Injecting AI into Medicine: The NEJM AI Launch - a conversation

Host 1: Welcome to our deep dive today.

Host 2: Oh, this should be good.

Host 1: We're diving into AI in medicine.

Host 2: Okay.

Host 1: And our source material comes from a new publication called, uh, NEJM AI.

Host 2: Mm, interesting.

Host 1: And, uh, it's put out by the same folks behind the New England Journal of Medicine.

Host 2: Oh, wow. So they're the authority.

Host 1: They're focusing on navigating the rapidly developing world of AI in healthcare.

Host 2: Yeah, it's changing so fast, right?

Host 1: It is, and I think it's safe to say we're all kind of wondering what this could mean for the future of medicine, you know?

Host 2: For sure.

Host 1: Imagine AI helping doctors make better decisions and personalized care.

Host 2: It's mind-blowing to think about where this is all headed.

Host 1: It really is, so we're going to find out more. What I find really interesting is how AI in medicine isn't like some brand new idea.

Host 2: Right.

Host 1: It's been on the minds of medical experts for decades.

Host 2: It has. The editorial we're looking at actually talks about that history.

Host 1: Oh, that's cool.

Host 2: And, um, the idea of using machines to enhance medicine goes back to at least the 1950s, believe it or not.

Host 1: Wow, that's amazing. That's a long time.

Host 2: It is.

Host 1: So was there anything like concrete back then or was it just ideas?

Host 2: There were definitely ideas and even some early attempts,

Host 1: Okay.

Host 2: but the technology just wasn't there yet.

Host 1: Right.

Host 2: But, you know, there was even a 1970 article in the New England Journal of Medicine

Host 1: Yeah, wow.

Host 2: where a doctor named, uh, William Schwartz predicted the major influence computers would have on healthcare.

Host 1: Mhm.

Host 2: And he saw both the good and the bad that could come from it, which I think is really interesting.

Host 1: Yeah, because, you know, with any new technology, there's always like that double-edged sword.

Host 2: For sure.

Host 1: It's amazing how accurately he predicted the impact all those years ago.

Host 2: It really is.

Host 1: So what's really remarkable is how these early visions of AI are now becoming a reality.

Host 2: Absolutely.

Host 1: We're already seeing AI perform tasks that like rival human ability, especially when it comes to image recognition, right?

Host 2: Yeah, definitely.

Host 1: It's kind of crazy to think about, like computers analyzing medical images with the same skill as an experienced radiologist. Is that what they mean by like AI quote-unquote rivaling

Host 2: Mhm.

Host 1: human ability?

Host 2: Exactly, and that's just the beginning. One of the biggest advancements in AI is the development of large language models,

Host 1: Okay.

Host 2: or LLMs.

Host 1: LLMs.

Host 2: Yeah, LLMs.

Host 1: And what is that?

Host 2: Think of it like this, um,

Host 1: Right.

Host 2: imagine a chatbot

Host 1: Uh-huh.

Host 2: that has absorbed every single medical textbook and research paper ever written.

Host 1: Oh, wow.

Host 2: That's essentially what an LLM is.

Host 1: Okay, now that's what I call a super-powered medical librarian.

Host 2: Right.

Host 1: So it's no wonder people are both excited and a little bit nervous about this technology, right?

Host 2: Definitely.

Host 1: Because if it has all of that information, it could do like amazing things, but it also could maybe

Host 2: Make mistakes.

Host 1: Yeah, or be used in the wrong way.

Host 2: Exactly.

Host 1: So how do we make sure that this incredible technology is used safely and effectively?

Host 2: Well, the editorial really stresses that AI tools need to be put through the same rigorous testing as any new drug or treatment.

Host 1: Right.

Host 2: It's all about making sure these tools are both safe and effective for patients.

Host 1: That makes total sense, but isn't that like a pretty high bar to set,

Host 2: Yeah.

Host 1: especially with how quickly AI is evolving?

Host 2: It is a high bar, and the truth is the editorial points out that less than 1% of AI evaluations in 2020 were the kind of rigorous randomized controlled trials we need.

Host 1: Wow, only 1%?

Host 2: Less than 1%, yeah.

Host 1: That's surprising.

Host 2: It is. And, you know, these are the gold standard for proving whether a medical treatment really works.

Host 1: Right.

Host 2: In these trials, patients are randomly assigned to receive either the AI intervention or a standard approach. And this helps researchers isolate the impact of the AI and determine if it's truly beneficial.

Host 1: So it really takes out all the guesswork.

Host 2: Exactly.

Host 1: So there's a lot of catching up to do when it comes to evaluating AI in medicine.

Host 2: Absolutely. And the editorial highlights how this becomes even more complex with LLMs,

Host 1: Okay.

Host 2: because they are so versatile,

Host 1: Mhm.

Host 2: it's hard to know what data they've already been trained on,

Host 1: Oh, I see.

Host 2: which makes evaluating them that much harder.

Host 1: It's like trying to hit a moving target. So how do we address this? Where do we even go from here?

Host 2: Mhm. Well, the editorial emphasizes two key factors: collaboration and transparency.

Host 1: Mmm, okay.

Host 2: They're calling for a true team effort, bringing together clinicians, patients, and AI experts to ensure these tools are developed responsibly.

Host 1: I like that, yeah, because it's not just about the technology itself. It's about how it's going to be used in the real world.

Host 2: Exactly.

Host 1: So I mean transparency is crucial for building trust, right?

Host 2: Absolutely.

Host 1: Patients and doctors need to understand how AI is being used, what data it's based on, and what the potential risks might be.

Host 2: You're absolutely right, and it's not just about evaluating the technology, it's about shaping the future of AI in healthcare.

Host 1: Yeah, like proactively thinking about where we want this to go.

Host 2: Right, exactly.

Host 1: So, what is NEJM AI doing?

Host 2: Well, they want to be a platform for discussion and debate about all of this.

Host 1: Oh, okay. So they're not just publishing research, they're trying to facilitate like an open conversation.

Host 2: Exactly. They want to bring everyone to the table to talk about these important issues.

Host 1: I can see how that's so important, especially in a field that's changing so rapidly.

Host 2: It is.

Host 1: We need to be having these conversations now.

Host 2: We do.

Host 1: So what kinds of things are they discussing?

Host 2: Well, they're addressing the ethical concerns, trying to figure out who should be guiding the development of AI in medicine,

Host 1: Okay.

Host 2: and, you know, how to make sure it truly benefits both patients and doctors.

Host 1: It sounds like a lot to tackle.

Host 2: It is.

Host 1: But it's really important.

Host 2: It is.

Host 1: So the editorial also raises a really critical point about the data used to train these AI models, right?

Host 2: Yeah, the data is everything

Host 1: Yeah.

Host 2: because it shapes how the AI learns and makes decisions.

Host 1: So tell me about that.

Host 2: Well, they emphasize the need for diverse and accessible data sets.

Host 1: Okay.

Host 2: That means data from a wide range of patients and populations.

Host 1: Because if the data's biased or incomplete, then the algorithms themselves will be biased and that could lead to unfair differences in how people receive care.

Host 2: Precisely. That's why it's so important for the data to reflect the diverse communities we serve.

Host 1: Right, so it's not just a technical challenge, it's really a matter of social responsibility.

Host 2: Exactly.

Host 1: So we've got this incredible technology with the potential to like revolutionize medicine,

Host 2: Uh-huh.

Host 1: but we need to be very careful and thoughtful about how we develop and use it.

Host 2: Absolutely.

Host 1: And the New England Journal of Medicine launching this new journal dedicated to AI seems to reflect how seriously they're taking this, right?

Host 2: Well, I think so.

Host 1: They're really trying to like guide the conversation in a responsible direction.

Host 2: I agree.

Host 1: And with that in mind, they pose a very thought-provoking question in this editorial.

Host 2: Oh, what's that?

Host 1: They ask if AI will make the practice of medicine more rewarding.

Host 2: Hmm. Interesting.

Host 1: That's a great question to think about.

Host 2: Yeah, it really is.

Host 1: It makes you wonder, could AI actually free up doctors to spend more time connecting with patients, you know?

Host 2: Yeah.

Host 1: Less paperwork, more FaceTime, that sounds like a win-win to me.

Host 2: It does, but we need to be careful about assuming it'll be that simple.

Host 1: Oh, yeah.

Host 2: You know, the editorial brings up the possibility of AI, um, encroaching on some very human aspects of medicine.

Host 1: Like what?

Host 2: Well, for example, like what happens when AI starts making diagnoses

Host 1: Hmm.

Host 2: or recommending treatments that clash with a doctor's own judgment?

Host 1: That's a really good point. It raises all sorts of questions about how much autonomy doctors should have when AI is like providing these insights.

Host 2: For sure.

Host 2: And like how do we make sure the patient's voice is still heard in the decision-making process?

Host 1: Exactly, it's a delicate balance.

Host 2: It really is.

Host 1: And then there are the emotional aspects of healthcare to consider, right?

Host 2: Oh, yeah.

Host 1: Like can AI truly replicate the empathy

Host 2: Right.

Host 1: and compassion that patients often seek from their doctors?

Host 2: You know, that's such a core part of, you know, feeling cared for.

Host 1: Exactly, those are questions we absolutely need to be thinking about as AI becomes more integrated into healthcare.

Host 2: Absolutely.

Host 1: It's not just about the technology, it's about how we use it to enhance the human connection.

Host 2: Right.

Host 1: That's at the very core of medicine.

Host 2: So are there any potential downsides to like relying too heavily on AI in medicine?

Host 1: Well, you know, the editorial reminds us that data plays a critical role in shaping the future of AI in healthcare.

Host 2: And if the data used to train these AI models is biased or incomplete, the algorithms themselves could perpetuate those biases.

Host 1: Right, so if the AI is learning from flawed data, it could actually end up reinforcing existing inequalities in healthcare.

Host 2: That's the concern, and it highlights why the editorial is calling for a real commitment to building diverse and accessible datasets.

Host 1: Right, like we were just talking about.

Host 2: Data that truly reflects the wide range of patients and communities we serve.

Host 1: It sounds like the success of AI in medicine depends not just on the technology itself, but on how we develop and use it responsibly.

Host 2: It certainly does.

Host 1: And a crucial part of that responsibility is transparency, right?

Host 2: Absolutely.

Host 1: We need to know what data's being used to train these models

Host 2: Yeah.

Host 1: and how those models are being applied in real-world medical settings.

Host 2: Exactly.

Host 1: So circling back to that question posed by the NEJM AI editorial,

Host 2: Yeah.

Host 1: will AI make the practice of medicine more rewarding? What are your thoughts?

Host 2: It's a complex question.

Host 1: Yeah, there's a lot to unpack there.

Host 2: There is, and I think the answer ultimately depends on how we choose to integrate AI into healthcare.

Host 1: Mhm.

Host 2: If we use it thoughtfully and ethically,

Host 1: Right.

Host 2: with a focus on enhancing the doctor-patient relationship,

Host 1: Mhm.

Host 2: I believe AI has the potential to make medicine more rewarding for everyone involved.

Host 1: But if we let AI become a substitute for that human connection,

Host 2: then we risk losing something truly essential about what it means to care for others.

Host 1: Wow, that's a powerful statement.

Host 2: It is.

Host 1: We can't forget that AI is a tool,

Host 2: Right.

Host 1: a powerful tool,

Host 2: It is.

Host 1: but still just a tool.

Host 2: Right, exactly.

Host 1: It's up to us to decide how we use it to shape the future of healthcare.

Host 2: I agree.

Host 1: That's a great point to remember. We've covered a lot of ground in this deep dive, from the history of AI in healthcare to the ethical challenges it presents, but we're not done yet. In the final part, we'll explore NEJM AI's vision for the future of this exciting field and how you can get involved. Stay with us.

Host 2: I'm excited to hear more.

Host 1: So let's get back to that question the NEJM AI editorial asked.

Host 2: Yeah, about AI and

Host 1: Yeah, will AI make practicing medicine more rewarding?

Host 2: It's a good one.

Host 1: It is. It makes you think about like what it really means to be a doctor.

Host 2: Mhm, totally.

Host 1: We often talk about AI automating tasks, but this question makes us think about the human side of medicine.

Host 2: Right, the impact on doctors themselves.

Host 1: Exactly. So could AI actually free up doctors to spend more time with their patients?

Host 2: Potentially.

Host 1: You know, less paperwork, more FaceTime.

Host 2: It's an interesting idea.

Host 1: That sounds pretty good to me.

Host 2: It does, but we have to be realistic. It might not be that straightforward.

Host 1: Oh, okay.

Host 2: The editorial brings up a good point, actually.

Host 1: What's that?

Host 2: It talks about the possibility of AI, like encroaching on some of the very human aspects of medicine.

Host 1: So like what does that look like?

Host 2: Well, think about it. What happens when AI starts making diagnoses

Host 1: Okay.

Host 2: or recommending treatments that clash with a doctor's own judgment?

Host 1: Mmm, that's a good point. It raises a lot of questions about how much autonomy doctors should have

Host 2: Right.

Host 1: when AI is giving them all these insights.

Host 2: Exactly, and how do we make sure the patient's voice is still heard

Host 1: Yeah, it's a tricky balance.

Host 2: in the decision-making process? It is, and then there are the emotional aspects of healthcare to think about, too, right?

Host 1: Like can AI really understand what a patient's going through?

Host 2: Right. Can it provide that empathy and compassion that people need from their doctors?

Host 1: Yeah, because sometimes just knowing that someone really cares makes a huge difference.

Host 2: Absolutely. So those are questions we really need to consider as AI becomes more integrated into healthcare.

Host 1: It's not just about the technology.

Host 2: Oh.

Host 1: It's about how it impacts the human connection.

Host 2: That's at the heart of medicine.

Host 1: So are there any other downsides to relying too much on AI in medicine?

Host 2: Well, the data itself is a big one.

Host 1: Okay, how so?

Host 2: The editorial reminds us that the data used to train AI models is crucial,

Host 1: Right.

Host 2: because if that data is biased

Host 1: Yeah.

Host 2: or incomplete,

Host 1: Mhm.

Host 2: then the algorithms will be biased, too.

Host 1: Right, they'll just perpetuate those problems.

Host 2: Exactly, which means AI could actually end up making existing inequalities in healthcare even worse.

Host 1: That's a big concern.

Host 2: It is, and that's why they're emphasizing the need for diverse and accessible data sets,

Host 1: Yeah.

Host 2: data that represents everyone.

Host 1: So we need to make sure these AI systems are learning from good data,

Host 2: Absolutely.

Host 1: data that reflects the real world.

Host 2: That's the key.

Host 1: It seems like the success of AI in medicine hinges on more than just the technology.

Host 2: It does.

Host 1: It's about being responsible

Host 2: and ethical.

Host 1: about how we use it.

Host 2: Exactly.

Host 1: And a big part of that is transparency, wouldn't you say?

Host 2: I would. We have to know how these models are being trained

Host 1: Right.

Host 2: and how they're being used in real-world settings.

Host 1: Transparency builds trust

Host 2: Definitely.

Host 1: and accountability.

Host 2: So let's go back to that question from the NEJM AI editorial.

Host 1: Okay.

Host 2: Will AI make the practice of medicine more rewarding?

Host 1: Big question.

Host 2: It is, and I don't think there's an easy answer.

Host 1: No, probably not.

Host 2: It depends on us,

Host 1: on how we use AI.

Host 2: Yeah. If we use it thoughtfully and ethically,

Host 1: Okay.

Host 2: with the focus on improving that connection between doctors and patients,

Host 1: Mhm.

Host 2: then I think AI has the potential to make medicine more rewarding for everyone involved.

Host 1: But what if we let AI replace that human connection?

Host 2: That's the danger. We could lose something essential about what it means to care for others.

Host 1: It's a good point to keep in mind.

Host 2: AI is a tool.

Host 1: Right, a very powerful tool.

Host 2: But it's still just a tool.

Host 1: And how we use it is up to us.

Host 2: We shape the future.

Host 1: That's a powerful thought. All right, so in this final part of our deep dive,

Host 2: Okay.

Host 1: we're going to shift gears a bit

Host 2: Okay.

Host 1: and take a closer look at NEJM AI itself.

Host 2: Sounds good.

Host 1: It's more than just a new medical journal, isn't it?

Host 2: It is, you're right.

Host 1: Like they're trying to be a real force in this whole AI in healthcare thing.

Host 2: Exactly. NEJM AI wants to actually shape the future of AI in medicine.

Host 1: Okay, that's a pretty big goal. How do they plan on doing that?

Host 2: Well, they lay out three main goals in this editorial.

Host 1: Okay.

Host 2: First, they want to see more rigorous evaluation of AI tools.

Host 1: Like we talked about before.

Host 2: Yeah, they're really pushing for high quality clinical trials.

Host 1: Which makes sense because there's been a lot of hype around AI,

Host 2: Right.

Host 1: but not enough evidence.

Host 2: Exactly, and that's where they want to step in.

Host 1: Okay, so what's the second goal?

Host 2: It's all about transparency and collaboration.

Host 1: Makes sense.

Host 2: They want to bring everyone together.

Host 1: Like who?

Host 2: Clinicians, patients, AI experts, policymakers,

Host 1: Wow, everybody.

Host 2: to openly discuss the challenges and opportunities of AI in medicine.

Host 1: It sounds like they're trying to create a space for real dialogue.

Host 2: That's exactly it.

Host 1: Which is so important in a field that's changing so fast.

Host 2: It is. It's crucial to have those open conversations.

Host 1: All right, so what about the third goal? What's that focused on?

Host 2: That one's all about the data.

Host 1: The data, okay.

Host 2: NEJM AI wants to advocate for diverse and accessible data sets.

Host 1: Right, because that's what the AI models are trained on.

Host 2: Exactly.

Host 1: And if the data is biased, the algorithms will be biased, too.

Host 2: Which is a huge problem.

Host 1: It is. It could really exacerbate health disparities.

Host 2: And they want to avoid that, so they're pushing for data that represents the full spectrum of human diversity.

Host 1: So it's not just about the technology. It's about ethics and social responsibility.

Host 2: You got it.

Host 1: It sounds like NEJM AI is really tackling a lot.

Host 2: They are, and they're not limiting themselves to just the US.

Host 1: Oh, interesting.

Host 2: They're planning international meetings

Host 1: Wow.

Host 2: and collaborations with journals from all over the world.

Host 1: To get a global perspective.

Host 2: Exactly. It's encouraging to see that kind of commitment.

Host 1: It really seems like they're going to be a major voice in this field.

Host 2: I think so, too.

Host 1: They're taking it seriously.

Host 2: They are, and you know what?

Host 1: What?

Host 2: They want everyone to be a part of this conversation.

Host 1: Really? So it's not just for experts?

Host 2: Not at all. Whether you're a doctor, a patient, or just someone who's curious about the future of medicine,

Host 1: Anyone can contribute.

Host 2: Exactly. Your voice matters.

Host 1: I like that.

Host 2: Because we all have a stake in this.

Host 1: That's a good point. So as we wrap up this deep dive,

Host 2: Okay.

Host 1: let's go back to that question posed by the NEJM AI editorial:

Host 1: will AI make the practice of medicine more rewarding?

Host 2: That's the big one.

Host 1: It is, and we've explored a lot of different angles.

Host 2: We have. The benefits, the challenges, the importance of data and transparency.

Host 1: Now we want to hear from you.

Host 2: What do you think?

Host 1: How do you see AI impacting your own healthcare in the future?

Host 2: Head over to NEJM AI and keep exploring these questions.

Host 1: It's an exciting time to be following this field, and your voice really does matter. So thank you for joining us on this deep dive into the world of AI and medicine.

Host 2: It's been a fascinating conversation.

Host 1: And we'll be back soon with another one.

Host 2: Looking forward to it.

Host 1: So stay tuned.