescribing/refills) and sees the biggest near-term impact as democratizing access to scarce medical expertise rather than “curing all diseases.”

IDEAS WORTH REMEMBERING

5 ideas

“Human + AI” isn’t automatically better; integration and training determine gains.

Chen cites studies where GPT-4 alone beat physicians who had access to it, largely because many clinicians lacked chatbot fluency and because humans introduce anchoring and workflow friction. With training and better task design, human+AI improves—just not automatically.

Never lead the chatbot with your guess; feed it objective observations instead.

If you tell a model your suspected diagnosis (“Is this scabies?”), it tends to agree and elaborate, pulling you deeper into the wrong branch. He recommends describing only observable facts (symptoms, timing, photos, vitals) and asking for differential possibilities and what would change urgency.

AI’s biggest medical risk is omission and context confusion, not wild hallucinations.

Models often err by leaving out important details when summarizing long records (e.g., a lung nodule that needs follow-up). They also struggle with dates/numbers and can mix old issues with current ones (“context rot”), especially when you dump years of data at once.

Use AI to translate and prepare for appointments, not to self-prescribe.

A practical, high-value use is turning your labs/doctor notes into plain-language explanations, a concise timeline, and a question list for your next appointment—so scarce clinician time is spent on decisions rather than decoding jargon. This helps patients be better prepared without pretending the AI is the decision-maker.

Match AI use to the stakes: double-check when consequences are serious.

For low-stakes choices (e.g., minor OTC decisions), AI guidance may be “good enough,” but for high-stakes triage and major interventions (surgery, chemo selection), Chen urges verification with accountable professionals. He notes chatbots may also bias toward higher-acuity advice to reduce liability.

WORDS WORTH SAVING

5 quotes

The AI probably is smarter than most doctors, but knowing everything doesn't make you somebody who's worthy of trust.

— Dr. Jonathan Chen

If you don't have the medical context, the training, the subject matter, um, expertise, you might ask the question in the wrong way. You don't even know how to formulate it, and if you get down the wrong rabbit hole, the AI will very confidently, sycophantically agree with you and lead you all the way down the wrong path.

— Dr. Jonathan Chen

If you tell it what you're thinking, it's gonna have a very strong tendency to agree with you.

— Dr. Jonathan Chen

You have to say, "My daughter had a rash. She's scratching very heavily. Um, this has never happened before. This is what it looks like." And just, like, objective statements. Don't tell it what you're thinking.

— Dr. Jonathan Chen

It's you have competence, communication, and character, and you need all of them in one person.

— Dr. Jonathan Chen

Read the available transcript
AI vs doctor diagnostic performanceSycophancy and anchoring from leading promptsObjective symptom description and follow-up questioningSummarization errors and omissionsContext rot, dates/numbers limitationsChart lore and bias propagation in recordsStakes-based double-checking and accountability

Ask ChatGPT about this episode

Get its key ideas, plus a summary of each chapter linked to that point in the video, right in your chat.

  1. Add uListen AI to ChatGPT.

  2. Paste this prompt into a new chat.

    Use uListen AI to give me this episode's key ideas and a chapter-by-chapter summary, with timestamps: https://www.youtube.com/watch?v=oOpF0ZXSXuk

SPEAKERS

  • Dr. Jonathan Chen

    guest

    Stanford physician and computer scientist focused on applying and evaluating AI/LLMs in clinical care and medical education.

  • Marina Mogilko

    host

    Host of “Silicon Valley Girl with Marina Mogilko,” interviewing researchers and founders about AI careers and applications.

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

Read along on your phone, search it beside the video, or ask your favorite AI app what was said.