From the course: AI Applications in Healthcare: A Conversation with Dr. Matthew Lungren
What is the future of AI in healthcare?
From the course: AI Applications in Healthcare: A Conversation with Dr. Matthew Lungren
What is the future of AI in healthcare?
- I would love to discuss the future, and you already say that you feel like you're living in a science fiction novel, but let's go even deeper and look at one year, five years, whatever you like from now. What emerging AI technologies are you most excited about in the context of healthcare, and then how far away are we from seeing widespread implementation of these technologies? - Yeah, I mean, so this is where I like to spend my time personally just 'cause someone who loves to do the new thing or try to push the boundaries as much as possible, and so for me, one of the most obvious things is when I look at, particularly in healthcare, if I look at the amount of data that I use every day to take care of my patients, it's not just text. In general, the vast majority of data for most health health systems is not all in text. And as you've seen with the commercial version of GPT-4, you can take a picture of, I don't know, your broken bike and say, "Hey, my bike's not working. Maybe my seat doesn't raise or lower, whatever it is. Can you help me fix it?" And you can have a conversation with an image and you can show, the model's like, "Well, yeah, I don't, can you take a picture of this area a little more closely?" And you do that and it's like, "oh, okay, you need this kind of an Allen wrench," and you go through this interactive experience with an image and text combo, and to me, there's nothing more exciting about that same kind of capability in healthcare, right? So take this X-ray, can you help me find, is this pneumonia? Okay, well, what medications would you want to treat this with based on what you're seeing in the pattern, like having that conversation, or here's a brain MRI, do you think that there's any challenges with a stroke in the right frontal, whatever your questions might be, that dynamic interactivity with the healthcare data outside of just text, to me, is my north star. I just feel like this can unlock so much more that we can do both, again, within our current sort of workflows, but even brand new workflows. And then take that a step further and now we're pushing this even further out in the future, but now imagine with that same information, you as a patient, you have all your healthcare data or you have the data of your aged parent or your child and you're trying to understand, what should I be doing? Or what does this really show? Or what medication should I worry about? Just having simple conversations with your own data, what a powerful idea that is. And I think that that's where this generative AI technology, again, kind of opens up some doors that, again, we would never have thought possible before. So where I see the future going is certainly, number one, building competencies for these models in medical data that's not just text, and I think that we're on a really great trajectory for that as a field. Next step to that is, well, how do I make interactive experiences that can really, really operationalize some of the powerful capabilities of these models for everyone, whether you're a patient, whether you're a physician, nurse, whatever, right? And that to me would be a really powerful North Star, because again, you start to see that information asymmetry kind of unlocked at scale. - What a wonderful point about sort of empowering the patient to learn more about their condition, a little bit better than going down the WebMD rabbit hole. - Right, right. Those never end well, do they? - No, they never do. It's a very scary place to spend the middle of the night, that's for sure. - That's right.