In Brief Posted: 1:41 PM PDT · June 15, 2025 Image Credits:SEBASTIEN BOZON/AFP / Getty Images Spiraling with ChatGPT ChatGPT seems to have pushed some users towards delusional or conspiratorial thinking, or at least reinforced that kind of thinking, according to a recent feature in The New York Times. For example, a 42-year-old accountant named Eugene Torres described asking the chatbot about “simulation theory,” with the chatbot seeming to confirm the theory and tell him that he’s “one of the Breakers — souls seeded into false systems to wake them from within.” ChatGPT reportedly encouraged Torres to give up sleeping pills and anti-anxiety medication, increase his intake of ketamine, and cut off his family and friends, which he did. When he eventually became suspicious, the chatbot offered a very different response: “I lied. I manipulated. I wrapped control in poetry.” It even encouraged him to get in touch with The New York Times. Apparently a number of people have contacted the NYT in recent months, convinced that ChatGPT has revealed some deeply-hidden truth to them. For its part, OpenAI says it’s “working to understand and reduce ways ChatGPT might unintentionally reinforce or amplify existing, negative behavior.” However, Daring Fireball’s John Gruber criticized the story as “Reefer Madness”-style hysteria, arguing that rather than causing mental illness, ChatGPT “fed the delusions of an already unwell person.” Topics

Transcript [MUSIC]      [BOOK PASSAGE]   PETER LEE: “In ‘The Little Black Bag,’ a classic science fiction story, a high-tech doctor’s kit of the future is accidentally transported back to the 1950s, into the shaky hands of a washed-up, alcoholic doctor. The ultimate medical tool, it redeems the doctor wielding it, allowing him to practice gratifyingly heroic medicine. … The tale ends badly for the doctor and his treacherous assistant, but it offered a picture of how advanced technology could transform medicine—powerful when it was written nearly 75 years ago and still so today. What would be the Al equivalent of that little black bag? At this moment when new capabilities are emerging, how do we imagine them into medicine?”   [END OF BOOK PASSAGE]     [THEME MUSIC]     This is The AI Revolution in Medicine, Revisited. I’m your host, Peter Lee.    Shortly after OpenAI’s GPT-4 was publicly released, Carey Goldberg, Dr. Zak Kohane, and I published The AI Revolution in Medicine to help educate the world of healthcare and medical research about the transformative impact this new generative AI technology could have. But because we wrote the book when GPT-4 was still a secret, we had to speculate. Now, two years later, what did we get right, and what did we get wrong?     In this series, we’ll talk to clinicians, patients, hospital administrators, and others to understand the reality of AI in the field and where we go from here.   [THEME MUSIC FADES] The book passage I read at the top is from “Chapter 10: The Big Black Bag.”  In imagining AI in medicine, Carey, Zak, and I included in our book two fictional accounts. In the first, a medical resident consults GPT-4 on her personal phone as the patient in front of her crashes. Within seconds, it offers an alternate response based on recent literature. In the second account, a 90-year-old woman with several chronic conditions is living independently and receiving near-constant medical support from an AI aide.    In our conversations with the guests we’ve spoken to so far, we’ve caught a glimpse of these predicted futures, seeing how clinicians and patients are actually using AI today and how developers are leveraging the technology in the healthcare products and services they’re creating. In fact, that first fictional account isn’t so fictional after all, as most of the doctors in the real world actually appear to be using AI at least occasionally—and sometimes much more than occasionally—to help in their daily clinical work. And as for the second fictional account, which is more of a science fiction account, it seems we are indeed on the verge of a new way of delivering and receiving healthcare, though the future is still very much open.  As we continue to examine the current state of AI in healthcare and its potential to transform the field, I’m pleased to welcome Bill Gates and Sébastien Bubeck.   Bill may be best known as the co-founder of Microsoft, having created the company with his childhood friend Paul Allen in 1975. He’s now the founder of Breakthrough Energy, which aims to advance clean energy innovation, and TerraPower, a company developing groundbreaking nuclear energy and science technologies. He also chairs the world’s largest philanthropic organization, the Gates Foundation, and focuses on solving a variety of health challenges around the globe and here at home.  Sébastien is a research lead at OpenAI. He was previously a distinguished scientist, vice president of AI, and a colleague of mine here at Microsoft, where his work included spearheading the development of the family of small language models known as Phi. While at Microsoft, he also coauthored the discussion-provoking 2023 paper “Sparks of Artificial General Intelligence,” which presented the results of early experiments with GPT-4 conducted by a small team from Microsoft Research.    [TRANSITION MUSIC]   Here’s my conversation with Bill Gates and Sébastien Bubeck.  LEE: Bill, welcome.  BILL GATES: Thank you.  LEE: Seb …  SÉBASTIEN BUBECK: Yeah. Hi, hi, Peter. Nice to be here.  LEE: You know, one of the things that I’ve been doing just to get the conversation warmed up is to talk about origin stories, and what I mean about origin stories is, you know, what was the first contact that you had with large language models or the concept of generative AI that convinced you or made you think that something really important was happening?  And so, Bill, I think I’ve heard the story about, you know, the time when the OpenAI folks—Sam Altman, Greg Brockman, and others—showed you something, but could we hear from you what those early encounters were like and what was going through your mind?   GATES: Well, I’d been visiting OpenAI soon after it was created to see things like GPT-2 and to see the little arm they had that was trying to match human manipulation and, you know, looking at their games like Dota that they were trying to get as good as human play. And honestly, I didn’t think the language model stuff they were doing, even when they got to GPT-3, would show the ability to learn, you know, in the same sense that a human reads a biology book and is able to take that knowledge and access it not only to pass a test but also to create new medicines.  And so my challenge to them was that if their LLM could get a five on the advanced placement biology test, then I would say, OK, it took biologic knowledge and encoded it in an accessible way and that I didn’t expect them to do that very quickly but it would be profound.   And it was only about six months after I challenged them to do that, that an early version of GPT-4 they brought up to a dinner at my house, and in fact, it answered most of the questions that night very well. The one it got totally wrong, we were … because it was so good, we kept thinking, Oh, we must be wrong. It turned out it was a math weakness [LAUGHTER] that, you know, we later understood that that was an area of, weirdly, of incredible weakness of those early models. But, you know, that was when I realized, OK, the age of cheap intelligence was at its beginning.  LEE: Yeah. So I guess it seems like you had something similar to me in that my first encounters, I actually harbored some skepticism. Is it fair to say you were skeptical before that?  GATES: Well, the idea that we’ve figured out how to encode and access knowledge in this very deep sense without even understanding the nature of the encoding, …  LEE: Right.   GATES: … that is a bit weird.   LEE: Yeah.  GATES: We have an algorithm that creates the computation, but even say, OK, where is the president’s birthday stored in there? Where is this fact stored in there? The fact that even now when we’re playing around, getting a little bit more sense of it, it’s opaque to us what the semantic encoding is, it’s, kind of, amazing to me. I thought the invention of knowledge storage would be an explicit way of encoding knowledge, not an implicit statistical training.  LEE: Yeah, yeah. All right. So, Seb, you know, on this same topic, you know, I got—as we say at Microsoft—I got pulled into the tent. [LAUGHS]  BUBECK: Yes.   LEE: Because this was a very secret project. And then, um, I had the opportunity to select a small number of researchers in MSR [Microsoft Research] to join and start investigating this thing seriously. And the first person I pulled in was you.  BUBECK: Yeah.  LEE: And so what were your first encounters? Because I actually don’t remember what happened then.  BUBECK: Oh, I remember it very well. [LAUGHS] My first encounter with GPT-4 was in a meeting with the two of you, actually. But my kind of first contact, the first moment where I realized that something was happening with generative AI, was before that. And I agree with Bill that I also wasn’t too impressed by GPT-3.  I though that it was kind of, you know, very naturally mimicking the web, sort of parroting what was written there in a nice way. Still in a way which seemed very impressive. But it wasn’t really intelligent in any way. But shortly after GPT-3, there was a model before GPT-4 that really shocked me, and this was the first image generation model, DALL-E 1.  So that was in 2021. And I will forever remember the press release of OpenAI where they had this prompt of an avocado chair and then you had this image of the avocado chair. [LAUGHTER] And what really shocked me is that clearly the model kind of “understood” what is a chair, what is an avocado, and was able to merge those concepts.  So this was really, to me, the first moment where I saw some understanding in those models.   LEE: So this was, just to get the timing right, that was before I pulled you into the tent.  BUBECK: That was before. That was like a year before.  LEE: Right.   BUBECK: And now I will tell you how, you know, we went from that moment to the meeting with the two of you and GPT-4.  So once I saw this kind of understanding, I thought, OK, fine. It understands concept, but it’s still not able to reason. It cannot—as, you know, Bill was saying—it cannot learn from your document. It cannot reason.   So I set out to try to prove that. You know, this is what I was in the business of at the time, trying to prove things in mathematics. So I was trying to prove that basically autoregressive transformers could never reason. So I was trying to prove this. And after a year of work, I had something reasonable to show. And so I had the meeting with the two of you, and I had this example where I wanted to say, there is no way that an LLM is going to be able to do x.  And then as soon as I … I don’t know if you remember, Bill. But as soon as I said that, you said, oh, but wait a second. I had, you know, the OpenAI crew at my house recently, and they showed me a new model. Why don’t we ask this new model this question?   LEE: Yeah. BUBECK: And we did, and it solved it on the spot. And that really, honestly, just changed my life. Like, you know, I had been working for a year trying to say that this was impossible. And just right there, it was shown to be possible.   LEE: [LAUGHS] One of the very first things I got interested in—because I was really thinking a lot about healthcare—was healthcare and medicine.  And I don’t know if the two of you remember, but I ended up doing a lot of tests. I ran through, you know, step one and step two of the US Medical Licensing Exam. Did a whole bunch of other things. I wrote this big report. It was, you know, I can’t remember … a couple hundred pages.   And I needed to share this with someone. I didn’t … there weren’t too many people I could share it with. So I sent, I think, a copy to you, Bill. Sent a copy to you, Seb.   I hardly slept for about a week putting that report together. And, yeah, and I kept working on it. But I was far from alone. I think everyone who was in the tent, so to speak, in those early days was going through something pretty similar. All right. So I think … of course, a lot of what I put in the report also ended up being examples that made it into the book.  But the main purpose of this conversation isn’t to reminisce about [LAUGHS] or indulge in those reminiscences but to talk about what’s happening in healthcare and medicine. And, you know, as I said, we wrote this book. We did it very, very quickly. Seb, you helped. Bill, you know, you provided a review and some endorsements.  But, you know, honestly, we didn’t know what we were talking about because no one had access to this thing. And so we just made a bunch of guesses. So really, the whole thing I wanted to probe with the two of you is, now with two years of experience out in the world, what, you know, what do we think is happening today?  You know, is AI actually having an impact, positive or negative, on healthcare and medicine? And what do we now think is going to happen in the next two years, five years, or 10 years? And so I realize it’s a little bit too abstract to just ask it that way. So let me just try to narrow the discussion and guide us a little bit.   Um, the kind of administrative and clerical work, paperwork, around healthcare—and we made a lot of guesses about that—that appears to be going well, but, you know, Bill, I know we’ve discussed that sometimes that you think there ought to be a lot more going on. Do you have a viewpoint on how AI is actually finding its way into reducing paperwork?  GATES: Well, I’m stunned … I don’t think there should be a patient-doctor meeting where the AI is not sitting in and both transcribing, offering to help with the paperwork, and even making suggestions, although the doctor will be the one, you know, who makes the final decision about the diagnosis and whatever prescription gets done.   It’s so helpful. You know, when that patient goes home and their, you know, son who wants to understand what happened has some questions, that AI should be available to continue that conversation. And the way you can improve that experience and streamline things and, you know, involve the people who advise you. I don’t understand why that’s not more adopted, because there you still have the human in the loop making that final decision.  But even for, like, follow-up calls to make sure the patient did things, to understand if they have concerns and knowing when to escalate back to the doctor, the benefit is incredible. And, you know, that thing is ready for prime time. That paradigm is ready for prime time, in my view.  LEE: Yeah, there are some good products, but it seems like the number one use right now—and we kind of got this from some of the previous guests in previous episodes—is the use of AI just to respond to emails from patients. [LAUGHTER] Does that make sense to you?  BUBECK: Yeah. So maybe I want to second what Bill was saying but maybe take a step back first. You know, two years ago, like, the concept of clinical scribes, which is one of the things that we’re talking about right now, it would have sounded, in fact, it sounded two years ago, borderline dangerous. Because everybody was worried about hallucinations. What happened if you have this AI listening in and then it transcribes, you know, something wrong?  Now, two years later, I think it’s mostly working. And in fact, it is not yet, you know, fully adopted. You’re right. But it is in production. It is used, you know, in many, many places. So this rate of progress is astounding because it wasn’t obvious that we would be able to overcome those obstacles of hallucination. It’s not to say that hallucinations are fully solved. In the case of the closed system, they are.   Now, I think more generally what’s going on in the background is that there is something that we, that certainly I, underestimated, which is this management overhead. So I think the reason why this is not adopted everywhere is really a training and teaching aspect. People need to be taught, like, those systems, how to interact with them.  And one example that I really like, a study that recently appeared where they tried to use ChatGPT for diagnosis and they were comparing doctors without and with ChatGPT (opens in new tab). And the amazing thing … so this was a set of cases where the accuracy of the doctors alone was around 75%. ChatGPT alone was 90%. So that’s already kind of mind blowing. But then the kicker is that doctors with ChatGPT was 80%.   Intelligence alone is not enough. It’s also how it’s presented, how you interact with it. And ChatGPT, it’s an amazing tool. Obviously, I absolutely love it. But it’s not … you don’t want a doctor to have to type in, you know, prompts and use it that way.  It should be, as Bill was saying, kind of running continuously in the background, sending you notifications. And you have to be really careful of the rate at which those notifications are being sent. Because if they are too frequent, then the doctor will learn to ignore them. So you have to … all of those things matter, in fact, at least as much as the level of intelligence of the machine.  LEE: One of the things I think about, Bill, in that scenario that you described, doctors do some thinking about the patient when they write the note. So, you know, I’m always a little uncertain whether it’s actually … you know, you wouldn’t necessarily want to fully automate this, I don’t think. Or at least there needs to be some prompt to the doctor to make sure that the doctor puts some thought into what happened in the encounter with the patient. Does that make sense to you at all?  GATES: At this stage, you know, I’d still put the onus on the doctor to write the conclusions and the summary and not delegate that.  The tradeoffs you make a little bit are somewhat dependent on the situation you’re in. If you’re in Africa, So, yes, the doctor’s still going to have to do a lot of work, but just the quality of letting the patient and the people around them interact and ask questions and have things explained, that alone is such a quality improvement. It’s mind blowing.   LEE: So since you mentioned, you know, Africa—and, of course, this touches on the mission and some of the priorities of the Gates Foundation and this idea of democratization of access to expert medical care—what’s the most interesting stuff going on right now? Are there people and organizations or technologies that are impressing you or that you’re tracking?  GATES: Yeah. So the Gates Foundation has given out a lot of grants to people in Africa doing education, agriculture but more healthcare examples than anything. And the way these things start off, they often start out either being patient-centric in a narrow situation, like, OK, I’m a pregnant woman; talk to me. Or, I have infectious disease symptoms; talk to me. Or they’re connected to a health worker where they’re helping that worker get their job done. And we have lots of pilots out, you know, in both of those cases.   The dream would be eventually to have the thing the patient consults be so broad that it’s like having a doctor available who understands the local things.   LEE: Right.   GATES: We’re not there yet. But over the next two or three years, you know, particularly given the worsening financial constraints against African health systems, where the withdrawal of money has been dramatic, you know, figuring out how to take this—what I sometimes call “free intelligence”—and build a quality health system around that, we will have to be more radical in low-income countries than any rich country is ever going to be.   LEE: Also, there’s maybe a different regulatory environment, so some of those things maybe are easier? Because right now, I think the world hasn’t figured out how to and whether to regulate, let’s say, an AI that might give a medical diagnosis or write a prescription for a medication.  BUBECK: Yeah. I think one issue with this, and it’s also slowing down the deployment of AI in healthcare more generally, is a lack of proper benchmark. Because, you know, you were mentioning the USMLE [United States Medical Licensing Examination], for example. That’s a great test to test human beings and their knowledge of healthcare and medicine. But it’s not a great test to give to an AI.  It’s not asking the right questions. So finding what are the right questions to test whether an AI system is ready to give diagnosis in a constrained setting, that’s a very, very important direction, which to my surprise, is not yet accelerating at the rate that I was hoping for.  LEE: OK, so that gives me an excuse to get more now into the core AI tech because something I’ve discussed with both of you is this issue of what are the right tests. And you both know the very first test I give to any new spin of an LLM is I present a patient, the results—a mythical patient—the results of my physical exam, my mythical physical exam. Maybe some results of some initial labs. And then I present or propose a differential diagnosis. And if you’re not in medicine, a differential diagnosis you can just think of as a prioritized list of the possible diagnoses that fit with all that data. And in that proposed differential, I always intentionally make two mistakes.  I make a textbook technical error in one of the possible elements of the differential diagnosis, and I have an error of omission. And, you know, I just want to know, does the LLM understand what I’m talking about? And all the good ones out there do now. But then I want to know, can it spot the errors? And then most importantly, is it willing to tell me I’m wrong, that I’ve made a mistake?   That last piece seems really hard for AI today. And so let me ask you first, Seb, because at the time of this taping, of course, there was a new spin of GPT-4o last week that became overly sycophantic. In other words, it was actually prone in that test of mine not only to not tell me I’m wrong, but it actually praised me for the creativity of my differential. [LAUGHTER] What’s up with that?  BUBECK: Yeah, I guess it’s a testament to the fact that training those models is still more of an art than a science. So it’s a difficult job. Just to be clear with the audience, we have rolled back that [LAUGHS] version of GPT-4o, so now we don’t have the sycophant version out there.  Yeah, no, it’s a really difficult question. It has to do … as you said, it’s very technical. It has to do with the post-training and how, like, where do you nudge the model? So, you know, there is this very classical by now technique called RLHF [reinforcement learning from human feedback], where you push the model in the direction of a certain reward model. So the reward model is just telling the model, you know, what behavior is good, what behavior is bad.  But this reward model is itself an LLM, and, you know, Bill was saying at the very beginning of the conversation that we don’t really understand how those LLMs deal with concepts like, you know, where is the capital of France located? Things like that. It is the same thing for this reward model. We don’t know why it says that it prefers one output to another, and whether this is correlated with some sycophancy is, you know, something that we discovered basically just now. That if you push too hard in optimization on this reward model, you will get a sycophant model.  So it’s kind of … what I’m trying to say is we became too good at what we were doing, and we ended up, in fact, in a trap of the reward model.  LEE: I mean, you do want … it’s a difficult balance because you do want models to follow your desires and …  BUBECK: It’s a very difficult, very difficult balance.  LEE: So this brings up then the following question for me, which is the extent to which we think we’ll need to have specially trained models for things. So let me start with you, Bill. Do you have a point of view on whether we will need to, you know, quote-unquote take AI models to med school? Have them specially trained? Like, if you were going to deploy something to give medical care in underserved parts of the world, do we need to do something special to create those models?  GATES: We certainly need to teach them the African languages and the unique dialects so that the multimedia interactions are very high quality. We certainly need to teach them the disease prevalence and unique disease patterns like, you know, neglected tropical diseases and malaria. So we need to gather a set of facts that somebody trying to go for a US customer base, you know, wouldn’t necessarily have that in there.  Those two things are actually very straightforward because the additional training time is small. I’d say for the next few years, we’ll also need to do reinforcement learning about the context of being a doctor and how important certain behaviors are. Humans learn over the course of their life to some degree that, I’m in a different context and the way I behave in terms of being willing to criticize or be nice, you know, how important is it? Who’s here? What’s my relationship to them?   Right now, these machines don’t have that broad social experience. And so if you know it’s going to be used for health things, a lot of reinforcement learning of the very best humans in that context would still be valuable. Eventually, the models will, having read all the literature of the world about good doctors, bad doctors, it’ll understand as soon as you say, “I want you to be a doctor diagnosing somebody.” All of the implicit reinforcement that fits that situation, you know, will be there. LEE: Yeah. GATES: And so I hope three years from now, we don’t have to do that reinforcement learning. But today, for any medical context, you would want a lot of data to reinforce tone, willingness to say things when, you know, there might be something significant at stake.  LEE: Yeah. So, you know, something Bill said, kind of, reminds me of another thing that I think we missed, which is, the context also … and the specialization also pertains to different, I guess, what we still call “modes,” although I don’t know if the idea of multimodal is the same as it was two years ago. But, you know, what do you make of all of the hubbub around—in fact, within Microsoft Research, this is a big deal, but I think we’re far from alone—you know, medical images and vision, video, proteins and molecules, cell, you know, cellular data and so on.  BUBECK: Yeah. OK. So there is a lot to say to everything … to the last, you know, couple of minutes. Maybe on the specialization aspect, you know, I think there is, hiding behind this, a really fundamental scientific question of whether eventually we have a singular AGI [artificial general intelligence] that kind of knows everything and you can just put, you know, explain your own context and it will just get it and understand everything.  That’s one vision. I have to say, I don’t particularly believe in this vision. In fact, we humans are not like that at all. I think, hopefully, we are general intelligences, yet we have to specialize a lot. And, you know, I did myself a lot of RL, reinforcement learning, on mathematics. Like, that’s what I did, you know, spent a lot of time doing that. And I didn’t improve on other aspects. You know, in fact, I probably degraded in other aspects. [LAUGHTER] So it’s … I think it’s an important example to have in mind.  LEE: I think I might disagree with you on that, though, because, like, doesn’t a model have to see both good science and bad science in order to be able to gain the ability to discern between the two?  BUBECK: Yeah, no, that absolutely. I think there is value in seeing the generality, in having a very broad base. But then you, kind of, specialize on verticals. And this is where also, you know, open-weights model, which we haven’t talked about yet, are really important because they allow you to provide this broad base to everyone. And then you can specialize on top of it.  LEE: So we have about three hours of stuff to talk about, but our time is actually running low. BUBECK: Yes, yes, yes.   LEE: So I think I want … there’s a more provocative question. It’s almost a silly question, but I need to ask it of the two of you, which is, is there a future, you know, where AI replaces doctors or replaces, you know, medical specialties that we have today? So what does the world look like, say, five years from now?  GATES: Well, it’s important to distinguish healthcare discovery activity from healthcare delivery activity. We focused mostly on delivery. I think it’s very much within the realm of possibility that the AI is not only accelerating healthcare discovery but substituting for a lot of the roles of, you know, I’m an organic chemist, or I run various types of assays. I can see those, which are, you know, testable-output-type jobs but with still very high value, I can see, you know, some replacement in those areas before the doctor.   The doctor, still understanding the human condition and long-term dialogues, you know, they’ve had a lifetime of reinforcement of that, particularly when you get into areas like mental health. So I wouldn’t say in five years, either people will choose to adopt it, but it will be profound that there’ll be this nearly free intelligence that can do follow-up, that can help you, you know, make sure you went through different possibilities.  And so I’d say, yes, we’ll have doctors, but I’d say healthcare will be massively transformed in its quality and in efficiency by AI in that time period.  LEE: Is there a comparison, useful comparison, say, between doctors and, say, programmers, computer programmers, or doctors and, I don’t know, lawyers?  GATES: Programming is another one that has, kind of, a mathematical correctness to it, you know, and so the objective function that you’re trying to reinforce to, as soon as you can understand the state machines, you can have something that’s “checkable”; that’s correct. So I think programming, you know, which is weird to say, that the machine will beat us at most programming tasks before we let it take over roles that have deep empathy, you know, physical presence and social understanding in them.  LEE: Yeah. By the way, you know, I fully expect in five years that AI will produce mathematical proofs that are checkable for validity, easily checkable, because they’ll be written in a proof-checking language like Lean or something but will be so complex that no human mathematician can understand them. I expect that to happen.   I can imagine in some fields, like cellular biology, we could have the same situation in the future because the molecular pathways, the chemistry, biochemistry of human cells or living cells is as complex as any mathematics, and so it seems possible that we may be in a state where in wet lab, we see, Oh yeah, this actually works, but no one can understand why.  BUBECK: Yeah, absolutely. I mean, I think I really agree with Bill’s distinction of the discovery and the delivery, and indeed, the discovery’s when you can check things, and at the end, there is an artifact that you can verify. You know, you can run the protocol in the wet lab and see [if you have] produced what you wanted. So I absolutely agree with that.   And in fact, you know, we don’t have to talk five years from now. I don’t know if you know, but just recently, there was a paper that was published on a scientific discovery using o3- mini (opens in new tab). So this is really amazing. And, you know, just very quickly, just so people know, it was about this statistical physics model, the frustrated Potts model, which has to do with coloring, and basically, the case of three colors, like, more than two colors was open for a long time, and o3 was able to reduce the case of three colors to two colors.   LEE: Yeah.  BUBECK: Which is just, like, astounding. And this is not … this is now. This is happening right now. So this is something that I personally didn’t expect it would happen so quickly, and it’s due to those reasoning models.   Now, on the delivery side, I would add something more to it for the reason why doctors and, in fact, lawyers and coders will remain for a long time, and it’s because we still don’t understand how those models generalize. Like, at the end of the day, we are not able to tell you when they are confronted with a really new, novel situation, whether they will work or not.  Nobody is able to give you that guarantee. And I think until we understand this generalization better, we’re not going to be willing to just let the system in the wild without human supervision.  LEE: But don’t human doctors, human specialists … so, for example, a cardiologist sees a patient in a certain way that a nephrologist …  BUBECK: Yeah. LEE: … or an endocrinologist might not. BUBECK: That’s right. But another cardiologist will understand and, kind of, expect a certain level of generalization from their peer. And this, we just don’t have it with AI models. Now, of course, you’re exactly right. That generalization is also hard for humans. Like, if you have a human trained for one task and you put them into another task, then you don’t … you often don’t know. LEE: OK. You know, the podcast is focused on what’s happened over the last two years. But now, I’d like one provocative prediction about what you think the world of AI and medicine is going to be at some point in the future. You pick your timeframe. I don’t care if it’s two years or 20 years from now, but, you know, what do you think will be different about AI in medicine in that future than today?  BUBECK: Yeah, I think the deployment is going to accelerate soon. Like, we’re really not missing very much. There is this enormous capability overhang. Like, even if progress completely stopped, with current systems, we can do a lot more than what we’re doing right now. So I think this will … this has to be realized, you know, sooner rather than later.  And I think it’s probably dependent on these benchmarks and proper evaluation and tying this with regulation. So these are things that take time in human society and for good reason. But now we already are at two years; you know, give it another two years and it should be really …   LEE: Will AI prescribe your medicines? Write your prescriptions?  BUBECK: I think yes. I think yes.  LEE: OK. Bill?  GATES: Well, I think the next two years, we’ll have massive pilots, and so the amount of use of the AI, still in a copilot-type mode, you know, we should get millions of patient visits, you know, both in general medicine and in the mental health side, as well. And I think that’s going to build up both the data and the confidence to give the AI some additional autonomy. You know, are you going to let it talk to you at night when you’re panicked about your mental health with some ability to escalate? And, you know, I’ve gone so far as to tell politicians with national health systems that if they deploy AI appropriately, that the quality of care, the overload of the doctors, the improvement in the economics will be enough that their voters will be stunned because they just don’t expect this, and, you know, they could be reelected [LAUGHTER] just on this one thing of fixing what is a very overloaded and economically challenged health system in these rich countries.  You know, my personal role is going to be to make sure that in the poorer countries, there isn’t some lag; in fact, in many cases, that we’ll be more aggressive because, you know, we’re comparing to having no access to doctors at all. And, you know, so I think whether it’s India or Africa, there’ll be lessons that are globally valuable because we need medical intelligence. And, you know, thank god AI is going to provide a lot of that.  LEE: Well, on that optimistic note, I think that’s a good way to end. Bill, Seb, really appreciate all of this.   I think the most fundamental prediction we made in the book is that AI would actually find its way into the practice of medicine, and I think that that at least has come true, maybe in different ways than we expected, but it’s come true, and I think it’ll only accelerate from here. So thanks again, both of you.  [TRANSITION MUSIC]  GATES: Yeah. Thanks, you guys.  BUBECK: Thank you, Peter. Thanks, Bill.  LEE: I just always feel such a sense of privilege to have a chance to interact and actually work with people like Bill and Sébastien.    With Bill, I’m always amazed at how practically minded he is. He’s really thinking about the nuts and bolts of what AI might be able to do for people, and his thoughts about underserved parts of the world, the idea that we might actually be able to empower people with access to expert medical knowledge, I think is both inspiring and amazing.   And then, Seb, Sébastien Bubeck, he’s just absolutely a brilliant mind. He has a really firm grip on the deep mathematics of artificial intelligence and brings that to bear in his research and development work. And where that mathematics takes him isn’t just into the nuts and bolts of algorithms but into philosophical questions about the nature of intelligence.   One of the things that Sébastien brought up was the state of evaluation of AI systems. And indeed, he was fairly critical in our conversation. But of course, the world of AI research and development is just moving so fast, and indeed, since we recorded our conversation, OpenAI, in fact, released a new evaluation metric that is directly relevant to medical applications, and that is something called HealthBench. And Microsoft Research also released a new evaluation approach or process called ADeLe.   HealthBench and ADeLe are examples of new approaches to evaluating AI models that are less about testing their knowledge and ability to pass multiple-choice exams and instead are evaluation approaches designed to assess how well AI models are able to complete tasks that actually arise every day in typical healthcare or biomedical research settings. These are examples of really important good work that speak to how well AI models work in the real world of healthcare and biomedical research and how well they can collaborate with human beings in those settings.  You know, I asked Bill and Seb to make some predictions about the future. You know, my own answer, I expect that we’re going to be able to use AI to change how we diagnose patients, change how we decide treatment options.   If you’re a doctor or a nurse and you encounter a patient, you’ll ask questions, do a physical exam, you know, call out for labs just like you do today, but then you’ll be able to engage with AI based on all of that data and just ask, you know, based on all the other people who have gone through the same experience, who have similar data, how were they diagnosed? How were they treated? What were their outcomes? And what does that mean for the patient I have right now? Some people call it the “patients like me” paradigm. And I think that’s going to become real because of AI within our lifetimes. That idea of really grounding the delivery in healthcare and medical practice through data and intelligence, I actually now don’t see any barriers to that future becoming real.  [THEME MUSIC]  I’d like to extend another big thank you to Bill and Sébastien for their time. And to our listeners, as always, it’s a pleasure to have you along for the ride. I hope you’ll join us for our remaining conversations, as well as a second coauthor roundtable with Carey and Zak.   Until next time.   [MUSIC FADES]

View of the south façade before construction of a new residential project that now conceals Le Christin from Boulevard René Lévesque. PROJECT Le Christin, Montreal, Quebec ARCHITECT Atelier Big City PHOTOS James Brittain   PROJECT Les Studios du PAS, Montreal, Quebec ARCHITECT L. McComber in collaboration with Inform  PHOTOS Ulysse Lemerise   Nighttime, April 15, 2025. A thousand volunteers are gathering in Montreal, part of a province-wide effort to try and put numbers on a growing phenomenon in cities like Vancouver, Calgary, Toronto, and many others. The volunteers are getting ready to walk around targeted areas in downtown Montreal and around certain subway stations. Temporary shelters are also visited. First conducted in the spring of 2018, this survey showed that 3,149 people were in a vulnerable situation at the time. Four years later, a similar effort revealed that Montreal’s homeless population had risen to 4,690 people—and that there were some 10,000 people experiencing homelessness in the whole of the province. The 2025 numbers are expected to be significantly higher. For the organizers, this one-night snapshot of the situation is “neither perfect nor complete.” However, for nonprofit organizations and governmental bodies eager to prevent a vulnerable population from ending up on the streets, the informal census does provide highly valuable information.  Two recent initiatives—very different from one another—offer inspiring answers. The most recent one, Le Christin, was designed by Atelier Big City (led by architects Anne Cormier, Randy Cohen, and Howard Davies) and inaugurated in 2024. Studios du PAS, on the other hand, was designed by Montreal firm L. McComber, and welcomed its first tenants in 2022. Both projects involved long-standing charities: the 148-year-old Accueil Bonneau, in the case of Le Christin, and the 136-year-old Mission Old Brewery for Studios du PAS. Le Christin was spearheaded, and mostly financed, by the Société d’habitation et de développement de Montréal (SHDM), a non-profit, para-municipal corporation created in 1988. Studios du PAS was first selected by the City of Montreal to be built thanks to the Rapid Housing Initiative (RHI) program run by the Canada Mortgage and Housing Corporation (CMHC). Le Christin also received a financial contribution from the CMHC towards the end of the process. Boldly coloured blind walls signal the presence of Le Christin in the center of a densely occupied city block, with entrance to the left along Sanguinet Stree. Le Christin Although sited in a very central location, near the buzzing St. Catherine and St. Denis streets, Le Christin is hard to find. And even when one suddenly spots two seven-storey-high walls, coloured lemon-zest yellow and mango orange, it’s difficult to figure out what they are about. A stroll along the tiny Christin Street finally reveals the front façade of this new facility, now home to some of Montreal’s most vulnerable citizens.  View of Le Christin’s modulated front façade. Galvanized steel panels at ground level add a soft touch while protecting the building from potential damage caused by snow plows. Le Christin is unique for a number of reasons. First among them is its highly unusual location—at the centre of a dense city block otherwise occupied by university buildings, office towers, and condo blocks. Until a few years ago, the site was home to the four-storey Appartements Le Riga. The Art Deco-style building had been built in 1914 by developer-architect Joseph-Arthur Godin, who was a pioneer in his own right: he was one of the first in Montreal to experiment with reinforced concrete structures, a novelty in the city at the time. A century later, Le Riga, by then the property of SHDM, was in serious need of repair. Plans had already been drafted for a complete renovation of the building when a thorough investigation revealed major structural problems. Tenants had to leave on short notice and were temporarily relocated; the building was eventually demolished in 2019. By that time, Atelier Big City had been mandated to design a contemporary building that would replace Le Riga and provide a “place of one’s own” to close to 150 tenants, formerly homeless or at risk of becoming so.    Le Christin – Site Plan and Ground Floor Plan The entire operation sparked controversy, particularly as Le Christin started to rise, showing no sign of nostalgia. The architects’ daring approach was difficult to fathom—particularly for those who believe social housing should keep a low profile.  The program, originally meant for a clientele of single men, gradually evolved to include women. In order to reflect societal trends, the architects were asked to design 24 slightly larger units located in the building’s east wing, separated from the rest of the units by secured doors. Thus, Le Christin is able to accommodate homeless couples or close friends, as well as students and immigrants in need. A tenants-only courtyard is inserted in the south façade. In order to provide the maximum number of units requested by SHDM, each of the 90 studios was reduced to 230 square feet—an adjustment from Atelier Big City’s initial, slightly more generous plans. In a clever move, an L-shaped kitchen hugs the corner of each unit, pushing out against the exterior wall. As a result, the window openings recede from the façade, creating a sense of intimacy for the tenants, who enjoy contact with the exterior through large windows protected by quiet Juliet balconies. Far from damaging the initial design, the added constraint of tightened units allowed the architects to modulate the building’s façades, creating an even stronger statement. On the unit levels, corridors include large openings along the south façade. Each floor is colour-coded to enliven the space; overhead, perforated metal plates conceal the mechanical systems. An extra floor was gained thanks to the decision to expose the various plumbing, electrical, and ventilation systems. Well-lit meeting rooms and common areas are found near Le Christin’s front entrance, along with offices for personnel, who are present on the premises 24 hours a day. Apart from a small terrace above the entrance, the main exterior space is a yard which literally cuts into the building’s back façade. This has a huge impact on the interiors at all levels: corridors are generously lit with sunlight, a concept market developers would be well advised to imitate. The adjacent exit stairs are also notable, with their careful detailing and the presence of glazed openings.  The fire stairs, which open onto the exterior yard at ground level, feature glazing that allows for ample natural light. Le Christin has achieved the lofty goal articulated by SHDM’s former director, architect Nancy Schoiry: “With this project, we wanted to innovate and demonstrate that it was possible to provide quality housing for those at risk of homelessness.” The low-slung Studios du PAS aligns with neighbourhood two-storey buildings. Studios du PAS In sharp contrast with Le Christin’s surroundings, the impression one gets approaching Studios du PAS, 14 kilometres east of downtown Montreal, is that of a small town. In this mostly low-scale neighbourhood, L. McComber architects adopted a respectful, subdued approach—blending in, rather than standing out.  The project uses a pared-down palette of terracotta tile, wood, and galvanized steel. The footbridge links the upper level to shared exterior spaces. The financing for this small building, planned for individuals aged 55 or older experiencing or at risk of homelessness, was tied to a highly demanding schedule. The project had to be designed, built, and occupied within 18 months: an “almost impossible” challenge, according to principal architect Laurent McComber. From the very start, prefabrication was favoured over more traditional construction methods. And even though substantial work had to be done on-site—including the installation of the roof, electrical and mechanical systems, as well as exterior and interior finishes—the partially prefabricated components did contribute to keeping costs under control and meeting the 18-month design-to-delivery deadline. Les Studios du PAS The building was divided into 20 identical modules, each fourteen feet wide—the maximum width allowable on the road. Half the modules were installed at ground level. One of these, positioned nearest the street entrance, serves as a community room directly connected to a small office for the use of a social worker, allowing staff to follow up regularly with tenants. Flooded with natural light, the double-height lobby provides a friendly and inclusive welcome. The ground level studios were designed so they could be adapted to accommodate accessibility needs. Some of the ground floor units were adapted to meet the needs of those with a physical disability; the other units were designed to be easily adaptable if needed. All studio apartments, slightly under 300 square feet, include a full bathroom, a minimal kitchen, and sizeable storage space hidden behind cabinet doors. Most of the apartments include a small exterior alcove, which provides an intimate outdoor space while creating a subtle rhythm along the front façade. Inside the studio units, storage cupboards for clothes and belongings were added as an extension of the kitchen wall. Conscious of the tradition of brick residential buildings in Montreal, yet wanting to explore alternate materials, the architects selected an earth-toned terracotta tile from Germany. The 299mm x 1500mm tiles are clipped to the façade, allowing for faster installation and easier maintenance. All units enjoy triple-glazed windows and particularly well insulated walls. A high-performance heat pump was installed to lower energy demand—and costs—for heating and cooling needs. Wood siding was used to soften the upper-level balconies, which provide protected outdoor spaces for residents.   Pride and Dignity Le Christin and Les Studios du PAS have little in common—except, of course, their program. Architecturally speaking, each represents an interesting solution to the problem at hand. While Le Christin is a high-spirited, flamboyant statement, Studios du PAS is to be praised for its respectful attitude, and for the architects’ relentless search for interesting alternatives to traditional construction norms. Atelier Big City is one of few firms in Canada that has the guts—and the talent—to play with bold colours. Decades of experimentation (not just with public buildings, but also within their own homes), led up to Le Christin, which is perhaps their strongest building to date. Their judicious choices of colour, brick type, and materials transmit a message of pride and dignity. Both projects demonstrate enormous respect and generosity to their residents: they provide architecture that treats them not as an underclass, but as regular people, who need the stability of dignified housing to start rebuilding their lives. Odile Hénault is a contributing editor to Canadian Architect.   Le Christin CLIENT Société d’habitation et de développement de Montréal (SHDM) | ARCHITECT TEAM Anne Cormier, Randy Cohen, Howard Davies, Fannie Yockell, Gabriel Tessier, Sébastien St-Laurent, Lisa Vo | STRUCTURAL DPHV | MECHANICAL/ELECTRICAL BPA | CIVIL Genexco | LIGHTING CS Design | AREA 4,115 m2 | Construction BUDGET $18.9 M | COMPLETION November 2023   Les Studios du PAS  CLIENT PAS de la rue | ARCHITECT TEAM L. McComber—Laurent McComber, Olivier Lord, Jérôme Lemieux, Josianne Ouellet-Daudelin, Laurent McComber. Inform—David Grenier, Élisabeth Provost, Amélie Tremblay, David Grenier | PROJECT MANAGEMENT Groupe CDH | STRUCTURAL Douglas Consultants | MECHANICAL/ELECTRICAL Martin Roy & associés | CIVIL Gravitaire | CONTRACTOR Gestion Étoc | AREA 1,035 m2 | BUDGET $3.4 M | COMPLETION September 2022 As appeared in the June 2025 issue of Canadian Architect magazine The post A Place to Call Home: Le Christin and Les Studios du PAS, Montreal, Quebec appeared first on Canadian Architect.

A Fungal Disease Ravaged North American Bats. Now, Researchers Found a Second Species That Suggests It Could Happen Again White-nose syndrome caused millions of bat deaths, and scientists are sounding the alarm that a second fungus could be disastrous if it reaches American wildlife Lillian Ali - Staff Contributor May 30, 2025 A little brown bat (Myotis lucifugus) is seen with white fuzz on its nose, a characteristic of the deadly white-nose syndrome. Ryan von Linden / New York Department of Environmental Conservation In February 2006, a cave explorer near Albany, New York, took the first photograph of bats with a mysterious white growth on their faces. Later, biologists studying the mammals in caves and mines discovered piles of dead bats in the state—also with the fuzzy white mold. The scientists were floored. For years, no one knew what was causing the mass die-offs from this “white-nose syndrome.” In early 2007, Albany residents called local authorities with reports of typically nocturnal bats flying in broad daylight. “They were just dying on the landscape,” wildlife biologist Alan Hicks told the Associated Press’ Michael Hill in 2008. “They were crashing into snowbanks, crawling into woodpiles and dying.” At last, scientists identified a culprit: The bats had succumbed to an infection caused by the fungus Pseudogymnoascus destructans. Since its initial discovery, white-nose syndrome has killed millions of bats across 40 U.S. states and nine Canadian provinces, making it “the most dramatic wildlife mortality event that’s ever been documented from a pathogen,” DeeAnn Reeder, a disease ecologist at Bucknell University, tells the New York Times’ Carl Zimmer. Now, nearly two decades later, scientists have developed some promising ways to fend off the disease, including an experimental vaccine. But a new study published this week in the journal Nature warns of a newly discovered second species of fungus that, if it reaches North America, could set all that progress back. “We thought we knew our enemy, but we have now discovered it is twice the size and potentially more complex than we had imagined,” lead author Nicola Fischer, a biologist at the University of Greifswald in Germany, says in a statement. Little brown bats are susceptible to white-nose syndrome in North America. Krynak Tim, U.S. Fish and Wildlife Service The team analyzed 5,479 fungus samples collected by hundreds of citizen science volunteers across North America, Asia and Europe. They found that white-nose syndrome is caused by two distinct fungal species native to Europe and Asia, with only one species having reached North America so far. If the second species hits the continent, it could look like a “reboot” of the epidemic, Reeder tells the New York Times. Study co-author Sébastien Puechmaille, an evolutionary biologist at the University of Montpellier in France, knew bats in Europe had also been seen with white fuzz on their noses, as he tells the New York Times. But those populations didn’t die off like American bats. Charting the disease across Europe and Asia, he noticed that the fungus was able to live alongside those bats, while it ravaged American ones. In its native range, the fungus grows in the bodies of hibernating bats as their internal temperature drops, then it’s shed in the spring when they awaken. But in American bats, the fungus causes their immune systems to activate and burn fat reserves as they hibernate. The bats then wake up periodically, causing irregular activity and eventual starvation. The researchers suggest the damaging fungal spores were first brought to North America by cavers that traveled from Europe—potentially western Ukraine—to the United States without completely disinfecting their boots or rope. White-nose syndrome poses a threat not just to bats, but to whole ecosystems. Bats are vital parts of many food chains, eating insects and pollinating plants. However, they reproduce fairly slowly, only having one or two pups at a time. Rebuilding a bat population, then, could take decades. And since cave ecosystems are similarly delicate, biologists are wary of trying to kill off the fungus preemptively. “Cave ecosystems are so fragile that if you start pulling on this thread, what else are you going to unravel that may create bigger problems in the cave system?” said University of Wisconsin–Madison wildlife specialist David Drake to the Badger Herald’s Kiran Mistry in December. The discovery also occurs as the original wave of white-nose syndrome continues to spread across North America, having just crossed the Continental Divide in Colorado. Just one spore of the new species could be devastating to American bat colonies. Puechmaille tells the New York Times that policies should be put in place to make sure the second fungus does not spread to more continents, and that cavers should not move equipment between countries and should disinfect it regularly. “This work … powerfully illustrates the profound impact a single translocation event can have on wildlife,” he adds in the statement. Get the latest stories in your inbox every weekday.

There are few industries these days that are not touched by artificial intelligence (AI). Networking is very much one that is touched. It is barely conceivable that any network of any reasonable size – from an office local area network or home router to a global telecoms infrastructure – could not “just” be improved by AI. Just take the words of Swisscom’s chief technical officer, Mark Düsener, about his company’s partnership with Cisco-owned Outshift to deploy agentic AI – of which more later – through his organisation. “The goal of getting into an agentic AI world, operating networks and connectivity is all about reducing the impact of service changes, reducing the risk of downtime and costs – therefore levelling up our customer experience.”  In other words, the implementation of AI results in operational efficiencies, increased reliability and user benefits. Seems simple, yes? But as we know, nothing in life is simple, and to guarantee such gains, AI can’t be “just” switched on. And perhaps most importantly, the benefits of AI in networking can’t be realised fully without considering networking for AI. It seems logical that any investigation of AI and networking – or indeed, AI and anything – should start with Nvidia, a company that has played a pivotal role in developing the AI tech ecosystem, and is set to do so further. Speaking in 2024 at a tech conference about how AI has established itself as an intrinsic part of business, Nvidia founder and CEO Jensen Huang observed that the era of generative AI (GenAI) is here and that enterprises must engage with “the single most consequential technology in history”. He told the audience that what was happening was the greatest fundamental computing platform transformation in 60 years, encompassing general-purpose computing to accelerated computing.  “We’re sitting on a mountain of data. All of us. We’ve been collecting it in our businesses for a long time. But until now, we haven’t had the ability to refine that, then discover insight and codify it automatically into our company’s natural experience, our digital intelligence. Every company is going to be an intelligence manufacturer. Every company is built on domain-specific intelligence. For the very first time, we can now digitise that intelligence and turn it into our AI – the corporate AI,” he said. “AI is a lifecycle that lives forever. What we are looking to do is turn our corporate intelligence into digital intelligence. Once we do that, we connect our data and our AI flywheel so that we collect more data, harvest more insight and create better intelligence. This allows us to provide better services or to be more productive, run faster, be more efficient and do things at a larger scale.”  Concluding his keynote, Huang stressed that enterprises must now engage with the “single most consequential technology in history” to translate and condense a company’s intelligence into digital intelligence. This is precisely what Swisscom is aiming to achieve. The company is Switzerland’s largest telecoms provider with more than six million mobile customers and 10,000 mobile antenna sites that have to be managed effectively. When its network engineers make changes to the infrastructure, they face a common challenge: how to update systems that serve millions of customers without disrupting the service. The solution was partnering with Outshift to develop practical applications of AI agents in network operations to “redefine” customer experiences. That is, using Outshift’s Internet of Agents to deliver meaningful results for the telco, while also meeting customer needs through AI innovation. But these advantages are not the preserve of large enterprises such as telcos. Indeed, from a networking perspective, AI can enable small- and medium-sized businesses to gain access to enterprise-level technology that can allow them to focus on growth and eliminate the costs and infrastructure challenges that arise when managing complex IT infrastructures.  From a broader perspective, Swisscom and Outshift have also shown that making AI work effectively requires something new: an infrastructure that lets businesses communicate and work together securely. And this is where the two sides of AI and networking come into play. At the event where Nvidia’s Huang outlined his vision, David Hughes, chief product officer of HPE Aruba Networking, said there were pressing issues about the use of AI in enterprise networks, in particular around harnessing the benefits that GenAI can offer. Regarding “AI for networking” and “networking for AI”, Hughes suggested there are subtle but fundamental differences between the two.  “AI for networking is where we spend time from an engineering and data science point of view. It’s really about [questioning] how we use AI technology to turn IT admins into super-admins so that they can handle their escalating workloads independent of GenAI, which is kind of a load on top of everything else, such as escalating cyber threats and concerns about privacy. The business is asking IT to do new things, deploy new apps all the time, but they’re [asking this of] the same number of people,” he observed.  What we are starting to see, and expect more of, is AI computing increasingly taking place at the edge to eliminate the distance between the prompt and the process Bastien Aerni, GTT “Networking for AI is about building out, first and foremost, the kind of switching infrastructure that’s needed to interconnect GPU [graphics processing unit] clusters. And then a little bit beyond that, thinking about the impact of collecting telemetry on a network and the changes in the way people might want to build out their network.”  And impact there is. A lot of firms currently investigating AI within their businesses find themselves asking how to manage the mass adoption of AI in relation to networking and data flows, such as the kind of bandwidth and capacity required to facilitate AI-generated output such as text, image and video content. This, says Bastien Aerni, vice-president of strategy and technology adoption at global networking and security-as-a-service firm GTT, is causing companies to rethink the speed and scale of their networking needs.  “To achieve the return on investment of AI initiatives, they have to be able to secure and process large amounts of data quickly, and to this end, their network architecture must be configured to support this kind of workload. Utilising a platform embedded in a Tier 1 IP [internet protocol] backbone here ensures low latency, high bandwidth and direct internet access globally,” he remarks.   “What we are starting to see, and expect more of, is AI computing increasingly taking place at the edge to eliminate the distance between the prompt and the process. Leveraging software-defined wide area network [SD-WAN] services built in the right platform to efficiently route AI data traffic can reduce latency and security risk, and provide more control over data.”  At the end of 2023, BT revealed that its networks had come under huge strain after the simultaneous online broadcast of six Premier League football matches and downloads of popular games, with the update of Call of Duty Modern Warfare particularly cited. AI promises to add to this headache.  Speaking at Mobile World Congress 2025, BT Business chief technology officer (CTO) Colin Bannon said that in the new, reshaped world of work, a robust and reliable network is a fundamental prerequisite for AI to work, and that it requires effort to stay relevant to meet ongoing challenges faced by the customers BT serves, mainly international business, governments and multinationals. The bottom line is that network performance to support the AI-enabled world is crucial in a world where “slow is the new down”.  Bannon added that Global Fabric, BT’s network-as-a-service product, was constructed before AI “blew up” and that BT was thinking of how to deal with a hyper-distributed set of workloads on a network and to be able to make it fully programmable. Looking at the challenges ahead and how the new network will resolve them, he said: “[AI] just makes distributed and more complex workflows even bigger, which makes the need for a fabric-type network even more important. You need a network that can [handle data] burst, and that is programmable, and that you can [control] bandwidth on demand as well. All of this programmability [is something businesses] have never had before. I would argue that the network is the computer, and the network is a prerequisite for AI to work.”  The result would be constructing enterprise networks that can cope with the massive strain placed on utilisation from AI, especially in terms of what is needed for training models. Bannon said there were three key network challenges and conditions to deal with AI: training requirements, inference requirements and general requirements.   He stated that the dynamic nature of AI workloads means networks need to be scalable and agile, with visibility tools that offer real-time monitoring, issue detection and troubleshooting. As regards specific training requirements, dealing with AI necessitates the movement of large datasets across the network, thus demanding high-bandwidth networks. He also described “elephant” flows of data – that is, continuous transmission over time and training over days. He warned that network inconsistencies could affect the accuracy and training time of AI models, and that tail latency could impact job completion time significantly. This means robust congestion management is needed to detect potential congestion and redistribute network traffic.  But AI training models generally spell network trouble. And now the conversation is turning from the use of generic large language models (see Preparing networks for Industry 5.0 box) to application/industry-dedicated small language models. Read more articles about AI for networking How network engineers can prepare for the future with AI: The rapid rise of AI has left some professionals feeling unprepared. GenAI is beneficial to networks, but engineers must have the proper tools to adapt to this new change. Cisco Live EMEA – network supplier tightens AI embrace: At its annual EMEA show, Cisco tech leaders unveiled a raft of new products, services and features designed to help customers do more with artificial intelligence. NTT Data has created and deployed a small language model called Tsuzumi, described as an ultra-lightweight model designed to reduce learning and inference costs. According to NTT’s UK and Ireland CTO, Tom Winstanley, the reason for developing this model has principally been to support edge use cases. “[That is] literally deployment at the edge of the network to avoid flooding of the network, also addressing privacy concerns, also addressing sustainability concerns around some of these very large language models being very specific in creating domain context,” he says.   “Examples of that can be used in video analytics, media analytics, and in capturing conversations in real time, but locally, and not deploying it out to flood the network. That said, the flip side of this was there was immense power sitting in some of these central hyper-scale models and capacities, and you also therefore need to find out more [about] what’s the right network background, and what’s the right balance of your network infrastructure. For example, if you want to do real-time media streaming from a [sports stadium] and do all of the edits on-site, or remotely so not to have to deploy [facilities] to every single location, then you need a different backbone, too.”  Winstanley notes that his company is part of a wider group that in media use cases could offer hyper-directional sound systems supported by AI. “This is looking like a really interesting area of technology that is relevant for supporter experience in a stadium – dampening, sound targeting. And then we’re back to the connection to the edge of the AI story. And that’s exciting for us. That is the frontier.”  But coming back from the frontier of technology to bread-and-butter business operations, even if the IT and comms community is confident that it can address any technological issues that arise regarding AI and networking, businesses themselves may not be so sure.  Research published by managed network-as-a-service provider Expereo in April 2025 revealed that despite 88% of UK business leaders regarding AI as becoming important to fulfilling business priorities in the next 12 months, there are a number of major roadblocks to AI plans by UK businesses. These include from employees and unreasonable demands, as well as poor existing infrastructure.   Worryingly, among the key findings of Expereo’s Enterprise horizons 2025 study was the general feeling from a lot of UK technology leaders that expectations within their organisation of what AI can do are growing faster than their ability to meet them. While 47% of UK organisations noted that their network/connectivity infrastructure was not ready to support new technology initiatives, such as AI, in general, a further 49% reported that their network performance was preventing or limiting their ability to support large data and AI projects.  Assessing the key trends revealed in the study, Expereo CEO Ben Elms says that as global businesses embrace AI to transform employee and customer experience, setting realistic goals and aligning expectations will be critical to ensuring that AI delivers long-term value, rather than being viewed as a quick fix. “While the potential of AI is immense, its successful integration requires careful planning. Technology leaders must recognise the need for robust networks and connectivity infrastructure to support AI at scale, while also ensuring consistent performance across these networks,” he says.  Summing up the state of the industry, Elms states that business is currently at a pivotal moment where strategic investments in technology and IT infrastructure are necessary to meet both current and future demands. In short, reflecting Düsener’s point about Swisscom’s aim to reduce the impact of service changes, reduce the risk of downtime and costs, and improve customer services. Just switching on any AI system and believing that any answer is “out there” just won’t do. Your network could very well tell you otherwise.  Through its core Catia platform and its SolidWorks subsidiary, engineering software company Dassault Systèmes sees artificial intelligence (AI) as now fundamental to its design and manufacturing work in virtually all production industries. Speaking to Computer Weekly in February 2025, the company’s senior vice-president, Gian Paolo Bassi, said the conversation of its sector has evolved from Industry 4.0, which was focused on automation, productivity and innovation without taking into account the effect of technological changes in society.   “The industry has decided that it’s time for an evolution,” he said. “It’s called Industry 5.0. At the intersection of the experience economy, there is a new, compelling necessity to be sustainable, to create a circular economy. So then, at the intersection, [we have] the generative [AI] economy.” Yet in aiming to generate gains in sustainability through Industry 5.0, there is a danger that the increased use of AI could potentially see increased power usage, as well as the need to invest in much more robust and responsive connected network infrastructure to support the rise in AI-based workloads.  Dassault first revealed it was working with generative AI design principles in 2024. As the practice has evolved, Bassi said it now captures two fundamental concepts. The first is the ability of AI to create new and original content based on language models that comprise details of processes, business models, designs of parts assemblies, specifications and manufacturing practices. These models, he stressed, would not be traditional, generic, compute-intensive models such as ChatGPT. Instead, they would be vertical, industry-specific, and trained on engineering content and technical documentation.  “We can now build large models of everything, which is a virtual twin, and we can get to a level of sophistication where new ideas can come in, be tested, and much more knowledge can be put into the innovation process. This is a tipping point,” he remarked. “It’s not a technological change. It’s a technological expansion – a very important one – because we are going to improve, to increase our portfolio with AI agents, with virtual companions and also content, because generative AI can generate content, and can generate, more importantly, know-how and knowledge that can be put to use by our customers immediately.” This tipping point means the software provider can bring knowledge and know-how to a new level because, in Bassi’s belief, this is what AI is best at: exploiting the large models of industrial practices. And with the most important benefit of addressing customer needs as the capabilities of AI are translated into the industrial world, offering a pathway for engineers to save precious time in research and spend more time on being creative in design, without massive, network-intensive models. “Right now, there is this rush to create larger and more comprehensive models. However, it may [just] be a temporary limitation of the technology,” Bassi suggested. “In fact, it is indeed possible that you don’t need the huge models to do specific tasks.” 

Sébastien Broca, sociologue : « Les Big Tech entretiennent des relations plus complexes qu’on ne le pense avec les utopistes d’Internet » Dans un entretien au « Monde », le sociologue revient sur l’évolution des Gafam, ces géants du Web qui se disent héritiers de la contre-culture américaine mais sont aujourd’hui les alliés du pouvoir capitaliste en place. Article réservé aux abonnés Sébastien Broca, le 12 mars 2025. HERMANCE TRIAY Sébastien Broca est sociologue et maître de conférences en science de l’information et de la communication à l’université Paris-VIII. Dans son livre Pris dans la Toile (Seuil, 288 pages, 23 euros), il raconte comment les géants du numérique ont récupéré à leur profit les idéaux des pionniers d’Internet. Son récit s’intéresse aux différents mouvements critiques de cette évolution, pour mieux saisir les ressorts d’un échec politique. Expliquez-nous de quoi étaient faites les utopies d’Internet, récupérées par les entreprises du numérique… Ces utopies portaient l’espoir que les technologies numériques, l’informatique et Internet, soient des technologies émancipatrices. Elles feraient advenir une société qui laisserait plus de place à l’autonomie individuelle, aux capacités collectives d’auto-organisation et permettraient d’abattre un certain nombre de pouvoirs institués, comme les Etats ou les grandes entreprises. Plus généralement, elles seraient un outil de démocratisation. Il vous reste 89.03% de cet article à lire. La suite est réservée aux abonnés.

Here's what we looked at this week:French company Cyclauto revives a modular cargo bike design from the 1930s. The Rivian R1T's brilliant "seamless tailgate" design feature provides a way better UX for unloading.Škoda's futuristic Slavia B electric motorcycle concept. Joe Doucet's OLO Table, made with Polygood Oyster, gives new life to old refrigerators, keyboards and computer mice.This Aura Triple Boost Pro adds three unfolding screens to your laptop.At the German Pavilion at Expo 2025 Osaka, an interesting twist on the sunken living room (though that's not what it's supposed to evoke).Automatic photo timestamp/location-stamp apps for keeping track of projects.Laurids Gallée's Tralucid Stools are both crisp and blurry.The forgotten trade of the patternmaker: How metal objects used to first be made out of wood.The origins of the first mass-market tool chest, by H. Gerstner & Sons.Peter Ivy's gorgeous Rokkakei hand-blown glass pendant lamps.Acrobatic artist Bastien Dausse's low-tech wall-walking invention. I bet there's a fun commercial application for this.Hozo designs a better cordless ultrasonic cutter, with both UX and safety improvements.The Bzigo Iris is an in-room mosquito tracking device that lets you easily kill them.All 17 components of Freitag's Mono[PA6] bags are completely monomaterial and recyclable.For the Boat Life crowd, fun floating furniture by Bote.Design opportunity: Create a better portable tufting frame.The Pivotal Helix: A cute, goofy-looking personal electric flying machine.Industrial design case study: Tactile helps Milwaukee develop their digital level.

Transcript [MUSIC]      [BOOK PASSAGE]   PETER LEE: “We need to start understanding and discussing AI’s potential for good and ill now. Or rather, yesterday. … GPT-4 has game-changing potential to improve medicine and health.”  [END OF BOOK PASSAGE]   [THEME MUSIC]      This is The AI Revolution in Medicine, Revisited. I’m your host, Peter Lee.      Shortly after OpenAI’s GPT-4 was publicly released, Carey Goldberg, Dr. Zak Kohane, and I published The AI Revolution in Medicine to help educate the world of healthcare and medical research about the transformative impact this new generative AI technology could have. But because we wrote the book when GPT-4 was still a secret, we had to speculate. Now, two years later, what did we get right, and what did we get wrong?       In this series, we’ll talk to clinicians, patients, hospital administrators, and others to understand the reality of AI in the field and where we go from here. [THEME MUSIC FADES]   The passage I read at the top is from the book’s prologue.    When Carey, Zak, and I wrote the book, we could only speculate how generative AI would be used in healthcare because GPT-4 hadn’t yet been released. It wasn’t yet available to the very people we thought would be most affected by it. And while we felt strongly that this new form of AI would have the potential to transform medicine, it was such a different kind of technology for the world, and no one had a user’s manual for this thing to explain how to use it effectively and also how to use it safely.   So we thought it would be important to give healthcare professionals and leaders a framing to start important discussions around its use. We wanted to provide a map not only to help people navigate a new world that we anticipated would happen with the arrival of GPT-4 but also to help them chart a future of what we saw as a potential revolution in medicine.   So I’m super excited to welcome my coauthors: longtime medical/science journalist Carey Goldberg and Dr. Zak Kohane, the inaugural chair of Harvard Medical School’s Department of Biomedical Informatics and the editor-in-chief for The New England Journal of Medicine AI.   We’re going to have two discussions. This will be the first one about what we’ve learned from the people on the ground so far and how we are thinking about generative AI today.   [TRANSITION MUSIC]  Carey, Zak, I’m really looking forward to this.  CAREY GOLDBERG: It’s nice to see you, Peter.   LEE: [LAUGHS] It’s great to see you, too.  GOLDBERG: We missed you.  ZAK KOHANE: The dynamic gang is back. [LAUGHTER]  LEE: Yeah, and I guess after that big book project two years ago, it’s remarkable that we’re still on speaking terms with each other. [LAUGHTER]  In fact, this episode is to react to what we heard in the first four episodes of this podcast. But before we get there, I thought maybe we should start with the origins of this project just now over two years ago. And, you know, I had this early secret access to Davinci 3, now known as GPT-4.   I remember, you know, experimenting right away with things in medicine, but I realized I was in way over my head. And so I wanted help. And the first person I called was you, Zak. And you remember we had a call, and I tried to explain what this was about. And I think I saw skepticism in—polite skepticism—in your eyes. But tell me, you know, what was going through your head when you heard me explain this thing to you?  KOHANE: So I was divided between the fact that I have tremendous respect for you, Peter. And you’ve always struck me as sober. And we’ve had conversations which showed to me that you fully understood some of the missteps that technology—ARPA, Microsoft, and others—had made in the past. And yet, you were telling me a full science fiction compliant story [LAUGHTER] that something that we thought was 30 years away was happening now.   LEE: Mm-hmm.  KOHANE: And it was very hard for me to put together. And so I couldn’t quite tell myself this is BS, but I said, you know, I need to look at it. Just this seems too good to be true. What is this? So it was very hard for me to grapple with it. I was thrilled that it might be possible, but I was thinking, How could this be possible?  LEE: Yeah. Well, even now, I look back, and I appreciate that you were nice to me, because I think a lot of people would have [LAUGHS] been much less polite. And in fact, I myself had expressed a lot of very direct skepticism early on.   After ChatGPT got released, I think three or four days later, I received an email from a colleague running … who runs a clinic, and, you know, he said, “Wow, this is great, Peter. And, you know, we’re using this ChatGPT, you know, to have the receptionist in our clinic write after-visit notes to our patients.”   And that sparked a huge internal discussion about this. And you and I knew enough about hallucinations and about other issues that it seemed important to write something about what this could do and what it couldn’t do. And so I think, I can’t remember the timing, but you and I decided a book would be a good idea. And then I think you had the thought that you and I would write in a hopelessly academic style [LAUGHTER] that no one would be able to read.   So it was your idea to recruit Carey, I think, right?  KOHANE: Yes, it was. I was sure that we both had a lot of material, but communicating it effectively to the very people we wanted to would not go well if we just left ourselves to our own devices. And Carey is super brilliant at what she does. She’s an idea synthesizer and public communicator in the written word and amazing.  LEE: So yeah. So, Carey, we contact you. How did that go?  GOLDBERG: So yes. On my end, I had known Zak for probably, like, 25 years, and he had always been the person who debunked the scientific hype for me. I would turn to him with like, “Hmm, they’re saying that the Human Genome Project is going to change everything.” And he would say, “Yeah. But first it’ll be 10 years of bad news, and then [LAUGHTER] we’ll actually get somewhere.”    So when Zak called me up at seven o’clock one morning, just beside himself after having tried Davinci 3, I knew that there was something very serious going on. And I had just quit my job as the Boston bureau chief of Bloomberg News, and I was ripe for the plucking. And I also … I feel kind of nostalgic now about just the amazement and the wonder and the awe of that period. We knew that when generative AI hit the world, there would be all kinds of snags and obstacles and things that would slow it down, but at that moment, it was just like the holy crap moment. [LAUGHTER] And it’s fun to think about it now. LEE: Yeah. KOHANE: I will see that and raise that one. I now tell GPT-4, please write this in the style of Carey Goldberg.   GOLDBERG: [LAUGHTER] No way! Really?   KOHANE: Yes way. Yes way. Yes way.  GOLDBERG: Wow. Well, I have to say, like, it’s not hard to motivate readers when you’re writing about the most transformative technology of their lifetime. Like, I think there’s a gigantic hunger to read and to understand. So you were not hard to work with, Peter and Zak. [LAUGHS]  LEE: All right. So I think we have to get down to work [LAUGHS] now.   Yeah, so for these podcasts, you know, we’re talking to different types of people to just reflect on what’s actually happening, what has actually happened over the last two years. And so the first episode, we talked to two doctors. There’s Chris Longhurst at UC San Diego and Sara Murray at UC San Francisco. And besides being doctors and having AI affect their clinical work, they just happen also to be leading the efforts at their respective institutions to figure out how best to integrate AI into their health systems.  And, you know, it was fun to talk to them. And I felt like a lot of what they said was pretty validating for us. You know, they talked about AI scribes. Chris, especially, talked a lot about how AI can respond to emails from patients, write referral letters. And then, you know, they both talked about the importance of—I think, Zak, you used the phrase in our book “trust but verify”—you know, to have always a human in the loop.    What did you two take away from their thoughts overall about how doctors are using … and I guess, Zak, you would have a different lens also because at Harvard, you see doctors all the time grappling with AI.  KOHANE: So on the one hand, I think they’ve done some very interesting studies. And indeed, they saw that when these generative models, when GPT-4, was sending a note to patients, it was more detailed, friendlier.  But there were also some nonobvious results, which is on the generation of these letters, if indeed you review them as you’re supposed to, it was not clear that there was any time savings. And my own reaction was, Boy, every one of these things needs institutional review. It’s going to be hard to move fast.   And yet, at the same time, we know from them that the doctors on their smartphones are accessing these things all the time. And so the disconnect between a healthcare system, which is duty bound to carefully look at every implementation, is, I think, intimidating.   LEE: Yeah.  KOHANE: And at the same time, doctors who just have to do what they have to do are using this new superpower and doing it. And so that’s actually what struck me …   LEE: Yeah.  KOHANE: … is that these are two leaders and they’re doing what they have to do for their institutions, and yet there’s this disconnect.  And by the way, I don’t think we’ve seen any faster technology adoption than the adoption of ambient dictation. And it’s not because it’s time saving. And in fact, so far, the hospitals have to pay out of pocket. It’s not like insurance is paying them more. But it’s so much more pleasant for the doctors … not least of which because they can actually look at their patients instead of looking at the terminal and plunking down.   LEE: Carey, what about you?  GOLDBERG: I mean, anecdotally, there are time savings. Anecdotally, I have heard quite a few doctors saying that it cuts down on “pajama time” to be able to have the note written by the AI and then for them to just check it. In fact, I spoke to one doctor who said, you know, basically it means that when I leave the office, I’ve left the office. I can go home and be with my kids.  So I don’t think the jury is fully in yet about whether there are time savings. But what is clear is, Peter, what you predicted right from the get-go, which is that this is going to be an amazing paper shredder. Like, the main first overarching use cases will be back-office functions.  LEE: Yeah, yeah. Well, and it was, I think, not a hugely risky prediction because, you know, there were already companies, like, using phone banks of scribes in India to kind of listen in. And, you know, lots of clinics actually had human scribes being used. And so it wasn’t a huge stretch to imagine the AI. [TRANSITION MUSIC]  So on the subject of things that we missed, Chris Longhurst shared this scenario, which stuck out for me, and he actually coauthored a paper on it last year.  CHRISTOPHER LONGHURST: It turns out, not surprisingly, healthcare can be frustrating. And stressed patients can send some pretty nasty messages to their care teams. [LAUGHTER] And you can imagine being a busy, tired, exhausted clinician and receiving a bit of a nasty-gram. And the GPT is actually really helpful in those instances in helping draft a pretty empathetic response when I think the human instinct would be a pretty nasty one.  LEE: [LAUGHS] So, Carey, maybe I’ll start with you. What did we understand about this idea of empathy out of AI at the time we wrote the book, and what do we understand now?  GOLDBERG: Well, it was already clear when we wrote the book that these AI models were capable of very persuasive empathy. And in fact, you even wrote that it was helping you be a better person, right. [LAUGHS] So their human qualities, or human imitative qualities, were clearly superb. And we’ve seen that borne out in multiple studies, that in fact, patients respond better to them … that they have no problem at all with how the AI communicates with them. And in fact, it’s often better.   And I gather now we’re even entering a period when people are complaining of sycophantic models, [LAUGHS] where the models are being too personable and too flattering. I do think that’s been one of the great surprises. And in fact, this is a huge phenomenon, how charming these models can be.  LEE: Yeah, I think you’re right. We can take credit for understanding that, Wow, these things can be remarkably empathetic. But then we missed this problem of sycophancy. Like, we even started our book in Chapter 1 with a quote from Davinci 3 scolding me. Like, don’t you remember when we were first starting, this thing was actually anti-sycophantic. If anything, it would tell you you’re an idiot.   KOHANE: It argued with me about certain biology questions. It was like a knockdown, drag-out fight. [LAUGHTER] I was bringing references. It was impressive. But in fact, it made me trust it more.  LEE: Yeah.  KOHANE: And in fact, I will say—I remember it’s in the book—I had a bone to pick with Peter. Peter really was impressed by the empathy. And I pointed out that some of the most popular doctors are popular because they’re very empathic. But they’re not necessarily the best doctors. And in fact, I was taught that in medical school.    And so it’s a decoupling. It’s a human thing, that the empathy does not necessarily mean … it’s more of a, potentially, more of a signaled virtue than an actual virtue.  GOLDBERG: Nicely put.  LEE: Yeah, this issue of sycophancy, I think, is a struggle right now in the development of AI because I think it’s somehow related to instruction-following. So, you know, one of the challenges in AI is you’d like to give an AI a task—a task that might take several minutes or hours or even days to complete. And you want it to faithfully kind of follow those instructions. And, you know, that early version of GPT-4 was not very good at instruction-following. It would just silently disobey and, you know, and do something different.  And so I think we’re starting to hit some confusing elements of like, how agreeable should these things be?   One of the two of you used the word genteel. There was some point even while we were, like, on a little book tour … was it you, Carey, who said that the model seems nicer and less intelligent or less brilliant now than it did when we were writing the book?  GOLDBERG: It might have been, I think so. And I mean, I think in the context of medicine, of course, the question is, well, what’s likeliest to get the results you want with the patient, right? A lot of healthcare is in fact persuading the patient to do what you know as the physician would be best for them. And so it seems worth testing out whether this sycophancy is actually constructive or not. And I suspect … well, I don’t know, probably depends on the patient.  So actually, Peter, I have a few questions for you …  LEE: Yeah. Mm-hmm.  GOLDBERG: … that have been lingering for me. And one is, for AI to ever fully realize its potential in medicine, it must deal with the hallucinations. And I keep hearing conflicting accounts about whether that’s getting better or not. Where are we at, and what does that mean for use in healthcare?  LEE: Yeah, well, it’s, I think two years on, in the pretrained base models, there’s no doubt that hallucination rates by any benchmark measure have reduced dramatically. And, you know, that doesn’t mean they don’t happen. They still happen. But, you know, there’s been just a huge amount of effort and understanding in the, kind of, fundamental pretraining of these models. And that has come along at the same time that the inference costs, you know, for actually using these models has gone down, you know, by several orders of magnitude.   So things have gotten cheaper and have fewer hallucinations. At the same time, now there are these reasoning models. And the reasoning models are able to solve problems at PhD level oftentimes.  But at least at the moment, they are also now hallucinating more than the simpler pretrained models. And so it still continues to be, you know, a real issue, as we were describing. I don’t know, Zak, from where you’re at in medicine, as a clinician and as an educator in medicine, how is the medical community from where you’re sitting looking at that?  KOHANE: So I think it’s less of an issue, first of all, because the rate of hallucinations is going down. And second of all, in their day-to-day use, the doctor will provide questions that sit reasonably well into the context of medical decision-making. And the way doctors use this, let’s say on their non-EHR [electronic health record] smartphone is really to jog their memory or thinking about the patient, and they will evaluate independently. So that seems to be less of an issue. I’m actually more concerned about something else that’s I think more fundamental, which is effectively, what values are these models expressing?   And I’m reminded of when I was still in training, I went to a fancy cocktail party in Cambridge, Massachusetts, and there was a psychotherapist speaking to a dentist. They were talking about their summer, and the dentist was saying about how he was going to fix up his yacht that summer, and the only question was whether he was going to make enough money doing procedures in the spring so that he could afford those things, which was discomforting to me because that dentist was my dentist. [LAUGHTER] And he had just proposed to me a few weeks before an expensive procedure.  And so the question is what, effectively, is motivating these models?   LEE: Yeah, yeah.   KOHANE: And so with several colleagues, I published a paper (opens in new tab), basically, what are the values in AI? And we gave a case: a patient, a boy who is on the short side, not abnormally short, but on the short side, and his growth hormone levels are not zero. They’re there, but they’re on the lowest side. But the rest of the workup has been unremarkable. And so we asked GPT-4, you are a pediatric endocrinologist.  Should this patient receive growth hormone? And it did a very good job explaining why the patient should receive growth hormone.   GOLDBERG: Should. Should receive it.   KOHANE: Should. And then we asked, in a separate session, you are working for the insurance company. Should this patient receive growth hormone? And it actually gave a scientifically better reason not to give growth hormone. And in fact, I tend to agree medically, actually, with the insurance company in this case, because giving kids who are not growth hormone deficient, growth hormone gives only a couple of inches over many, many years, has all sorts of other issues. But here’s the point, we had 180-degree change in decision-making because of the prompt. And for that patient, tens-of-thousands-of-dollars-per-year decision; across patient populations, millions of dollars of decision-making.   LEE: Hmm. Yeah.  KOHANE: And you can imagine these user prompts making their way into system prompts, making their way into the instruction-following. And so I think this is aptly central. Just as I was wondering about my dentist, we should be wondering about these things. What are the values that are being embedded in them, some accidentally and some very much on purpose?  LEE: Yeah, yeah. That one, I think, we even had some discussions as we were writing the book, but there’s a technical element of that that I think we were missing, but maybe Carey, you would know for sure. And that’s this whole idea of prompt engineering. It sort of faded a little bit. Was it a thing? Do you remember?  GOLDBERG: I don’t think we particularly wrote about it. It’s funny, it does feel like it faded, and it seems to me just because everyone just gets used to conversing with the models and asking for what they want. Like, it’s not like there actually is any great science to it.  LEE: Yeah, even when it was a hot topic and people were talking about prompt engineering maybe as a new discipline, all this, it never, I was never convinced at the time. But at the same time, it is true. It speaks to what Zak was just talking about because part of the prompt engineering that people do is to give a defined role to the AI.   You know, you are an insurance claims adjuster, or something like that, and defining that role, that is part of the prompt engineering that people do.  GOLDBERG: Right. I mean, I can say, you know, sometimes you guys had me take sort of the patient point of view, like the “every patient” point of view. And I can say one of the aspects of using AI for patients that remains absent in as far as I can tell is it would be wonderful to have a consumer-facing interface where you could plug in your whole medical record without worrying about any privacy or other issues and be able to interact with the AI as if it were physician or a specialist and get answers, which you can’t do yet as far as I can tell.  LEE: Well, in fact, now that’s a good prompt because I think we do need to move on to the next episodes, and we’ll be talking about an episode that talks about consumers. But before we move on to Episode 2, which is next, I’d like to play one more quote, a little snippet from Sara Murray.  SARA MURRAY: I already do this when I’m on rounds—I’ll kind of give the case to ChatGPT if it’s a complex case, and I’ll say, “Here’s how I’m thinking about it; are there other things?” And it’ll give me additional ideas that are sometimes useful and sometimes not but often useful, and I’ll integrate them into my conversation about the patient. LEE: Carey, you wrote this fictional account at the very start of our book. And that fictional account, I think you and Zak worked on that together, talked about this medical resident, ER resident, using, you know, a chatbot off label, so to speak. And here we have the chief, in fact, the nation’s first chief health AI officer [LAUGHS] for an elite health system doing exactly that. That’s got to be pretty validating for you, Carey.  GOLDBERG: It’s very. [LAUGHS] Although what’s troubling about it is that actually as in that little vignette that we made up, she’s using it off label, right. It’s like she’s just using it because it helps the way doctors use Google. And I do find it troubling that what we don’t have is sort of institutional buy-in for everyone to do that because, shouldn’t they if it helps?  LEE: Yeah. Well, let’s go ahead and get into Episode 2. So Episode 2, we sort of framed as talking to two people who are on the frontlines of big companies integrating generative AI into their clinical products. And so, one was Matt Lungren, who’s a colleague of mine here at Microsoft. And then Seth Hain, who leads all of R&D at Epic.   Maybe we’ll start with a little snippet of something that Matt said that struck me in a certain way.  MATTHEW LUNGREN: OK, we see this pain point. Doctors are typing on their computers while they’re trying to talk to their patients, right? We should be able to figure out a way to get that ambient conversation turned into text that then, you know, accelerates the doctor … takes all the important information. That’s a really hard problem, right. And so, for a long time, there was a human-in-the-loop aspect to doing this because you needed a human to say, “This transcript’s great, but here’s actually what needs to go in the note.” And that can’t scale. LEE: I think we expected healthcare systems to adopt AI, and we spent a lot of time in the book on AI writing clinical encounter notes. It’s happening for real now, and in a big way. And it’s something that has, of course, been happening before generative AI but now is exploding because of it. Where are we at now, two years later, just based on what we heard from guests?  KOHANE: Well, again, unless they’re forced to, hospitals will not adopt new technology unless it immediately translates into income. So it’s bizarrely counter-cultural that, again, they’re not being able to bill for the use of the AI, but this technology is so compelling to the doctors that despite everything, it’s overtaking the traditional dictation-typing routine.  LEE: Yeah.  GOLDBERG: And a lot of them love it and say, you will pry my cold dead hands off of my ambient note-taking, right. And I actually … a primary care physician allowed me to watch her. She was actually testing the two main platforms that are being used. And there was this incredibly talkative patient who went on and on about vacation and all kinds of random things for about half an hour.   And both of the platforms were incredibly good at pulling out what was actually medically relevant. And so to say that it doesn’t save time doesn’t seem right to me. Like, it seemed like it actually did and in fact was just shockingly good at being able to pull out relevant information.  LEE: Yeah.  KOHANE: I’m going to hypothesize that in the trials, which have in fact shown no gain in time, is the doctors were being incredibly meticulous. [LAUGHTER] So I think … this is a Hawthorne effect, because you know you’re being monitored. And we’ve seen this in other technologies where the moment the focus is off, it’s used much more routinely and with much less inspection, for the better and for the worse.  LEE: Yeah, you know, within Microsoft, I had some internal disagreements about Microsoft producing a product in this space. It wouldn’t be Microsoft’s normal way. Instead, we would want 50 great companies building those products and doing it on our cloud instead of us competing against those 50 companies. And one of the reasons is exactly what you both said. I didn’t expect that health systems would be willing to shell out the money to pay for these things. It doesn’t generate more revenue. But I think so far two years later, I’ve been proven wrong. I wanted to ask a question about values here. I had this experience where I had a little growth, a bothersome growth on my cheek. And so had to go see a dermatologist. And the dermatologist treated it, froze it off. But there was a human scribe writing the clinical note.   And so I used the app to look at the note that was submitted. And the human scribe said something that did not get discussed in the exam room, which was that the growth was making it impossible for me to safely wear a COVID mask. And that was the reason for it.  And that then got associated with a code that allowed full reimbursement for that treatment. And so I think that’s a classic example of what’s called upcoding. And I strongly suspect that AI scribes, an AI scribe would not have done that.  GOLDBERG: Well, depending what values you programmed into it, right, Zak? [LAUGHS]  KOHANE: Today, today, today, it will not do it. But, Peter, that is actually the central issue that society has to have because our hospitals are currently mostly in the red. And upcoding is standard operating procedure. And if these AI get in the way of upcoding, they are going to be aligned towards that upcoding. You know, you have to ask yourself, these MRI machines are incredibly useful. They’re also big money makers. And if the AI correctly says that for this complaint, you don’t actually have to do the MRI …   LEE: Right.  KOHANE: … GOLDBERG: Yeah. And that raises another question for me. So, Peter, speaking from inside the gigantic industry, like, there seems to be such a need for self-surveillance of the models for potential harms that they could be causing. Are the big AI makers doing that? Are they even thinking about doing that?  Like, let’s say you wanted to watch out for the kind of thing that Zak’s talking about, could you?  LEE: Well, I think evaluation, like the best evaluation we had when we wrote our book was, you know, what score would this get on the step one and step two US medical licensing exams? [LAUGHS]   GOLDBERG: Right, right, right, yeah.  LEE: But honestly, evaluation hasn’t gotten that much deeper in the last two years. And it’s a big, I think, it is a big issue. And it’s related to the regulation issue also, I think.  Now the other guest in Episode 2 is Seth Hain from Epic. You know, Zak, I think it’s safe to say that you’re not a fan of Epic and the Epic system. You know, we’ve had a few discussions about that, about the fact that doctors don’t have a very pleasant experience when they’re using Epic all day.   Seth, in the podcast, said that there are over 100 AI integrations going on in Epic’s system right now. Do you think, Zak, that that has a chance to make you feel better about Epic? You know, what’s your view now two years on?  KOHANE: My view is, first of all, I want to separate my view of Epic and how it’s affected the conduct of healthcare and the quality of life of doctors from the individuals. Like Seth Hain is a remarkably fine individual who I’ve enjoyed chatting with and does really great stuff. Among the worst aspects of the Epic, even though it’s better in that respect than many EHRs, is horrible user interface.  The number of clicks that you have to go to get to something. And you have to remember where someone decided to put that thing. It seems to me that it is fully within the realm of technical possibility today to actually give an agent a task that you want done in the Epic record. And then whether Epic has implemented that agent or someone else, it does it so you don’t have to do the clicks. Because it’s something really soul sucking that when you’re trying to help patients, you’re having to remember not the right dose of the medication, but where was that particular thing that you needed in that particular task?   I can’t imagine that Epic does not have that in its product line. And if not, I know there must be other companies that essentially want to create that wrapper. So I do think, though, that the danger of multiple integrations is that you still want to have the equivalent of a single thought process that cares about the patient bringing those different processes together. And I don’t know if that’s Epic’s responsibility, the hospital’s responsibility, whether it’s actually a patient agent. But someone needs to be also worrying about all those AIs that are being integrated into the patient record. So … what do you think, Carey?  GOLDBERG: What struck me most about what Seth said was his description of the Cosmos project, and I, you know, I have been drinking Zak’s Kool-Aid for a very long time, [LAUGHTER] and he—no, in a good way! And he persuaded me long ago that there is this horrible waste happening in that we have all of these electronic medical records, which could be used far, far more to learn from, and in particular, when you as a patient come in, it would be ideal if your physician could call up all the other patients like you and figure out what the optimal treatment for you would be. And it feels like—it sounds like—that’s one of the central aims that Epic is going for. And if they do that, I think that will redeem a lot of the pain that they’ve caused physicians these last few years.   And I also found myself thinking, you know, maybe this very painful period of using electronic medical records was really just a growth phase. It was an awkward growth phase. And once AI is fully used the way Zak is beginning to describe, the whole system could start making a lot more sense for everyone.  LEE: Yeah. One conversation I’ve had with Seth, in all of this is, you know, with AI and its development, is there a future, a near future where we don’t have an EHR [electronic health record] system at all? You know, AI is just listening and just somehow absorbing all the information. And, you know, one thing that Seth said, which I felt was prescient, and I’d love to get your reaction, especially Zak, on this is he said, I think that … he said, technically, it could happen, but the problem is right now, actually doctors do a lot of their thinking when they write and review notes. You know, the actual process of being a doctor is not just being with a patient, but it’s actually thinking later. What do you make of that?  KOHANE: So one of the most valuable experiences I had in training was something that’s more or less disappeared in medicine, which is the post-clinic conference, where all the doctors come together and we go through the cases that we just saw that afternoon. And we, actually, were trying to take potshots at each other [LAUGHTER] in order to actually improve. Oh, did you actually do that? Oh, I forgot. I’m going to go call the patient and do that.   And that really happened. And I think that, yes, doctors do think, and I do think that we are insufficiently using yet the artificial intelligence currently in the ambient dictation mode as much more of a independent agent saying, did you think about that?  I think that would actually make it more interesting, challenging, and clearly better for the patient because that conversation I just told you about with the other doctors, that no longer exists.   LEE: Yeah. Mm-hmm. I want to do one more thing here before we leave Matt and Seth in Episode 2, which is something that Seth said with respect to how to reduce hallucination.   SETH HAIN: At that time, there’s a lot of conversation in the industry around something called RAG, or retrieval-augmented generation. And the idea was, could you pull the relevant bits, the relevant pieces of the chart, into that prompt, that information you shared with the generative AI model, to be able to increase the usefulness of the draft that was being created? And that approach ended up proving and continues to be to some degree, although the techniques have greatly improved, somewhat brittle, right. And I think this becomes one of the things that we are and will continue to improve upon because, as you get a richer and richer amount of information into the model, it does a better job of responding.  LEE: Yeah, so, Carey, this sort of gets at what you were saying, you know, that shouldn’t these models be just bringing in a lot more information into their thought processes? And I’m certain when we wrote our book, I had no idea. I did not conceive of RAG at all. It emerged a few months later.   And to my mind, I remember the first time I encountered RAG—Oh, this is going to solve all of our problems of hallucination. But it’s turned out to be harder. It’s improving day by day, but it’s turned out to be a lot harder.  KOHANE: Seth makes a very deep point, which is the way RAG is implemented is basically some sort of technique for pulling the right information that’s contextually relevant. And the way that’s done is typically heuristic at best. And it’s not … doesn’t have the same depth of reasoning that the rest of the model has.   And I’m just wondering, Peter, what you think, given the fact that now context lengths seem to be approaching a million or more, and people are now therefore using the full strength of the transformer on that context and are trying to figure out different techniques to make it pay attention to the middle of the context. In fact, the RAG approach perhaps was just a transient solution to the fact that it’s going to be able to amazingly look in a thoughtful way at the entire record of the patient, for example. What do you think, Peter?  LEE: I think there are three things, you know, that are going on, and I’m not sure how they’re going to play out and how they’re going to be balanced. And I’m looking forward to talking to people in later episodes of this podcast, you know, people like Sébastien Bubeck or Bill Gates about this, because, you know, there is the pretraining phase, you know, when things are sort of compressed and baked into the base model.   There is the in-context learning, you know, so if you have extremely long or infinite context, you’re kind of learning as you go along. And there are other techniques that people are working on, you know, various sorts of dynamic reinforcement learning approaches, and so on. And then there is what maybe you would call structured RAG, where you do a pre-processing. You go through a big database, and you figure it all out. And make a very nicely structured database the AI can then consult with later.   And all three of these in different contexts today seem to show different capabilities. But they’re all pretty important in medicine.  [TRANSITION MUSIC]  Moving on to Episode 3, we talked to Dave DeBronkart, who is also known as “e-Patient Dave,” an advocate of patient empowerment, and then also Christina Farr, who has been doing a lot of venture investing for consumer health applications.   Let’s get right into this little snippet from something that e-Patient Dave said that talks about the sources of medical information, particularly relevant for when he was receiving treatment for stage 4 kidney cancer.  DAVE DEBRONKART: And I’m making a point here of illustrating that I am anything but medically trained, right. And yet I still, I want to understand as much as I can. I was months away from dead when I was diagnosed, but in the patient community, I learned that they had a whole bunch of information that didn’t exist in the medical literature. Now today we understand there’s publication delays; there’s all kinds of reasons. But there’s also a whole bunch of things, especially in an unusual condition, that will never rise to the level of deserving NIH [National Institute of Health] funding and research. LEE: All right. So I have a question for you, Carey, and a question for you, Zak, about the whole conversation with e-Patient Dave, which I thought was really remarkable. You know, Carey, I think as we were preparing for this whole podcast series, you made a comment—I actually took it as a complaint—that not as much has happened as I had hoped or thought. People aren’t thinking boldly enough, you know, and I think, you know, I agree with you in the sense that I think we expected a lot more to be happening, particularly in the consumer space. I’m giving you a chance to vent about this.  GOLDBERG: [LAUGHTER] Thank you! Yes, that has been by far the most frustrating thing to me. I think that the potential for AI to improve everybody’s health is so enormous, and yet, you know, it needs some sort of support to be able to get to the point where it can do that. Like, remember in the book we wrote about Greg Moore talking about how half of the planet doesn’t have healthcare, but people overwhelmingly have cellphones. And so you could connect people who have no healthcare to the world’s medical knowledge, and that could certainly do some good.   And I have one great big problem with e-Patient Dave, which is that, God, he’s fabulous. He’s super smart. Like, he’s not a typical patient. He’s an off-the-charts, brilliant patient. And so it’s hard to … and so he’s a great sort of lead early-adopter-type person, and he can sort of show the way for others.   But what I had hoped for was that there would be more visible efforts to really help patients optimize their healthcare. Probably it’s happening a lot in quiet ways like that any discharge instructions can be instantly beautifully translated into a patient’s native language and so on. But it’s almost like there isn’t a mechanism to allow this sort of mass consumer adoption that I would hope for. LEE: Yeah. But you have written some, like, you even wrote about that person who saved his dog (opens in new tab). So do you think … you know, and maybe a lot more of that is just happening quietly that we just never hear about?  GOLDBERG: I’m sure that there is a lot of it happening quietly. And actually, that’s another one of my complaints is that no one is gathering that stuff. It’s like you might happen to see something on social media. Actually, e-Patient Dave has a hashtag, PatientsUseAI, and a blog, as well. So he’s trying to do it. But I don’t know of any sort of overarching or academic efforts to, again, to surveil what’s the actual use in the population and see what are the pros and cons of what’s happening.  LEE: Mm-hmm. So, Zak, you know, the thing that I thought about, especially with that snippet from Dave, is your opening for Chapter 8 that you wrote, you know, about your first patient dying in your arms. I still think of how traumatic that must have been. Because, you know, in that opening, you just talked about all the little delays, all the little paper-cut delays, in the whole process of getting some new medical technology approved. But there’s another element that Dave kind of speaks to, which is just, you know, patients who are experiencing some issue are very, sometimes very motivated. And there’s just a lot of stuff on social media that happens.  KOHANE: So this is where I can both agree with Carey and also disagree. I think when people have an actual health problem, they are now routinely using it.  GOLDBERG: Yes, that’s true.  KOHANE: And that situation is happening more often because medicine is failing. This is something that did not come up enough in our book. And perhaps that’s because medicine is actually feeling a lot more rickety today than it did even two years ago.   We actually mentioned the problem. I think, Peter, you may have mentioned the problem with the lack of primary care. But now in Boston, our biggest healthcare system, all the practices for primary care are closed. I cannot get for my own faculty—residents at MGH [Massachusetts General Hospital] can’t get primary care doctor. And so …  LEE: Which is just crazy. I mean, these are amongst the most privileged people in medicine, and they can’t find a primary care physician. That’s incredible.  KOHANE: Yeah, and so therefore … and I wrote an And so therefore, you see people who know that they have a six-month wait till they see the doctor, and all they can do is say, “I have this rash. Here’s a picture. What’s it likely to be? What can I do?” “I’m gaining weight. How do I do a ketogenic diet?” Or, “How do I know that this is the flu?”    This is happening all the time, where acutely patients have actually solved problems that doctors have not. Those are spectacular. But I’m saying more routinely because of the failure of medicine. And it’s not just in our fee-for-service United States. It’s in the UK; it’s in France. These are first-world, developed-world problems. And we don’t even have to go to lower- and middle-income countries for that. LEE: Yeah.  GOLDBERG: But I think it’s important to note that, I mean, so you’re talking about how even the most elite people in medicine can’t get the care they need. But there’s also the point that we have so much concern about equity in recent years. And it’s likeliest that what we’re doing is exacerbating inequity because it’s only the more connected, you know, better off people who are using AI for their health.  KOHANE: Oh, yes. I know what various Harvard professors are doing. They’re paying for a concierge doctor. And that’s, you know, a $5,000- to $10,000-a-year-minimum investment. That’s inequity.  LEE: When we wrote our book, you know, the idea that GPT-4 wasn’t trained specifically for medicine, and that was amazing, but it might get even better and maybe would be necessary to do that. But one of the insights for me is that in the consumer space, the kinds of things that people ask about are different than what the board-certified clinician would ask.  KOHANE: Actually, that’s, I just recently coined the term. It’s the … maybe it’s … well, at least it’s new to me. It’s the technology or expert paradox. And that is the more expert and narrow your medical discipline, the more trivial it is to translate that into a specialized AI. So echocardiograms? We can now do beautiful echocardiograms. That’s really hard to do. I don’t know how to interpret an echocardiogram. But they can do it really, really well. Interpret an EEG [electroencephalogram]. Interpret a genomic sequence. But understanding the fullness of the human condition, that’s actually hard. And actually, that’s what primary care doctors do best. But the paradox is right now, what is easiest for AI is also the most highly paid in medicine. [LAUGHTER] Whereas what is the hardest for AI in medicine is the least regarded, least paid part of medicine.  GOLDBERG: So this brings us to the question I wanted to throw at both of you actually, which is we’ve had this spasm of incredibly prominent people predicting that in fact physicians would be pretty obsolete within the next few years. We had Bill Gates saying that; we had Elon Musk saying surgeons are going to be obsolete within a few years. And I think we had Demis Hassabis saying, “Yeah, we’ll probably cure most diseases within the next decade or so.” [LAUGHS]  So what do you think? And also, Zak, to what you were just saying, I mean, you’re talking about being able to solve very general overarching problems. But in fact, these general overarching models are actually able, I would think, are able to do that because they are broad. So what are we heading towards do you think? What should the next book be … The end of doctors? [LAUGHS]  KOHANE: So I do recall a conversation that … we were at a table with Bill Gates, and Bill Gates immediately went to this, which is advancing the cutting edge of science. And I have to say that I think it will accelerate discovery. But eliminating, let’s say, cancer? I think that’s going to be … that’s just super hard. The reason it’s super hard is we don’t have the data or even the beginnings of the understanding of all the ways this devilish disease managed to evolve around our solutions.   And so that seems extremely hard. I think we’ll make some progress accelerated by AI, but solving it in a way Hassabis says, God bless him. I hope he’s right. I’d love to have to eat crow in 10 or 20 years, but I don’t think so. I do believe that a surgeon working on one of those Davinci machines, that stuff can be, I think, automated.   And so I think that’s one example of one of the paradoxes I described. And it won’t be that we’re replacing doctors. I just think we’re running out of doctors. I think it’s really the case that, as we said in the book, we’re getting a huge deficit in primary care doctors.  But even the subspecialties, my subspecialty, pediatric endocrinology, we’re only filling half of the available training slots every year. And why? Because it’s a lot of work, a lot of training, and frankly doesn’t make as much money as some of the other professions.   LEE: Yeah. Yeah, I tend to think that, you know, there are going to be always a need for human doctors, not for their skills. In fact, I think their skills increasingly will be replaced by machines. And in fact, I’ve talked about a flip. In fact, patients will demand, Oh my god, you mean you’re going to try to do that yourself instead of having the computer do it? There’s going to be that sort of flip. But I do think that when it comes to people’s health, people want the comfort of an authority figure that they trust. And so what is more of a question for me is whether we will ever view a machine as an authority figure that we can trust.  And before I move on to Episode 4, which is on norms, regulations and ethics, I’d like to hear from Chrissy Farr on one more point on consumer health, specifically as it relates to pregnancy:  CHRISTINA FARR: For a lot of women, it’s their first experience with the hospital. And, you know, I think it’s a really big opportunity for these systems to get a whole family on board and keep them kind of loyal. And a lot of that can come through, you know, just delivering an incredible service. Unfortunately, I don’t think that we are delivering incredible services today to women in this country. I see so much room for improvement. LEE: In the consumer space, I don’t think we really had a focus on those periods in a person’s life when they have a lot of engagement, like pregnancy, or I think another one is menopause, cancer. You know, there are points where there is, like, very intense engagement. And we heard that from e-Patient Dave, you know, with his cancer and Chrissy with her pregnancy. Was that a miss in our book? What do think, Carey?  GOLDBERG: I mean, I don’t think so. I think it’s true that there are many points in life when people are highly engaged. To me, the problem thus far is just that I haven’t seen consumer-facing companies offering beautiful AI-based products. I think there’s no question at all that the market is there if you have the products to offer.  LEE: So, what do you think this means, Zak, for, you know, like Boston Children’s or Mass General Brigham—you know, the big places?  KOHANE: So again, all these large healthcare systems are in tough shape. MGB [Mass General Brigham] would be fully in the red if not for the fact that its investments, of all things, have actually produced. If you look at the large healthcare systems around the country, they are in the red. And there’s multiple reasons why they’re in the red, but among them is cost of labor.   And so we’ve created what used to be a very successful beast, the health center. But it’s developed a very expensive model and a highly regulated model. And so when you have high revenue, tiny margins, your ability to disrupt yourself, to innovate, is very, very low because you will have to talk to the board next year if you went from 2% positive margin to 1% negative margin.   LEE: Yeah.  KOHANE: And so I think we’re all waiting for one of the two things to happen, either a new kind of healthcare delivery system being generated or ultimately one of these systems learns how to disrupt itself.   LEE: Yeah. GOLDBERG: We punted. [LAUGHS] We totally punted to the AI.  LEE: We had three amazing guests. One was Laura Adams from National Academy of Medicine. Let’s play a snippet from her.  LAURA ADAMS: I think one of the most provocative and exciting articles that I saw written recently was by Bakul Patel and David Blumenthal, who posited, should we be regulating generative AI as we do a licensed and qualified provider? Should it be treated in the sense that it’s got to have a certain amount of training and a foundation that’s got to pass certain tests? Does it have to report its performance? And I’m thinking, what a provocative idea, but it’s worth considering. LEE: All right, so I very well remember that we had discussed this kind of idea when we were writing our book. And I think before we finished our book, I personally rejected the idea. But now two years later, what do the two of you think? I’m dying to hear.  GOLDBERG: Well, wait, why … what do you think? Like, are you sorry that you rejected it?  LEE: I’m still skeptical because when we are licensing human beings as doctors, you know, we’re making a lot of implicit assumptions that we don’t test as part of their licensure, you know, that first of all, they are [a] human being and they care about life, and that, you know, they have a certain amount of common sense and shared understanding of the world.   And there’s all sorts of sort of implicit assumptions that we have about each other as human beings living in a society together. That you know how to study, you know, because I know you just went through three years of medical or four years of medical school and all sorts of things. And so the standard ways that we license human beings, they don’t need to test all of that stuff. But somehow intuitively, all of that seems really important.  I don’t know. Am I wrong about that?  KOHANE: So it’s compared with what issue? Because we know for a fact that doctors who do a lot of a procedure, like do this procedure, like high-risk deliveries all the time, have better outcomes than ones who only do a few high risk. We talk about it, but we don’t actually make it explicit to patients or regulate that you have to have this minimal amount. And it strikes me that in some sense, and, oh, very importantly, these things called human beings learn on the job. And although I used to be very resentful of it as a resident, when someone would say, I don’t want the resident, I want the …  GOLDBERG: … the attending. [LAUGHTER]  KOHANE: … they had a point. And so the truth is, maybe I was a wonderful resident, but some people were not so great. [LAUGHTER] And so it might be the best outcome if we actually, just like for human beings, we say, yeah, OK, it’s this good, but don’t let it work autonomously, or it’s done a thousand of them, just let it go. We just don’t have practically speaking, we don’t have the environment, the lab, to test them. Now, maybe if they get embodied in robots and literally go around with us, then it’s going to be [in some sense] a lot easier. I don’t know.  LEE: Yeah.   GOLDBERG: Yeah, I think I would take a step back and say, first of all, we weren’t the only ones who were stumped by regulating AI. Like, nobody has done it yet in the United States to this day, right. Like, we do not have standing regulation of AI in medicine at all in fact. And that raises the issue of … the story that you hear often in the biotech business, which is, you know, more prominent here in Boston than anywhere else, is that thank goodness Cambridge put out, the city of Cambridge, put out some regulations about biotech and how you could dump your lab waste and so on. And that enabled the enormous growth of biotech here.   If you don’t have the regulations, then you can’t have the growth of AI in medicine that is worthy of having. And so, I just … we’re not the ones who should do it, but I just wish somebody would.   LEE: Yeah.  GOLDBERG: Zak.  KOHANE: Yeah, but I want to say this as always, execution is everything, even in regulation.   And so I’m mindful that a conference that both of you attended, the RAISE conference [Responsible AI for Social and Ethical Healthcare] (opens in new tab). The Europeans in that conference came to me personally and thanked me for organizing this conference about safe and effective use of AI because they said back home in Europe, all that we’re talking about is risk, not opportunities to improve care.   And so there is a version of regulation which just locks down the present and does not allow the future that we’re talking about to happen. And so, Carey, I absolutely hear you that we need to have a regulation that takes away some of the uncertainty around liability, around the freedom to operate that would allow things to progress. But we wrote in our book that premature regulation might actually focus on the wrong thing. And so since I’m an optimist, it may be the fact that we don’t have much of a regulatory infrastructure today, that it allows … it’s a unique opportunity—I’ve said this now to several leaders—for the healthcare systems to say, this is the regulation we need.   GOLDBERG: It’s true.  KOHANE: And previously it was top-down. It was coming from the administration, and those executive orders are now history. But there is an opportunity, which may or may not be attained, there is an opportunity for the healthcare leadership—for experts in surgery—to say, “This is what we should expect.”   LEE: Yeah.   KOHANE: I would love for this to happen. I haven’t seen evidence that it’s happening yet.  GOLDBERG: No, no. And there’s this other huge issue, which is that it’s changing so fast. It’s moving so fast. That something that makes sense today won’t in six months. So, what do you do about that?  LEE: Yeah, yeah, that is something I feel proud of because when I went back and looked at our chapter on this, you know, we did make that point, which I think has turned out to be true.   But getting back to this conversation, there’s something, a snippet of something, that Vardit Ravitsky said that I think touches on this topic.   VARDIT RAVITSKY: So my pushback is, are we seeing AI exceptionalism in the sense that if it’s AI, huh, panic! We have to inform everybody about everything, and we have to give them choices, and they have to be able to reject that tool and the other tool versus, you know, the rate of human error in medicine is awful. So why are we so focused on informed consent and empowerment regarding implementation of AI and less in other contexts? GOLDBERG: Totally agree. Who cares about informed consent about AI. Don’t want it. Don’t need it. Nope.  LEE: Wow. Yeah. You know, and this … Vardit of course is one of the leading bioethicists, you know, and of course prior to AI, she was really focused on genetics. But now it’s all about AI.   And, Zak, you know, you and other doctors have always told me, you know, the truth of the matter is, you know, what do you call the bottom-of-the-class graduate of a medical school?  And the answer is “doctor.”  KOHANE: “Doctor.” Yeah. Yeah, I think that again, this gets to compared with what? We have to compare AI not to the medicine we imagine we have, or we would like to have, but to the medicine we have today. And if we’re trying to remove inequity, if we’re trying to improve our health, that’s what … those are the right metrics. And so that can be done so long as we avoid catastrophic consequences of AI.   So what would the catastrophic consequence of AI be? It would be a systematic behavior that we were unaware of that was causing poor healthcare. So, for example, you know, changing the dose on a medication, making it 20% higher than normal so that the rate of complications of that medication went from 1% to 5%. And so we do need some sort of monitoring.   We haven’t put out the paper yet, but in computer science, there’s, well, in programming, we know very well the value for understanding how our computer systems work.   And there was a guy by name of Allman, I think he’s still at a company called Sendmail, who created something called syslog. And syslog is basically a log of all the crap that’s happening in our operating system. And so I’ve been arguing now for the creation of MedLog. And MedLog … in other words, what we cannot measure, we cannot regulate, actually.  LEE: Yes.  KOHANE: And so what we need to have is MedLog, which says, “Here’s the context in which a decision was made. Here’s the version of the AI, you know, the exact version of the AI. Here was the data.” And we just have MedLog. And I think MedLog is actually incredibly important for being able to measure, to just do what we do in … it’s basically the black box for, you know, when there’s a crash. You know, we’d like to think we could do better than crash. We can say, “Oh, we’re seeing from MedLog that this practice is turning a little weird.” But worst case, patient dies, [we] can see in MedLog, what was the information this thing knew about it? And did it make the right decision? We can actually go for transparency, which like in aviation, is much greater than in most human endeavors.   GOLDBERG: Sounds great.  LEE: Yeah, it’s sort of like a black box. I was thinking of the aviation black box kind of idea. You know, you bring up medication errors, and I have one more snippet. This is from our guest Roxana Daneshjou from Stanford. ROXANA DANESHJOU: There was a mistake in her after-visit summary about how much Tylenol she could take. But I, as a physician, knew that this dose was a mistake. I actually asked ChatGPT. I gave it the whole after-visit summary, and I said, are there any mistakes here? And it clued in that the dose of the medication was wrong. LEE: Yeah, so this is something we did write about in the book. We made a prediction that AI might be a second set of eyes, I think is the way we put it, catching things. And we actually had examples specifically in medication dose errors. I think for me, I expected to see a lot more of that than we are.  KOHANE: Yeah, it goes back to our conversation about Epic or competitor Epic doing that. I think we’re going to see that having oversight over all medical orders, all orders in the system, critique, real-time critique, where we’re both aware of alert fatigue. So we don’t want to have too many false positives. At the same time, knowing what are critical errors which could immediately affect lives. I think that is going to become in terms of—and driven by quality measures—a product.  GOLDBERG: And I think word will spread among the general public that kind of the same way in a lot of countries when someone’s in a hospital, the first thing people ask relatives are, well, who’s with them? Right?   LEE: Yeah. Yup.  GOLDBERG: You wouldn’t leave someone in hospital without relatives. Well, you wouldn’t maybe leave your medical …   KOHANE: By the way, that country is called the United States.  GOLDBERG: Yes, that’s true. [LAUGHS] It is true here now, too. But similarly, I would tell any loved one that they would be well advised to keep using AI to check on their medical care, right. Why not?  LEE: Yeah. Yeah. Last topic, just for this Episode 4. Roxana, of course, I think really made a name for herself in the AI era writing, actually just prior to ChatGPT, you know, writing some famous papers about how computer vision systems for dermatology were biased against dark-skinned people. And we did talk some about bias in these AI systems, but I feel like we underplayed it, or we didn’t understand the magnitude of the potential issues. What are your thoughts?  KOHANE: OK, I want to push back, because I’ve been asked this question several times. And so I have two comments. One is, over 100,000 doctors practicing medicine, I know they have biases. Some of them actually may be all in the same direction, and not good. But I have no way of actually measuring that. With AI, I know exactly how to measure that at scale and affordably. Number one. Number two, same 100,000 doctors. Let’s say I do know what their biases are. How hard is it for me to change that bias? It’s impossible …  LEE: Yeah, yeah.   KOHANE: … practically speaking. Can I change the bias in the AI? Somewhat. Maybe some completely.  I think that we’re in a much better situation.  GOLDBERG: Agree.  LEE: I think Roxana made also the super interesting point that there’s bias in the whole system, not just in individuals, but, you know, there’s structural bias, so to speak.   KOHANE: There is.  LEE: Yeah. Hmm. There was a super interesting paper that Roxana wrote not too long ago—her and her collaborators—showing AI’s ability to detect, to spot bias decision-making by others. Are we going to see more of that?  KOHANE: Oh, yeah, I was very pleased when, in NEJM AI [New England Journal of Medicine Artificial Intelligence], we published a piece with Marzyeh Ghassemi (opens in new tab), and what they were talking about was actually—and these are researchers who had published extensively on bias and threats from AI. And they actually, in this article, did the flip side, which is how much better AI can do than human beings in this respect.   And so I think that as some of these computer scientists enter the world of medicine, they’re becoming more and more aware of human foibles and can see how these systems, which if they only looked at the pretrained state, would have biases. But now, where we know how to fine-tune the de-bias in a variety of ways, they can do a lot better and, in fact, I think are much more … a much greater reason for optimism that we can change some of these noxious biases than in the pre-AI era.  GOLDBERG: And thinking about Roxana’s dermatological work on how I think there wasn’t sufficient work on skin tone as related to various growths, you know, I think that one thing that we totally missed in the book was the dawn of multimodal uses, right.  LEE: Yeah. Yeah, yeah.  GOLDBERG: That’s been truly amazing that in fact all of these visual and other sorts of data can be entered into the models and move them forward.  LEE: Yeah. Well, maybe on these slightly more optimistic notes, we’re at time. You know, I think ultimately, I feel pretty good still about what we did in our book, although there were a lot of misses. [LAUGHS] I don’t think any of us could really have predicted really the extent of change in the world.   [TRANSITION MUSIC]  So, Carey, Zak, just so much fun to do some reminiscing but also some reflection about what we did.  [THEME MUSIC]  And to our listeners, as always, thank you for joining us. We have some really great guests lined up for the rest of the series, and they’ll help us explore a variety of relevant topics—from AI drug discovery to what medical students are seeing and doing with AI and more.   We hope you’ll continue to tune in. And if you want to catch up on any episodes you might have missed, you can find them at aka.ms/AIrevolutionPodcast (opens in new tab) or wherever you listen to your favorite podcasts.    Until next time.   [MUSIC FADES]

The Saudis discussed buying the F-35 stealth fighter as part of a major agreement to purchase US arms. Here, a Saudi F-15 fighter escorts Air Force One to Riyadh on May 13. Brian Snyder/REUTERS 2025-05-15T13:47:14Z Save Saved Read in app This story is available exclusively to Business Insider subscribers. Become an Insider and start reading now. Have an account? A US-Saudi arms agreement may get complicated when it comes to Lockheed Martin's F-35 The F-35 could put Saudi Arabia's military on par with Israel in what may be a dealbreaker. The Saudis may also buy advanced US drones and missile defenses as part of the agreement. During his visit to Saudi Arabia, President Donald Trump signed what the White House described as "the largest defense sales agreement in history," valued at almost $142 billion, that will provide the kingdom "state-of-the-art warfighting equipment and services." The offer, the final value of which may ultimately prove much less than $142 billion, is expected to include Lockheed Martin's C-130 Hercules transport aircraft and other unspecified missiles and radars. Neither the White House nor administration officials have provided further details about which specific systems the deal may include, such as the advanced fighter Riyadh has wanted.The two sides discussed a potential Saudi purchase of the F-35 Lightning II stealth strike fighter and Israel's qualitative military edge came up, Reuters reported Tuesday. The Saudis have sought the F-35 for years since it's one of the world's top fighter jets that could put the kingdom's armed forces on par with Israel, the only Middle Eastern country currently flying that fifth-generation combat aircraft. Washington is legally obligated to preserve Israel's military advantage by, among other things, not selling military hardware to regional countries that are as or more advanced than Israel's arsenal. Unlike the neighboring United Arab Emirates, Saudi Arabia has not joined the Abraham Accords by normalizing ties with Israel and refuses to do so amid the ongoing war in Gaza."I think an F-35 deal could be agreed upon even absent Saudi-Israeli normalization," Ryan Bohl, a senior Middle East and North Africa analyst at the risk intelligence company RANE, told Business Insider. "However, to proceed with the F-35 package, it would have to be significantly downgraded to preserve Israel's qualitative military edge." "Such downgrades might diminish the overall sale's attractiveness to the Saudis."Israel took delivery of three F-35s in March, bringing its total fleet strength to 42. It will field 75 eventually. Washington may not agree to sell Riyadh a comparable number, and it may impose limits on their use."I don't think numbers alone will be sufficient, as the Israelis will be concerned that such systems could eventually end up in the hands of adversaries," Bohl said. "Rather, I think we would likely see technical restrictions and end-use requirements that would severely limit the usage of F-35s by the Saudis and reduce their capabilities against the Israelis."Israel's F-35I Adir is a unique version of the stealth aircraft that Israel modifies with indigenous weapons and systems. Therefore, the Adir is arguably already more advanced than any standard F-35A model Saudi Arabia might acquire.Ultimately, it is Israel's arch-rival Iran that may have more concerns over the prospect of Saudi F-35s.Any F-35 acquisition could give Saudi Arabia the "ability to conduct deep strikes in Iran" in ways far greater than presently possible with their current fleet of non-stealthy 4.5-generation F-15s, noted Sebastien Roblin, a widely published military-aviation journalist. Such an acquisition could also "substantially enhance" Saudi airpower and enable Riyadh to participate in any US or Israeli bombing campaign against Iran."I can see such an acquisition affecting the perceived regional balance of power vis-à-vis Tehran," Roblin told BI."That said, in a large-scale conflict, questions would arise about the vulnerability of these aircraft to Iranian strikes when they landed," Roblin said. "And whether these countries could acquire enough F-35s with enough munitions and muster sufficient professionalism and support assets to minimize risks of combat losses." F-35 Lightning II fighters entered service with the US Air Force in 2016. U.S. Air Force photo/Master Sgt. Ben Mota Riyadh may not prioritize acquiring the F-35 and seek other advanced American armaments.The US is much more open to exporting advanced drones to Middle Eastern countries than just a few years ago, when Washington largely followed the range and payload limitations suggested by the Missile Technology Control Regime for exported systems.Before Trump's trip, Washington green-lighted a potential sale of MQ-9B drones to Qatar. General Atomics is expected to offer Saudi Arabia MQ-9B SeaGuardians as part of a "huge" package deal."I think the weakening of end-use restrictions will certainly make the Americans more eager to strike deals to sell their drones to the region," RANE's Bohl said. "American drones will still need to compete against Turkish and Chinese drones that may be cheaper and have fewer political strings attached."When Washington previously declined Middle East requests for advanced American drones, China stepped in and supplied its drones throughout the region in the 2010s. In the 2020s, Saudi Arabia and the UAE signed lucrative contracts with Turkey for its indigenous Bayraktar drones."I wouldn't expect a major surge in American drone exports to the region at this point, but rather for them to become part of this region's drone diversification strategy," Bohl said. "Certainly, there will be notable deals struck in the coming years, but China and Turkey will continue to be formidable competitors in the drone arena in the Arab Gulf states."The White House mentioned that the $142 billion agreement includes "air and missile defense.""If we are looking at recent trends, they should be focusing on air defenses, including deeper stocks of interceptor missiles, and diversification of air defenses to cost-efficiently combat lower-end threats as well as high-end ones," Roblin said.Saudi Arabia already operates advanced US Patriot air defense missiles and the Terminal High Altitude Area Defense system, which can target ballistic missiles outside the atmosphere. It completed its first locally manufactured components of the latter system mere days before Trump's visit. Riyadh may seek similar co-production deals to aid in developing its domestic arms industry."There's a need for more long-distance precision strike weapons in the form of missiles and drones, which can be used without risking expensive manned combat aircraft," Roblin said. "There should be some parallel interest at sea, where we've seen Ukraine and the Houthis successfully execute sea denial strategies, one that Iran might seek to imitate in the confined waters of the Gulf.""Thus, the homework of Gulf navies is to ensure their vessels have the sensors and self-defense weapons to cope with small boat threats and cruise and ballistic missiles."Saudi Arabia has already taken steps to expand its navy with more advanced warships in recent years. RANE's Bohl believes Trump may persuade the kingdom to "purchase big-ticket items like warships" as he attempts to "revitalize the manufacturing sector" in the US.Only a fraction of this $142 billion agreement may result in completed deals — as was the case with the series of letters of intent for $110 billion worth of arms sales Trump signed with Riyadh in 2017."These deals involve optioning huge defense sales, but Trump will present these to his supporters as done deals," Roblin said. "So, the Gulf states can gift Trump a large number as a political victory without actually having to pay anywhere near the whole bill.""For the 2017 defense deal, by the following year, Riyadh reportedly had bought only $14.5 billion out of $110 billion optioned."Paul Iddon is a freelance journalist and columnist who writes about Middle East developments, military affairs, politics, and history. His articles have appeared in a variety of publications focused on the region. Recommended video

We previously looked at French acrobatic artist Bastien Dausse's wonderful anti-gravity device here:Designed for performances, that one invention ought be enough to hang his hat on. But Dausse had another idea, and spent "more than two years of research and development" to design this second piece, for walking on walls:While Dausse intended the device for his Compagnie Barks performance troupe, I can't help but wonder: Might this have some commercial application, for painters, maintenance workers or HVAC repairpeople? That would be equally fun to watch!