In this episode of The Future of Medicine, we welcome Jonathan Chen, MD, PhD, clinician, AI researcher, and Associate Professor at Stanford, whose work focuses on combining human and artificial intelligence to improve clinical decision-making.
Dr. Chen reflects on the rapid rise of AI in medicine, and the moment he realized everything had changed. He also walks through surprising findings from his research, including studies showing that AI alone can sometimes outperform doctors using AI tools. He explains why this happens, from human bias and “automation errors” to the ways AI systems are designed to agree with users, even when they’re wrong.
Looking ahead, Dr. Chen shares his perspective on the future of AI in medicine, including the risks of overreliance, the importance of clinical judgment, and how these tools could transform everything from medical education to patient care. He also explores the concept of “do no harm” in AI systems—and why safety and accuracy are not the same thing.
The Intro
“Confabulations, hallucinations... Your judgment matters even more than it did before.”
Dr. Jonathan Chen is a clinician and AI researcher who's making medicine smarter by turning real-world data into tools that actually work at the bedside.
“When I first saw a preview version of GPT-4 as things were building up, I was like, ‘Holy smokes.’ I literally felt like I need to throw away half of my research program.”
When he's not advancing the future of clinical decision-making, he's an accomplished magician, so he knows a thing or two about making the impossible feel possible.
“I'm using magic again as that algorithm. Boy, does that look real, but you still have to have that judgment to tell the difference.”
In our interview, he dives into how AI can support rather than replace human judgment.
“AI could be the best teacher you have ever had, but if you misuse it, you miss the point.”
And shares what he's learning from bringing these systems into real clinical practice. Welcome to Stanford Department of Medicine's inside look at the future of medicine.
The Transcript
Dr. Euan Ashley: Well, Jonathan, welcome to the future of medicine. You just came from our grand rounds being a speaker to a packed house, presenting not just cutting-edge ideas in AI and a lot of your own work, but magic. So let us start there — medicine or magic, which came first? How did you get there? How did you arrive at today as a faculty member here at Stanford who practices both?
Dr. Jonathan Chen: Who would be crazy enough to try all of the above? Um, I don't know. I played a little bit of magic when I was a 12-year-old boy, as many 12-year-old boys do. And then after a few months, I stopped doing it because I had nobody to perform for. My sister wasn't going to watch it more than one time, right? I gave it up.
Dr. Ashley: These were card tricks.
Dr. Chen: I learned a three-card monte, I learned some trick decks — I actually learned a few basic things, but probably nothing you'd be impressed by. Just kind of kid stuff. And then I became a doctor, somewhat more because my parents wanted me to. There's a little bit of story there as well.
Dr. Ashley: Well, tell us.
Dr. Chen: I started college unusually young. I was 13 years old when I started college. And I don't know about you — when I was 13, I did not know what I wanted to be when I grew up. If you had asked me when I was a kid, I would have said I wanted to be a stand-up comedian because I like to make people laugh and that felt good. I never said I want to be a doctor or a scientist or something like that.
Dr. Ashley: But how did you end up at college so early?
Dr. Chen: So there's a program in Los Angeles at Cal State LA. It actually started, I think, out of their psychology department — like an experiment, basically. They studied precocious kids in junior high and asked, do you want to basically be in this program? It just perpetuated. I took a test, they said you did okay, why don't you come in for summer quarter? And now there's a whole program — if you do okay, not just academically but in terms of emotional readiness, you could just start college full-time through that program.
Dr. Ashley: Okay, but "did okay" among the people being considered for college at age 13, right? Yes, so the scale is a little different.
Dr. Chen: Not that it was anything extreme, but that kind of dynamic. That was also very important and powerful because it's not like it was just me on a giant college campus. It was me and like 30-ish other kids, essentially. So I had a cohort to hang out with. A couple of years after that, I transferred to UCLA. That really was tough — now I'm a 15-year-old kid living on a gigantic college campus by myself, living in the dorms. That was very challenging, but it kind of forced me to grow up real fast. It's not enough just to study anymore. I can study and ace all the exams — oh, that's just not enough anymore. You have to figure out life real quick.
Dr. Ashley: Yeah, for a 15-year-old, that's a challenge. So then you graduated —
Dr. Chen: I graduated from UCLA. I was in all the pre-med clubs, I took the MCAT —
Dr. Ashley: But you were 16 or something when you graduated?
Dr. Chen: I was a little older than that. I was a ripe 19 by the time I was out of there. But then kind of at the last minute, I hear all this: well, you're going to be a doctor, you've got to have a passion for healing, it's got to be like your life's mission. And I'm like, I'm 18, 19 years old — I feel nothing. Okay, I'm a kid basically. I'm not ready to commit my life to this. I need to go have some other experiences. And my real underlying nerd passion — I'm just a nerd, I just like to study, I like to learn things — and I'm a computer nerd, right? I like to program. So I actually worked as a software engineer, software development, nothing to do with biology or medicine, for a couple of years. The dot-com bubble was bursting around that era.
Dr. Ashley: And what were you programming back then? What kind of computer were you using? What language?
Dr. Chen: Gosh, I mean I was learning C++ in class, but most of industry at the time was Java programming, and then the internet had become a thing — making three-tiered applications, HTML, JavaScript, Java at the time. Now it would seem antiquated; things have evolved. But that gave me a lot of perspective on industry. Oh wow, suddenly when I can make six figures as a 19-year-old, my mom didn't care that much when I went to medical school. But then after a year or so, the company laid off a third of its staff during the dot-com bubble burst, including myself. I got another job — it was fine. But it gave me more perspective: well, this is a good job and I'm good at it, but I could foresee I'm not sure I'm going to find a rewarding long-term career I'm going to be happy with.
Dr. Ashley: Were you from a technical or academic or medical family?
Dr. Chen: Really not. I'm sort of the mom's aspiration — the first doctor in the family kind of situation. My dad was an electrical engineer with a computer science master's. That's where I get a lot of the nerdiness from. But this combination was very weird. Then an old friend said, "Why don't you consider medical school again?" I'm like, "What?" I'd left that all behind. I'm not going to go back; I have to go back and beg my old professors for a letter of recommendation. Forget it. It was my girlfriend at the time. She said, "No, Jonathan, you should totally apply for med school. You should do it because in life, you will regret what you didn't try, not what you did try." Once I heard that — putting the guilt trip on me — I was like, well, now I have to. But I said I'm going to apply for MD-PhD programs only, and the PhD has to be in computer science. That does sound crazy, but that's the point. If I'm going to do it, it has to be crazy enough to be something different. Otherwise, I might as well just stay in my job.
Dr. Ashley: MD-PhD with computer science — a fair number of places do that, but not many.
Dr. Chen: I give credit to University of California, Irvine. They were very open-minded to that combination. They actually literally had another MSTP — Medical Scientist Training Program — student a couple of years ahead of me doing that same combo. Whereas most places would look at me strange: "Shouldn't you be getting a PhD in biochemistry, immunology, neuroscience? Why computer science?" And this was around 2000, 2001, 2002.
Dr. Ashley: Yes. The internet had happened, you were programming in Java, JavaScript — but we were really just maturing in our sense of what the internet was, just after the crash had started. Okay. So then you headed to medical school.
Dr. Chen: I ended up doing the joint degree program. I got the PhD in computer science. What I was working on would nowadays be called rules-based expert systems and machine learning applications for chemistry — specifically organic chemistry, which was my subject domain. A lesson I took from my industry experience is that you need strong technical skills, otherwise it's all just talk. Medicine is really the subject domain I care most about now, but at the time I said, well, chemistry — I think that's fun and interesting. In so many words, I made AI systems that could do your organic chemistry homework for you. The irony is, if it can do your homework for you and help with pharmaceutical development, it could also teach you how to do your homework. Not what I expected, but one of the main outputs of my PhD was an education system that was used for over a decade by students around the world to learn organic chemistry.
Dr. Ashley: My daughter is currently suffering through it in college. So then you did your clinical work and became a doctor.
Dr. Chen: I actually did not intend to do residency. I was never intending to be a practicing doctor. I was like, I'll just get the knowledge, I'll graduate, maybe do consulting — I'm going to go back to industry in some form or another. And then I did my clerkships — my third-year internal medicine rotation — and it surprised me how much I liked it. I did not expect to like it. I was like, oh, you're going to be at the bottom of the totem pole, have sick people complain to you all the time — who wants to deal with this? And then I got there and it was like: this is actually really cool. You work in a team, you're working on good problems together, it's something that matters. What you do has very clear impact. People are hanging on your words, you've got to do it right. There's a very interesting applied expertise that has to manifest there. It doesn't matter what the abstract says. Either you prescribe the medicine or you don't — how do you make that kind of tough decision? I thought that was very compelling. So I did decide to do residency, and that's how I first arrived at Stanford. Third time was a charm — undergrad didn't quite work out, med school and PhD, and finally internal medicine residency. Very grateful to have matched at Stanford. I've been here since.
Dr. Ashley: We've been very happy to have you here in the Department of Medicine, part of our newly named Division of Computational Medicine. You have a number of different roles now, but I think it would be great to start with some of the really impactful work that's been covered in the press around the world — thinking about the application of modern-day AI architectures to medicine. In particular, maybe the thing that hit the biggest headline was this idea that when you compared human doctors plus AI tools to the AI tool alone, the AI tool appeared to outperform — the doctor was actually pulling down the performance. Talk about that work and some of the caveats and what it means for us.
Dr. Chen: That was definitely not what we expected. When I first saw a preview version of GPT-4 as things were building up, I was like, "Holy smokes." I literally felt like I need to throw away half of my research program. This thing had leapfrogged so many things I thought were capable, and it moved much faster than I expected. But then empirically — I didn't even bother checking multiple-choice questions because I know 100 people do that; it's so easy, and that's not what we as doctors would care about. Complex case reasoning with expert consensus grading — very hard to do, but that's the key. Our educators are key to unlocking that, and the result was not what we expected. We thought our hypothesis was: oh, you could look up stuff in UpToDate, PubMed, or use this AI tool that seems really smart, and I bet doctors could be even better — we can show this combination is so great. And that's not what we found. We found the combination didn't make that much difference. Very surprising to us.
Dr. Ashley: And these were doctors who were already somewhat familiar with the tools.
Dr. Chen: At the time we did that study — about two and a half years ago — a third of the doctors had never touched ChatGPT or a chatbot before in their life, and another third had maybe used it once or twice. It was clear many of them didn't know what it was and didn't know how to use it, or if they did, they did not trust it. And I agree — the first time you chat with an AI chatbot, it felt weird. You didn't know what to ask, you didn't know what it could say. It was a very bizarre feeling.
Dr. Ashley: For decades in computer science, this idea of the Turing test from the 1950s — and suddenly we leapfrogged past it almost overnight. It became almost irrelevant. For medicine, where it's been clearly one of the major applications — with the internet, it wasn't immediately obvious that health was one of the most interesting applications. With the language models we're seeing now, it really is to the forefront. We don't see patients anymore who haven't already consulted a language model. The idea two years later that there would be a doctor who doesn't know how to use a language model — that's not very credible anymore. But what exactly do you think it was — that naivety about how to use the system, or do you genuinely think the human was dragging down the reasoning within the model?
Dr. Chen: There is anchoring and sycophancy that happens. If you say, "Hey AI, I've got this patient with shortness of breath — do you think it's congestive heart failure?" It'll probably just say, "Sure, good idea, doctor. Congestive heart failure sounds great." When maybe it's not giving you the objective assessment it could have. We've empirically shown that many times.
Dr. Ashley: Why does it do that?
Dr. Chen: At the very base level, I like to call them autocomplete on steroids. It's just read the internet, in so many words, and can guess the next word, filling in a sentence. But if it just did that, it's not very interesting to work with. So they did supervised fine-tuning, reinforcement learning from human feedback. What does that mean? It'll output 10 different answers, and humans said, "I like this answer better — this is answering my question better, this one is more coherent." So it gives more coherent answers. But that also means humans are instilling their values of what they like. And humans really like it when people — or computers — agree with them, whether it's right or not.
Dr. Ashley: So this leads the model into a part of its space where it not only wants to please, but it doesn't fully separate the facts it needs to provide from its sense of wanting to please, which was instilled through that reinforcement learning feedback cycle. This is something you've advised — a very practical piece of advice is to let the model operate first without giving it a pre-baked probability of what the answer might be, without biasing it with your own idea.
Dr. Chen: And for what it's worth, a lot of these concepts aren't really new — humans do the same thing. If you get an admission from the emergency room and they tell you, "Hey, this patient has systolic heart failure exacerbation," well, maybe 80 to 90% of the time that's correct, but sometimes — wait, no, this patient has pneumonia. We anchored on the wrong thing. The other danger with AI is automation bias. It's right so often that you eventually stop double-checking all the time — and that gets you in a bad situation when it really matters.
Dr. Ashley: I find this with the digital scribe we have available in the clinic. It is so articulate, so well-written — great punctuation, capital letters in the right place, commas, em dashes everywhere — it speaks so well that our brains, which are used to reading text that has been checked by humans, associate that with quality. And we have to revisit that now.
Dr. Chen: Absolutely. Confabulations are photorealistic, the prose is very eloquent — those are no longer reliable indicators of truth anymore. Your judgment matters even more than it did before. And beyond accidental errors, there are bad actors out there who are straight-up trying to scam you with things that look very credible. It's actually a kind of scary world we're emerging into, and I don't know that we fully know how to cope with it yet.
Dr. Ashley: You also presented some really interesting data around how many of these models have been tested on benchmarks where there's a right answer and a wrong answer — multiple choice, not very real-world. But in the real world, we have a rubric in medicine: first, do no harm. You've been exploring this with your fantastic team here.
Dr. Chen: What were you actually trying to do at first? We were prototyping an AI-provided consult — like, what would that be like? But before we unleash that in the real world, could we just make sure it's at least safe? Let's get some representative cases and make sure the thing doesn't give answers that would actually harm somebody, let alone whether it's even useful. And that's not a trivial question to answer. In this study, which is just emerging as a preprint right now, accuracy and harm are not the same thing. You can have a really bright student or resident who knows everything, and sometimes you go — whoa, whoa, whoa, that's dangerous. Don't do that. And it's not because they're not smart — it's, oh, was I not supposed to do that? Was that important? They have no basis to make that distinction. That required a very deliberate type of evaluation, and it showed harm was not correlated with accuracy alone.
Dr. Ashley: And we probably should subordinate accuracy to "do no harm."
Dr. Chen: It's kind of like a clinical trial progression — before you check efficacy, can we just make sure we're not accidentally harming people? And what's been surprising us: for these kinds of tough questions, 10 to 20% of the time, even frontier models will sometimes cause harm — and often the harm was that they didn't do something. They didn't recommend something that was actually important to address.
Dr. Ashley: We often miss sins of omission because we're so focused on sins of commission.
Dr. Chen: Doing nothing doesn't mean doing no harm. You actually do a lot of harm when you should have done something and you saw it.
Dr. Ashley: Another area worth exploring briefly is what falls under the term RAG — retrieval-augmented generation. We're moving from a world where these models are essentially next-token predictors built on a huge mass of data, trying to please you — all the things we talked about — to a world where they're much more likely to go to a source and use their language skills to bring back data directly from that source. Talk about that evolution, specifically for medical AI. It's a great migration.
Dr. Chen: I used to talk about confabulation, hallucination — this thing has Wernicke-Korsakoff syndrome; it's just making up words even if they don't make sense. It's not trying to lie to you. It doesn't know what a lie is. It's just filling in the blanks as it goes. I've decided not to cover that as much anymore because it's not a solved problem, but it's much less of an issue, because most models have deliberate RAG — retrieval-augmented generation — baked in, where it goes: well, you're asking a factual or evidence-specific question, let me go look for that evidence and kind of read a document for you. That's a much more grounded approach. You still can't rely on it perfectly, but you also don't have to just guess — you can go double-check the article yourself. As clinicians, we're actually in a very good position to deal with this. Just like in real life, your consultant gives you a lot of advice on the wards, and most of the time they're right. But sometimes — wait a minute, it really matters if we go to surgery or not. So this one is worth double-checking. Whether the antibiotics are for five or seven days — well, it's probably about the same, I'm not going to worry too much about it. You know when it matters and you know how to dig in deep when it does.
Dr. Ashley: And the link is right there — you can go directly from a model summary to the source. For medical decisions, that's how we mostly use the frontier models at this point. Our residents and colleagues tend towards using Open Evidence, which is a specific language model built from medical data with doctors in mind — although most patients use the frontier models.
Dr. Chen: Open Evidence requires an NPI number to get in; they're trying to restrict it. Sometimes models are restricted to providers, probably partly for liability reasons, because otherwise these models are basically giving medical advice. They have this disclaimer — "not for medical advice, cannot be used for any medical purpose" — but give me a break. That's clearly what people are doing. It creates a very odd tension about who's really responsible for the decisions being made.
Dr. Ashley: We've obviously been very quick to embrace the potential benefits of AI here at Stanford, including in our healthcare system. You mentioned the engineering team that visited rounds with me in the hospital last week.
Dr. Chen: That's right —
Dr. Ashley: Which was really fun! The tool allows a language model, under closed and safe conditions, to access a single patient's record and use these real strengths — assimilating data, going to raw data, and bringing back answers — but in the context of a single individual. Maybe some listeners don't realize how much time our residents spend clicking through different screens in the electronic health record. Talk about the problem first, then the potential solution.
Dr. Chen: For better or worse, so much of our practice — the electronic medical record is the central hub for what happens in medicine. The first time I thought about ChatGPT, the first thing I thought was: I want that attached to the medical record. But for all sorts of privacy, security, and engineering reasons, it's actually very difficult to do. But for a couple of years, teams have been putting this together, and now we have at least the first forms of it. Very powerful — because most of a clinician's time on the computer isn't writing notes or putting in orders. Most of their time is chart review. They're looking stuff up, reading notes, making sense of results, collating. And then at the last minute they summarize it.
Dr. Ashley: Between each piece of information, there might be five or six clicks.
Dr. Chen: I was recently reviewing a medical case record — they sent me 5,000 pages of documents, and most of those 4,900 pages were worthless. Just sifting through and making sense of it was hours and hours of work. That's a very common job we give medical students first — why don't you scour those records and summarize what's happened. And now — not perfect, not exactly ready for prime time, but you can see where it's going — AI can do
a lot of that, and we can get back to: okay, thank you for the summary, now I can actually think about how to approach this patient's case. And it can catch things humans didn't. My wife is a pathologist and she's doing a prototype there — it caught things the humans missed. "Has this patient ever had a hematologic malignancy?" The person who scoured the records said, "I don't see it." But wait — there was a note from nine months ago. That totally changes the interpretation.
Dr. Ashley: And I think the deep learning era for AI — where we were used to the idea that imaging could come under the spell of AI — now we're seeing that really all the data, anything to do with reading and writing, which is basically everything, is in scope.
Dr. Chen: Absolutely. That's why when I saw the emerging GPT-3 and GPT-4, I was like, holy smokes — this is actually going to change the world. For a lot of stuff — my most cited paper, for better or worse, is that machine learning is way overhyped in medicine. That's a two-page perspective, and it's the most cited thing I have. But here I'm like, okay, there is hype too, and people think AI is magic a little bit too much — marketers basically apply that word to anything. But here, I think this is real.
Dr. Ashley: Is that your vision for the next few years? I know this is something we both get asked a lot — these models are only going to get better from here. All of this really started with the transformer paper in 2017, but practically, GPT-3 — that was 2021, 2022. Three years ago. So it feels almost ridiculous to ask about five years from now.
Dr. Chen: Gosh, that is tough. Five years in this kind of epoch — look at where we were five years ago. We weren't even talking about this. Within five years, I'm pretty sure most doctors will be using ambient scribes, or at least that'll be very common, routine technology. It won't even seem novel anymore. Other predictions? I'm surprised there's not a class action lawsuit against big tech over AI harming people — giving bad medical advice or harmful things.
Dr. Ashley: There are class action lawsuits around creative content — use of movie actors' likenesses, publishing and Hollywood — but not yet for medical harm.
Dr. Chen: I'm really kind of surprised it hasn't happened yet. More and more, I think AI is just going to be embedded in our systems. Your kids will still know a little bit of the difference — you were there before the internet age as I was — but can your kids comprehend a world without the internet?
Dr. Ashley: No, they cannot.
Dr. Chen: It'll get to the point where it just seems routine. Anytime you read or write anything, which is basically anytime you interact with a computer, AI is going to be right over your shoulder. And it'll really abstract the nature of the way we interact with all things.
Dr. Ashley: And I think it'll get safer as well.
Dr. Chen: It has to, and it will — but it's a process to get there.
Dr. Ashley: So you recently took on a new position as Director of AI Education. We're obviously an educational facility — school of medicine, medical students, PA students, nurses and other trainees, residency programs, fellows. We're training a whole generation of doctors for the future, as well as providing continuing medical education. How do you even go about planning a curriculum around AI?
Dr. Chen: It's a chance — Senior Associate Dean Reena Thomas tapped me for this one. Dr. Thomas said, "Jonathan, you're not here for advice. You're here to apply for this job." And it's great. I think this is actually a very nice coalescence of so many themes that actually fits my history. I kind of couldn't help it — organic chemistry education, I kind of landed there. I have a fortune cookie from when I first started my job that said you could do well in education — like, no, I'm trying so hard not to be an educator!
Dr. Ashley: But you're a Stanford professor, so you're by definition an educator.
Dr. Chen: And I realize I kind of can't help it — whenever I talk to a young person, I start teaching something. And I knew it's not just myself. I found a crew — many clinical informatics fellows, former and current. I have associate directors: Dong Yao, Shivani Vadok, my current fellow Idan Zahash, and multiple others. There's a whole crew assembled for whom education is their passion, and I can give the outline, but they've done the curricular design, outlined things, reviewed the literature, established a framework, and reused a lot of resources. One of the best contributions actually came from a first-year student — Mitra Alkani. She gave one of the best presentations at our first-ever AI and Medical Education Symposium this past June. Her perspective was: she's just a student. She's not here to do research or science. She said, "Hey, these are the three tools I figured out, and nobody told me. I just tried them myself — they help me make flashcards, they help me practice interviewing." I thought that was very compelling. She showed: I'm a student, and this is how I help myself learn in ways that weren't possible before.
Dr. Ashley: And what are the principles or things you worry most about, or want to make sure you're passing on through the curriculum?
Dr. Chen: Oh gosh, there are some real dilemmas I don't have easy answers for. For example, should students be allowed to use AI on their homework? The default answer — and actually Stanford University's default answer — is assume it's disallowed unless your instructor says so. But I push back: that's hopeless, it cannot be the default policy, because it's unenforceable. It puts students in an awkward, unfair position where the honor code says they have to follow the rules, but they know all their classmates are doing it. So we've shifted toward: homework is just practice — we're not even going to bother grading it because it's so easy for AI to do. It's got to be the closed-book exams. But there are some real tensions. Some of our medical student classes were killing it on the homework for medical reasoning — doing great. But when it came to the closed-book exams, they did not do very well. Much worse than in prior years. What happened? It's obvious. They used AI to do their homework and never bothered to actually learn. AI could be the best teacher you have ever had, but if you misuse it, you miss the point. The point of homework isn't to do the homework — it's to make you struggle. And that struggle is actually what makes you learn and makes this knowledge innate.
Dr. Ashley: We often use the calculator example. We don't force kids to do long division, but maybe they should still learn it — so they have the ability, and then the calculator amplifies that. But there are fundamental medical skills where, if you start out using these models at the level they're at — which is beyond what any calculator could do — the metaphor breaks down. If you never learn those skills to begin with, you really do lose something.
Dr. Chen: We looked at other schools' policies, and I like the principle: you can use these tools once you've demonstrated you could have done it on your own. It's a great principle; it's just hard to enforce. Would you allow an intern or an MS3 to use an ambient scribe to write their notes? How will they ever know how to write a plan if AI always does it for them? That's quite a tension. Ultimately, we want everyone to have judgment — let the AI remember everything, who cares. But you can't have judgment about something if you've never learned the underlying knowledge.
Dr. Ashley: I also worry because these models are trained on text created by humans, which came from science and human experience. If all our generated text ends up being AI-generated, there's a feedback loop —
Dr. Chen: AI slop, the snake eating its own tail. This is already happening. A lot of the tech companies building frontier models would say they're not looking for more data at this point; they're trying to get higher-quality text, which is why many newspapers are suing them. "Yeah, we know you're using our work, and we know you like it because we produce high-quality writing." I suspect what will happen is text and the way we talk and write — without even realizing it — is going to become more and more homogenized. You can detect it now. When somebody writes you an AI-generated note, you can kind of tell. But how many times did you not notice?
Dr. Ashley: There's actually a whole Wikipedia page dedicated to tells of language models. Some are obvious — the word "delve," the word "critical," the word "deep,"
Dr. Chen: And the notorious em dash.
Dr. Ashley: I used to love em dashes. Now I actively delete every one I see. I have to stop using them. It's like — is this AI or is this Dr. Ashley? I'm not sure anymore. I feel personally sad about that. But it's a challenge. With so many places where we as professors or attendings are asked to generate text, if these become shortcuts, we end up in a world where everything is generated.
Dr. Chen: It will certainly homogenize our style of reading and writing. Maybe it abstracts us to the level where we can get more to the actual thinking and meaning. It's not a perfect analogy, but think about programming languages. Who wants to write bytecode? That's crazy. So we got assembly language, then Pascal or FORTRAN, then C++, then Python. Now it's: just write in English and have AI generate the lower-level code for me. That's a great abstraction. But I think when it comes to treading on the domain of normal human interactions, we haven't had to deal with that before.
Dr. Ashley: That really is a bit different. You began your talk with one of my favorite quotes: any sufficiently advanced technology is indistinguishable from magic. So let's get back to magic — because you are the only person any of us know who provides lectures and talks that include both up-to-date AI research data and actual magic tricks. How did you get back into magic?
Dr. Chen: I picked it up again maybe five or six years ago, just before the pandemic. Actually, it was a pandemic thing. My oldest child — he was probably eight or nine — we went to just some street magic show, and somebody pulled a rabbit out of a box, and his face just lit up, and he squealed. If you see a child really experience that wonder, it's a really magical thing — not the magic trick itself, but seeing somebody have that experience. So I thought, oh, my kid enjoys magic; I should show him a magic trick. I bought him a little basic kid set. And as I was showing him and my colleague a trick, my colleague looked over and said, "Jonathan should work on his sleight of hand." I was like — what? I'm just trying to entertain your child and you're giving me feedback. So I went and learned some sleight-of-hand magic. And it was a fun way to interact with students — look, I'm in a nerdy profession, but I'm approachable too. It kind of spiraled out of control during the pandemic. We're all locked indoors; some people learned to bake sourdough bread. I literally spent three months learning a very advanced Rubik's cube magic trick. And it spiraled — I've entered and won multiple competitions; I've had paid gigs just to perform magic. I performed in Vegas — on the main stage in Vegas, actually, at a health conference. Not something I was trying to do, but it's unearthing some childhood aspirations and inner child. It also helps me be better at presentations. You learn a lot of empathy when you do magic because it's all about — who cares what I'm doing; I have to anticipate what you're thinking while I do this, because I don't want you thinking the wrong thing. You have to really understand what another person is thinking. Directing a narrative, setting up expectations — I found that a very powerful combination. A few years ago, University Medical Partners invited me to give a keynote on AI and medicine. But they also had a "magic of medicine" theme that week. I thought, actually, I can perform some magic too — can I do that as a bonus? And I blame it on my wife. She was like, "Hey, why not do them together? What if you had some magic in the middle of your talk?" Like, that's crazy — they're going to laugh me off the stage. But if there's a thematic connection, that could be very fun and compelling. Especially as generative AI was emerging — you can't tell what's real anymore. The Turing test. Is that a real human or a chatbot? That image looks so real. That video of me — that was all AI-generated. And so I'm using magic again as that algorithm. Boy, does that look real — but you still have to have the judgment to tell the difference.
Dr. Ashley: It's a really powerful combination. Jonathan, we're so happy to have you in our department. I'm very happy to be here. We're proud of the work you've done in an academic way and for really bringing us safely into this future — and of course for your role in education and training the next generation. Thanks for joining us on the Future of Medicine.
Dr. Chen: Thank you, Dr. Ashley.
