Will AI Replace Doctors?

That looks pretty good. What is that? It's the good stuff, man. >> Is that your uh your nicarette gum? >> No. Little nicotine. >> I wish a cardiograph is trying to quit smoking. >> I know, right? Do as I say, not as I do. >> No, no, that's my Hey, smoking kills, you know. So, so no, that's my uh nicotine gum. No, I'm kidding. It's uh it's my caffeine pills that I'm taking this. It's too late, man. I'm tired.

I've been tired of uh just exhausted and these things, this narrow gum thing I found, uh just a little bit of mint helps me just stay awake and like get that jolt of energy. Uh which >> Yeah, man. It helps when we we're recording these late at night and it's a long week already. >> Yeah. Yeah. I mean, long week of uh writing notes and uh seeing patients, talking to like 20 25 patients a day, writing all their notes, you know.

Um I use a lot of uh AI softwares and doing my notes uh and uh even then I'm still so behind. I use so much technology. >> I hear you. The notew writing is the worst. >> What AI stuff are you using? I haven't actually dabbled into AI. I'm just hand typing all my notes and it's killing me. >> Yeah, I mean you know it I've as you know I'm a pretty early adopter in uh a lot of things.

So I started uh using a lot of the AI um speech to text softwares and it's not really speech to text really understands what you're talking about. So I started trying them out almost a year and a half ago or a year you know a year and a few months ago. >> You've been using them for a while. >> Yeah I've been using them for a while.

Uh I've tried maybe four or five different ones and uh you know finally found one that suits my practice and suits my style and I started using I've been using that one consistently for six to eight months and uh it's interesting. It took me a while to find the right one. Uh and I tried different ones. So I used one software I I like it for a while then someone suggest oh maybe this one's better. I see a I get an Instagram ad about a different one and I try it and I always come back to this one.

So >> So the Instagram ads work, huh? >> Huh? Yeah. It's it's awesome, man. Uh I think it's changed the way I do my clinic. I mean, you know, a big complaint of a lot of patients is that uh they go to a physician's office, they wait in the waiting room for 45 minutes, 30 minutes, 45 minutes, they get room, then they wait in the room for another 30, 35 minutes, 40 minutes sometimes. uh for some doctors, not for me.

But when the doctor comes in, they're on the computer just typing away and not even looking at the patient. There's literally patients that told me uh the doctor did not look at me at my eyes or at my face a single time. He was just staring at his computer. And uh I think that changed a lot with me. You know, there's a pressure on us. I'm doing our notes and documenting everything.

Um but with the AI softwares that I've been using it it has been pretty amazing in the sense that I can just get there turn it on. Uh it's pretty hippos secured where all the data is secured. There's a dual factor authentication uh you know uh and the data gets deleted once I'm done with it. Uh but I sit there I turn it on and I just talk to the patient look look at them talk talk about other things.

talk about sometimes the favorite sports teams, what they've done, what they do, what their job is like, and also talk about cardiology, but uh it it kind of, you know, uh know listens to it, listens to the conversation and then creates a whole soap note for me at the end and remembers all the details and forgets all the fluff.

So, so which is pretty amazing and and the best part is you can talk in any language and it records it and and changes it into English and and makes a a soap note out of it. >> Man, that's awesome.

I mean, you you know, it's it's very much creates an impersonal atmosphere in the clinic like for myself sometimes when I'm trying to juggle talking with the patient sitting in front of me doing the physical exam which that part in orthopedics it forces you to have a little bit of a personal touch because you're you're actually touching the patient. you got to look at them during the exam, but then you kind of go back to the computer and you're typing notes.

You got to quickly type out cuz I don't want to take a bunch of notes home with me. I don't want to see 40 45 patients in clinic and then have to do 40 notes at home and now you're working two extra hours and taking time away from your family. So, I mean, that sounds awesome, the ability to use that if it really makes you more efficient. I mean, that's huge. >> Yeah.

And even, you know, I liked my notes to have a little bit more detail and sometimes if you are late on your notes and you're trying to focus on the patient, you don't remember what exactly happened. And this really helps me go back. Hey, this is what I did and this is what I was thinking and this is the conversation I had and now I can write my assessment plan based on the conversation I had because it's summarized for me.

So I think you know um there's a lot of talk about AI in medicine, AI in technology, AI in business and all of us are scared, curious, excited about what AI is going to do for us. Uh what's AI going to do to us? A lot of doctors have this this argument or conversation almost every day about is AI going to replace us? What do you think, man? Is AI going to replace us? >> Yeah, it's a hot button question, right?

And it's kind of like a an incomplete thing that it's very headline catching like will we be replaced by AI? I mean, I think replacing a doctor and the doctor patient relationship is a far stretch. I mean will AI replace or more so I I think augment some of the skills that we already have or that we could do better that that I think is much more realistic and probable and probably better in a lot of ways.

I mean as as an example right in visual diagnosis in imagebased diagnosis in pattern recognition like a lot of diagnostic radiology pathology assessment of slides even dermatology which is visual diagnosis of the skin those things are very well done already. I mean there's real data, real studies showing exceptional results and even superior performance to humans that these artificial intelligence algorithms can do.

Um but other things I mean making a connection with the patient talking with the patient when a complication occurs when a side effect of a medication occurs that counseling aspect a patient brings up something random that's really troubling them like the death of a loved one that has interfered with their initial plan to do a surgery. I mean things like that happen all the time. This is human life things occur.

I think those things and the ability to pivot and have that emotional connection that I mean it would be hardressed to be replaced by an AI algorithm but augmenting our current skills and predictive ability. I mean those things for sure I think would be beneficial. >> Yeah. You know uh we keep hearing this u good old saying medicine is an art. You know medicine is science but medicine is also an art.

The science of it is uh the knowledge that we learn years and years and years educating ourselves on you know we read all these books and go through Robins and uh our anatomy books and do our training in anatomy and physiology and biochemistry and all these things that we learn that's the knowledge that we gain over the 8 n 10 12 years that we train but the art of medicine is also something you learn uh not everyone is uh good at the art of medicine uh not everyone is good at looking uh at patient in the eye uh and telling them when something bad is about to happen or when something bad happened and you're trying to tell them the course forward.

Um, so yeah, a lot of us have knowledge. There's doctors that there's a lot of good doctors that have much more knowledge than I have and much more knowledge than you have. They know every single study, every single details about uh the medical aspect of things. But what makes a good doctor is someone who can sit with the patient, show compassion, uh look them in the eye, and tell them what's going to happen and how they'll be fine. They'll be okay.

Uh and that's that's an art and I don't know how how far uh AI will go that it'll be able to do that for patients. >> Yeah, it's it's a big ask, right? I mean that's really what being a doctor in a large aspect is is converting that knowledge into real human connection and interacting with people in that way.

I mean I think you know as as a concrete example things that AI can and already in some ways does very well uh like in point of care acute injuries for example a patient comes in after a fall they think they broke the wrist they go to an ER and they get an X-ray your average ER physician nothing against ER physicians does not look at as many X-rays of the wrist as an orthopedic surgeon or radiologist it's simply a matter of volume they don't and so their ability to diagnose whether there's a fracture and where the fracture is the pattern of the fracture, all that stuff is simply not as good.

And that's okay. They do a lot of other things. But now, if you had an AI algorithm that could be an extra tool to help you in that setting, whether you're a physician in the ER or a lot of ERS and urgent cares are staffed by ancillary providers, NPs, PAs, it could be a great tool. And that that was a study that was done on something like 70,000 X-rays. This huge repository of wrist X-rays demonstrating if there's a fracture or not.

and if that could be a tool to help the physician in the point of care urgent care setting and it marketkedly improved their ability to not only diagnose whether there was a fracture or not but also if this was likely to be an operative injury meaning higher chance and higher need for them to see a hand or wrist surgeon soon and so there's already there's a lot of stuff that's just one example there's a lot of stuff that simple AI tools like this can be an augment and a benefit to our practice but also our patients >> yeah I mean look at the uh speaking of ER physicians I ERS are overwhelmed and overloaded.

I mean, there's lines out the door. >> Totally. >> Sometimes patients that are much more complicated are getting missed waiting in the hallway. And sometimes simple simple little injuries that could be taken care of by someone not as highly trained as an ER doctor uh that could be managed outside of the ER are being managed in the ER and taking space and time.

I mean, you know, AI can play a huge role in triaging early on, offload that ER, assessing things based on initial X-rays and labs and findings that, hey, this patient can be seen by someone as outpatient right away versus someone that needs to be seen by a physician right away. I mean, you know, those kind of algorithms that's kind of, you know, and that's kind of what a lot of the ER is, right? A lot of it is algorithm based as well.

uh a lot of the early triaging medicine triage is a lot of algorithm based and when you're doing that through AI that can be a huge way to offload not just ERS but primary care physicians right I mean a lot of primary care physicians get calls get appointments get overloaded by a lot of simple questions that could be answered uh by someone that you know it doesn't require a patient to come in and an AI app that's connected to their physicians with the physician's knowledge and supervision can be triaging those patient answering some simple questions about medications so the patients does not have to come to the primary care physician so the primary care physician can focus on those really complicated patients and I think that's one way that AI will augment uh you know our healthc care shortage of doctors that there is right now or healthcare providers that there is right now >> yeah you know it brings up an interesting point right like if those I mean not even if some of them do exist already but these AI tools tools if they become more ubiquitous at what point does it become simply a tool that may or may not offload a physician allow a patient to get a little more direct uh and immediate information but if they're still supervised in some way like a physician is still in some way responsible this is something I think about a lot like where does the liability fall is Microsoft Google open AI are these companies ever going to actually take responsibility and liability if an error occurs if a patient injury, a death or something occurs because of information that was provided by this machine learning algorithm that simply used a giant data set in a probabilistic model.

Is that really going to be something that is medical legally liable? Or is it always going to be the doctor who still remains liable and god forbid something occurs is the one that gets sued even though they were basically relying on these algorithms and now they're supervising potentially tons of these algorithms and thousands of patient queries that they didn't have to do before. That's an added element of their job in a way.

And it's impossible really to supervise numerous inflows of questions and questions and questions from patients or patients just checking in. Hey, I saw this on Chad GBT. Hey, I saw this on Gemini. Is this true? Is this legit? Can I do this? I mean, it opens up a whole new Pandora's box of things that potentially could now come down to the physician. Would I want to would you want to be liable potentially for something that is said via a large language model? >> I don't know.

I mean you know we live in a latigenous uh society you know uh doctors take on a lot of uh responsibility uh for patients and answering questions. I mean it's it's easy to you know you hear a lot of people say oh yeah my doctor missed it or my doctor uh did not diagnose this my doctor um uh you know told me this but I looked it up and this was the real things. So yeah, you're right.

I mean at this point at that point who takes the who takes the blame if if a doctor is being if a hospital system is imposing an AI AI system algorithm based ER triage system uh and a patient comes in and u based on t tons of algorithms something gets missed that was supposed to be seen by a physician and being taken care of right away rather than they get sent home. Who is responsible? is the hospital responsible. I don't think uh a corporation will be willing to take on the responsibility.

I mean you know doctors are always end up being the ones that have to be medically liable for these kind of decisions and um corporations can always say oh uh we're not physicians. This was an AI algorithm that was implemented and it was under the supervision of the physician.

So at what point you're absolutely right at what point it becomes too much for a doctor to monitor all and at what point we can let AI work autonomously right um that's why I think uh medicine will be the one of the slowest to change because who will take that responsibility when AI messes up. >> Yeah. I mean really nobody uh other than the companies that made the AI should right.

um or it should be a situation where maybe this is too idealistic but where it shouldn't be a thing in the medical legal environment right if you as a citizen happily choose to use this software let's say someone gets advice about chest pain via chat GPT and if chat GPT is somehow integrated down the road or a tool like it into the EHR with some method of access like a my chart type interface to their physician's office if they get advice from there that plays it down unlike webm MD, which says basically everything is an emergency.

Go to the ER. If it plays it down and the patient does end up having a cardiac event, god forbid dies, something happens, right? What happens then? I mean, the patient did choose to use that rather than actually scheduling a visit, going to the doctor or going to an ER. So, at some point, it's a question of personal responsibility, which I think is a huge factor and something that isn't really discussed a lot in this whole space of AI.

the buzz and the hype is really about replacement of potential doctors or providers or tools, but where does the responsibility come in? I mean, people are still choosing to use this. There's not a mandate to use this stuff. And so, if you choose to use an added tool or a service or software, some of the onus should be on you. >> Yeah, I think it'll always be kind of a a combined partnership.

I was reading this book about uh AI and medicine and in that the author gives a pretty good example about one of the there was a professor in one of the small college towns and he goes to the ER uh he's having a lot of abdominal pain uh and the abdominal pain you know the ER physician checks him out hey it seems like it's it's a gastritis let's take some of these medications hydrate yourself rest follow with your primary care doctor and the professor was not very convinced and uh uh put in some of his symptoms into uh AI software I think Chad GBT and one of the differential diagnosis that was given to him was uh appendicitis uh which makes sense and he he asked the ER doctor at that point hey uh can you can you please do a CT scan because based on my symptoms I think it's appendicitis from what the AI software is telling me so uh the ER doctor ended up doing a CT scan it ended up being appendicitis and he was taken for surgery.

So at this point, I mean, you know, it's it's physicians do miss things without AI, physicians may miss things with AI. We'll just have to find a balance with patient partnership of how to use these tools to the best of our ability.

Sometimes patients will have to understand and they'll have to take some ownership of of what they're asking and what they're uh you know requesting because sometimes you might end up doing a lot more CTs than you don't you might find some kind of incidental nodule but end up might getting biopsied and um because a patient requested a CT scan and then who is to blame right because the doctor listened to the patient's request because Chad GPT told them it could be appendicitis and then they found this possible non-cancerous mass that did not need all this workup.

So, it it'll be a interesting surge. I don't think I have answers to that yet. Um, >> yeah, I mean I I don't think anyone does, but I mean they're they're interesting and they're important questions, you know, like even if you delve a little bit into the weeds of AI and how a lot of these algorithms are actually created and how these softwares work. Um, a big question is like they're only as good as the data you put in.

in in a a little more simplistic sense, it's kind of like any PubMed or literature study that's based off a database. The study and the outcomes and the results of the study and their ability to apply to mine and your practice are only as good as the initial data set that was put into it. It's the exact same thing with AI. AI is nothing but analyzing these massive data sets and coming up with a method to analyze them um with high probability of success.

And so if your initial data set, even if it's huge, if it's for some reason not generalizable to the population at large, then you can really only use AI with high fidelity if you ask it about questions relating to that initial data set. So if you don't have real generalizable data, you're going to get inadequate information.

And one thing that AI does, and this is one of the known problems with AI, I've done a lot of AI research, and one of the issues you come into is called overfitting, where the AI algorithm itself, its goal is to give you a response. Its goal is not to give you the right response. Its primary goal is to give you a response. And so, it will overfit. It will fit itself to the peculiarities within your set of data.

And it will only give a response based off that, but it won't tell you that cuz it can't know that. It doesn't know what exists outside of that data. And so it may not apply to the population at large. It may only apply to that very select little cohort. But the person asking it, a patient or myself in this case, and the AI algorithm itself are unaware of this. >> Yeah.

But you know, if you think about it, Adel, u aren't we all kind of same where there's certain sets of datas that have been input into us and we've learned through our experiences of seeing patients and uh the books we've read and the articles we read, the studies we read and the research we've done. uh we're taking those experiences and that data to help patients, right? So, uh how's AI any different?

Maybe just a version where uh a version of software that can process a lot more information than we can, right? Uh >> sure. uh may maybe the AI yet does not have the real life examples of seeing patients and seeing their body language and seeing their physical uh symptoms and kind of having a hunch about their symptoms and kind of you know using that hunch to maybe order some tests and diagnose them yet. uh but at some point it will uh because >> for sure and that that that's a very good point.

And I mean these AI things, these algorithms, these softwares, they are self-correcting, right? They learn as they go. And that's a huge benefit in a way. It's kind of like how humans learn. We learn as we go through experience.

Um, but if if we create a tool, for example, to assess someone's risk of a cardiac event or assess someone's risk of a fragility fracture and we feed it whatever demographic data, imaging, blah blah blah from the outset, initially that tool is only going to have that data. It's not going to have other population data.

It's only going to be limited to whatever if we decided to come up with it, whatever we could think of at the time, it would be limited by its human creators who are naturally flawed. >> Yeah. And and and we are limited by our own biases, too. I mean, you know, it's it's it's a known fact that uh women of color tend to get treated differently with chest pain compared to a white male, right? People have their own biases. And maybe in the United States, that's the way.

Maybe in in in Africa that's not the case, right? Uh so wherever that data is going into the positions and the biases that they go through throughout their training, those take a take a role in when they make that decision in medical care. And I think AI will be the same way.

it'll have its own biases and who knows I mean you know a lot of times uh I use a lot of chat GPT Gemini and a lot of these things and some of the previous questions I've asked it it keeps that in mind when giving me answers right so if I've looked up a lot of information about cancer in the past will that uh will it tailor my tailor the response based on my previous searches and the information I've been concerned about uh you know uh that will be interesting thing as well because if I'm a person that looks up am I gonna have a heart attack am I gonna have a heart attack am I gonna have a heart what's my risk of a heart attack and then when I look up symptoms of chest pain it indicates towards me having a heart attack versus something else maybe something like costtochondritis you know so how much of your own biases will it >> yeah no that's a good point like how much does the recency effect kick in and I mean it affects humans that's a known thing like whatever we're focused on thinking about researching reading in recent memory, we tend to try to find it.

We tend to see those examples in real life more often just because our mind is set on it. Perhaps these algorithms are the same way. >> Yeah. So, it's, you know, interesting. But I think there's another uh aspect of AI to think about as as physicians, you know, I think there's two things AI will do.

I think it will give physicians a lot more autonomy to be able to go independent on their own because the barrier to entry into private practice uh into entrepreneurship will be much lower with the advent of AI.

uh you know um just to give you an example um you know when when when a patient comes in and they make an appointment uh there's a whole process of a human having to call another human at an insurance company giving that patient's information and writing down their benefits and putting them in into uh the EMR so then the office can see how much will the patient pay at their first visit or for a procedure uh getting prior authorizations uh you know taking in calls, collecting the information there.

It requires a lot of human time and capital. Uh uh with AI, those things will become much easier, much low cost uh for a person to be able to set shop, use simple softwares to set their clinic up and be start seeing patients, right?

Um but on the other hand in in in big corporations and in big systems and and through insurance companies it'll be easy to easy to monitor uh doctors pretty closely right if I'm using a AI software that's uh listening to the whole conversation and and assessing and writing a note for me and if everyone is using that through the EMR then insurance companies will exactly know how long I've spent time with the patient what kind of complex conversation I had and how I should be billing for it.

And uh hospital systems can definitely assess physician efficiency. Why did you talk to a patient about their their children's baseball game? You know, that was a waste of time. You should have spent less time doing that and you wasted three minutes doing that and said you could have seen another patient in that time. So there's good and bad. There there could be a lot more corporate control of physicians and maybe a lot more physician autonomy in some ways. >> Yeah.

I mean the the subverting of AI to create a type of surveillance um like a big brother type scenario. Um that's an interesting thought. I I hadn't even thought about that. But that I mean it's a potentially scary thought if you think about it that we adopt these tools to make ourselves in theory more efficient on a case- by case basis and to help more patients, right? Our goal is to see more people, more people can get in, the wait times decrease.

I I don't have to wait as long to see my patients.

Um but then if they're used like this by you know the powers that be the employers the administration whatever just to their goal is to make the enterprise more efficient the do the bottom line uh those are conflicting very much so and that's usually the problem in healthcare in general AI or not is the doctor patient relationship is often at odds with the corporate bottom line and if AI is used in that manner I mean it it could put even a further roadblock on a good doctor patient relationship and convert it even more into an assembly line type model.

>> Yeah. At that point, I mean, you know, would uh would you it's going to be tough for us to kind of manage our time then and be able to have that personalized conversation with patient. But I do want to also think positively about the idea of having more autonomy. I mean, you know, I don't have to hire two people right off the bat. I don't have to pay uh another company to manage my billing when all I have to do is come in have uh uh a lot of AI bots.

One may be answering phone calls, one may be uh doing benefits verification and uh one may be making calls and setting up procedures where I can come in see patients and maybe just hire one nurse who can room my patients.

Um, you know, I can have maybe possibly multiple um computer screens set up in my office where they can uh patients can come in and they can talk to a AI uh version of me on the computer asking the same exact questions I would be asking six patients at a time and create a summary for me for me to go in and counsel. >> I I love the idea of that. It's like we would have a Homayo version of Iron Man. It's like instead of Tony Stark, it's Homayu in his clinic with his white coat on in multiple rooms.

It's like his version of Jarvis just talking to the patients. Yeah. I mean, you can have a hologram of me in there uh six different versions of me uh who have been taught by me, have been taught by my u knowledge, my personality, and they've been trained by that. And they're in the room talking to the patient, collecting all the same information I would.

And then for the patients uh reassurance and I can go in based on the report that I've received and the suggestions that I have I can sign off on those uh talk to them and why I think that is a good idea. If they have questions maybe answer those and that way I can maybe see six patients at the same time and I don't have to worry about maybe you know instead of having doing uh 40 hours of work week I have a 20our work week. I can spend more time with my family.

I can uh I can do more things uh more work thinking about you know the deeper meaning of life rather do more podcasting.

Yeah, maybe I can spend more time podcasting with you rather than you know it's funny like um there was I forget gosh I forget where this was somewhere I think it was in Chicago Mayo but there was actually an orthopedic surgeon who this was before AI this was like I don't know maybe in like the mid 90s late 90s but he basically realized that he was the rate limiting step in his clinic that he had to see the patients talk to them about surgery sign them up so what he did is he basically came up with a series of videos like 2 three minute long videos of the 20 most commonly asked questions that the patients would ask him and he did pretty straightforward hip and knee replacement.

So it was not a huge breath of variety in these questions, right? It was these are the main heavy hitters and he just uploaded these little video clips of himself in all the computers in the rooms and he would just have his MAS go into the rooms and just start playing the videos and that would be the whole patient visit and he would just briefly pop his head in. Any other questions?

And if they asked a question that was answered in the video, he would just scroll real quick, click the video, and then leave the room, go to the next room. It seems crazy. And obviously like it was very off-putting to many of the patients. Um, but this would be I mean obviously not with the purpose of just running an assembly line, but this would be increased access to patients what you're describing.

I mean, in the future, but it would be more personal because it wouldn't just be a static video recorded in the past. it would actually be you or a likeness of you um that had learned and adapt and continues to just as you do. It would be like your surrogate in a way. But I mean it's an interesting it's an interesting idea for sure. >> Yeah.

You know I've t I've thought about a lot of these things and uh you know there there's tons of things you can talk about but one thing that always kind of boggles my mind and I think about is in mental health how grieving uh when someone dies is is such a big deal, right? Right. I mean, when you lose someone you love, um that's a big mental health uh change in your life. You know, it can cause depression. A lot of people, some people just have a hard time letting go.

Some people have a hard time grieving and miss people that they have. I mean, uh, you know, there's already, um, there's already, uh, things that Meta is doing, for example, in virtual reality where, uh, you can talk to an AI avatar that just exactly looks like the person that you love and know. Um, and they can be talking wearing their own headset and it it'll be emulating them in front of you on a virtual headset.

Now imagine uh teaching the avatar your personality uh and the things that you like, the quirks, your ideas, your wisdom, and it learning about it and then leaving that for your loved ones to come to you for advice once you're dead. Uh >> yeah, I mean it's like a it's like a deep fake but without the negative connotations that we associate the term deep fake with. It's like a legitimate real uh mimic of their personality. >> It's a a mimic of the personality.

maybe a way for people to slowly let go of you. Slowly uh go get through the grieving process. Uh go maybe even go back to you if you if they have any questions. Always have uh your dad, your brother, your mom with you to go to if you want to spend some time with them.

Yeah, maybe it's twisted in a way that uh you never really learn how to let go of them, but it could be therapeutic for some people who finally learn how to let go, but then know that comfort of being able to talk to their loved ones. Yeah.

Or even, you know, if you think like what if someone, you know, like a child, one of their parents died when they were still a child and they never really had an ability to build that interaction as they grew up and have a formative relationship, you know, like as they became a teenager or something, but the image and likeness was preserved with the advent of AI. >> That would be pretty interesting. >> We'll have to see what kind of uh, you know, mental health effects that has on paper people.

Uh, but it's something to think about. I mean the the the what I've learned through the past and we're you know uh as millennials we're one of the generations that has seen so many changes. I mean I still uh remember some of the early stages of a phone where you had to do the rotary uh dial and from that to a a wireless phone and then cell phone then iPhone now don't I can talk to patients from my people from my glasses. Dude, did you do you remember the car phones?

Like the old school car that was like >> in the car? Yeah. With the wires? >> Yes. >> We had one, but we never got it set up. It never worked. My brother and I just messed with it and pretend we're on the >> Yeah. So, we've seen that, right? We've seen the three different stages. We've seen the internet revol revolution. Now, we're we're living through the AI revolution. So, now I think nothing is nothing is impossible.

Like, I mean, just imagine a scenario in your head and it can possibly happen, right? There are some things I used to think that could never happen and now they're happening. So I've stopped limiting the the thought and ideas that oh yeah this could not ever happen. Um just just with the way things are moving there's so many different things that could happen. What excites me a little bit more uh speak kind of coming back to medicine is the use of AI and research you know and advancing.

You do a lot of AI research. I mean, you know, you you you're using a lot of AI tools and doing some research on AI, but AI and research can really fasten thing and maybe find cures for things, find uh uses of medications, uses of supplements, uses of uh procedures that we weren't even thinking of, right? Uh >> yeah, totally.

I mean there's even companies that are dedicated specifically to that that they these AI companies they primarily contract with like these huge academic medical centers like Mayo Clinic MD Anderson these centers because these centers have the best currency they have data they have huge amounts of data on patients of all kinds and so a lot of what these companies want is they want access to that data and you know they run through the data they identify like you're talking about clinical trial targets whether for drug design or radiation therapy whatever it is but they can just comb through this data way better than any human could.

And the question in this realm becomes who owns that data because in the past actually a lot of these academic centers they would just sell the data in some deidentified manner and it would then become the property of these AI companies and now they hold actually the most valuable resource. In more recent years these centers like MD Anderson have realized keeping the data is the most important thing.

you can contract with these companies that they can search your data and identify targets for these things, but keeping your data is the most critical thing because that's that's what these ASETs are trained on. >> Yeah. Yeah. I mean, the the data game is the strongest. I mean, someone once told me uh that the reason Tesla stock is so valuable because Tesla is not a car company. It's a data company.

You know, it's a company that's collecting data on traffic, on cars, on, you know, and that data is so valuable just like that. the data in medicine that is being collected through uh clinical trials and then that being processed through the AI softwares is it's the most valuable thing and a lot of the companies will be betting a lot of money on it.

Um, I think there was this company on Shark Tank uh I think a few years back that was uh collecting uh data on cat DNA uh and cats different breeds and what cats like and what kind of medications could be used and all it did was collect and this at at that time a lot of I don't think a lot of people bet big on that company but that company had a billion dollar exit recently because >> for the cat genome a billion dollar >> cat genome because so many different medications, so many different uh food types were created because of that data that was available.

>> So just like that, >> Marina Cat Chow, baby. >> Yeah. You know, that was a big scare with 23 and me when uh 23 and me went bankrupt. Um there was a big uh scare about where that data is going to go and what's going to happen to all those DNA samples. And they even offered to be able to delete all that data and destroy your DNA. and they gave that option to people to because uh once they were going to get bought out or or transferred no one knew what's going to happen to the data.

So that's a scary thought. I mean it's not it's a scary thought but it's also a hopeful thought because I think a lot of good can come out of the data right. >> Well for sure. Yeah. I mean you know it's like you know a lot of the research that I I've done in the last couple years is on like how patients use artificial intelligence.

like patients rather than just asking their primary care doctor, scheduling a visit, going to an orthopedic surgeon, they'll just query Chad GPT or Gemini and things like this and ask it like what should I do about my shoulder osteoarthritis? What should I do about my scafoid fracture in my wrist? And so we basically looked at some of the most common questions that patients ask regarding conditions like this and how accurate were the responses that these AI algorithms spit out.

And I mean they are actually remarkably accurate and we assess them on both accuracy and completeness because not only do you want to know if the response that was given is correct or not but also like is it missing a lot of stuff and in in fact it was very accurate and complete for the vast majority of things um which basically can help guide patients.

I mean it even provided initial treatment options like start with physical therapy you can do these exercises here's a link I can even come up with a custom physical therapy protocol for you here's some tools you can use. I mean like everything that they would essentially get for these simple treatment options. Obviously, it can't do invasive stuff. It can't do surgery. It can't do injections.

It can't it can't obtain or interpret imaging, but it had all of these initial very basic um steps for patients to take and without ever seeing a doctor. And for the most part, they were actually quite accurate. >> I mean, yeah, you know, in I I'm giving a lot of my data through this thing right here. You know, I wear the Whoop all the time. I've been wearing it for a year and a lot of insights I get on my health through that is is pretty amazing. But I know that data is going somewhere.

It's going to oop and they could sell it uh and use that data for maybe betterment maybe coming up with a little bit more information about her health. Uh they recently came out with this system where they can now monitor your blood pressure. All I have to do is check my blood pressure and input it into the app a few times at different timings and they can predict what my blood pressure is at a certain time.

And that makes me wonder a lot about wearables and the role of wearables in in in cardiology because big thing atrial fibrillation assess detection that's already happening with uh Apple watches right uh some of them some of the EKGs and AFA detection getting um FDA approval as well or FDA clearance but in the future it's going to get better.

I mean I put on halter monitors for my patients but in the future the wearables will always be able to detect hib and that's a big cause of strokes in patient right heart failure is a big big deal heart failure hospitalizations in patients is a big deal and there's a lot of work being done in that on how where wearables and different and blood pressure readings and weight readings can predict heart failure hospitalizations.

uh in my clinic uh a lot of my patients get a uh patients who have heart failure or hypertension. They I provide them with a blood pressure cuff and a weight monitor to check their weight and blood pressure daily.

So that data gets uh gets transmitted to me every day and my nurses monitor that data all the time and if something is if someone's blood pressure is persistently high and I can see their weight rising I know they're heading towards a heart failure hospitalization a heart failure exacerbation and I can write right there and then get in and act on it tell them to take their diuretics bring them in improve uh change their blood pressure medications and prevent that heart failure hospitalizations >> so that that right there is like a potential, you know, use case for AI rather than having a nurse or someone physically go and check every so often these metrics or whatever cut offs you set as an alert and these metrics, it could be automated using AI.

You know, >> it already is. It already is. It already is in a way that uh there's a lot of devices. I mean a lot of patients pacemakers have thoracic impedance set up in there where they can it can measure the pressure in someone's chest rising and they can uh alert the cardiologist that the thoracic impedance is rising and the cardiologist can right away act on it by diarrheasing them a little better at home.

Um, but the same is going to happen with wearable device where someone like maybe not having a pacemaker implanted but just by wearing a band like this can and maybe checking their blood pressure or weight every day can stay out of the hospital, stay healthier, live a better longer life. Um, and that kind of comes into uh m maintain their health. Another good way patients can benefit is as a cardiologist, as a physician, I can get bit better data from my EMR. Right now, my EMR is a billing tool.

I mean, all I'm doing is uh writing patients information for the sake of billing and maybe sending the reports to the the referring doctors, but in the future future, I can have my EMR work for me and work for my patients. uh I can have plugins or softwares within built into the EMR that can tell me, hey, 50% of your patients have a LDL or bad cholesterol higher than 70 even though they have heart disease.

Why don't you bring them in and improve their numbers a little better uh and can maybe highlight certain patients that need a little bit more urgent care based on their labs and their their metrics and that and that EMR can do two things. It can imp me improve my patients uh health by preventing hospitalizations uh preventing preventable illnesses like hypertension causing heart failure. Uh it can improve my business as well.

I can have patients come in more often uh identify patients that were lost to followup, bring them in because uh I can give them exact examples of how I can improve their health uh what needs to be done. All right. So those are the kind of things I think AI will improve healthcare in a lot of ways. There's a lot of negative aspects to it, but there's a lot of good things as well. >> No, for sure. I mean, I'm I'm way more excited about all the positives.

Like, it's important to talk about the negatives u like some of the stuff we've been highlighting, but I think the ability for increased efficiency, the ability to automate so many things, the ability for direct and instant communication, I think is huge. That's such a common complaint that patients have. And even myself when I am a patient or one of my family members is a patient.

It just the inability to talk to your doctor when you aren't physically in the room in those 10 in that 10-minute 15-minute window that the doctor's sitting in front of you. Like so many people have questions after the fact or you get an imaging study your labs, you want to talk about it. You want to know is everything okay? One value is on the cusp of normal, not normal. What does it mean? Is it okay? Do I need to do anything? um all of these things.

It it's very difficult for patients to try to speak to the doctor because there's so many different staff that are responsible for communication after hours with the physician or the physician's team. Some are more responsive than others. Some give complete incomplete responses. Some will communicate directly with the physician and make sure this is appropriate or not for me to give this response and some won't. I mean, there's so much variability.

I I think if there was, and I'm sure there are already ways and they're improving for AI to help with that aspect, too. Just the ability of patients to get their questions answered, I think is huge. It's very important. >> Yeah, for sure. I mean, you know, people a lot of times come to the doctor for that reassurance. Uh sometimes they know they might be okay. You know, this is just something they're nervous about.

A lot of people come to I'm sure this is just me being nervous or me having a panic attack. I just want some reassurance. Uh and yeah and and if I can just go in there, look them in the eye and provide them that reassurance instead of being on my computer and just typing away, you know, their whole story and just uh and then not having not not being overloaded by the idea of having to do all the work.

uh you know being offloaded by the need to write the notes, need to read all my imaging studies, need to make sure the staff is doing their job by making sure the benefits are verified and the patients are not complaining about the insurance charges. If that's all offloaded from me and I can just focus on patient care and be there uh for the reinsurance, then I can just be a better physician and a better doctor and provide better care.

So I think uh I think we as as as a healthcare community need to learn that these things we we can either use them in the right way and make our life easier or uh complain about them that they might replace us and and just uh kind of reject these things and then be left behind and that not not provide patients that that benefit that it could provide in the future based on all these things that could happen.

So I think uh I'm pretty positive, very optimistic about this and if it maybe replaces me as a physician but helps me as an entrepreneur make a better healthcare entity, I'm all all for it. >> There you go, baby. Stark, >> the iron man of medicine. >> Idea of six different versions of me in the room at the same time. >> I like it. I like it. No, >> maybe maybe the, you know, in the future we'll be the two different versions of us holograms doing this podcast. >> Now it's real us in in the flesh.

>> Yeah. Yeah. Two docs, several AIs, one mic. >> All right, man. This was a good episode. Let's wrap it up. Thanks for tuning in, guys, to Two Docs, one mic. We'll see you next time.