Artificial Intelligence Podcast: ChatGPT, Claude, Midjourney and all other AI Tools

Is Artificial Intelligence Going to Revolutionize Healthcare with Yauheni Solad

• Jonathan Green : Artificial Intelligence Expert and Author of ChatGPT Profits • Episode 346

Share episode

Welcome to the Artificial Intelligence Podcast with Jonathan Green! In this episode, we delve into the future of AI in healthcare with our esteemed guest, Dr. Yauheni Solad. Dr. Solad is at the forefront of integrating AI into medical practices, balancing innovation with the cautious pace necessitated by patient safety and regulatory environments.

Dr. Solad discusses the challenges and opportunities AI presents in healthcare. He emphasizes the importance of thoughtfully integrating technology, suggesting that while healthcare cannot adopt a "move fast and break things" mentality, there is room for more agile, yet cautious, advancements. The conversation covers the slow buying cycles for medical technology, data privacy concerns, and how current legal frameworks like HIPAA need modernization to address today's technological capabilities.

Notable Quotes:

  • "In healthcare, we cannot move fast and break things because breaking things here means a lot of negative things." - [Dr. Yauheni Solad] 
  • "We live in a free society. As long as I'm informed and I am an adult, I can do anything I want with my data period." - [Dr. Yauheni Solad] 
  • "The biggest assets you have, from an economic perspective, is your data." - [Dr. Yauheni Solad] 
  • "The way you deliver care... there is absolutely no societal or individual care benefits of automated, but systemically biased systems." - [Dr. Yauheni Solad] 

Connect with Dr. Yauheni Solad:

Podcast: https://realignedhealthcare.com/

LinkedIn: https://www.linkedin.com/in/solad/ 

Twitter: https://x.com/ysolad

Connect with Jonathan Green

Is artificial intelligence gonna revolutionize healthcare? Let's find out with today's special guest, Dr. Johanni Sahd on today's episode. Welcome to the Artificial Intelligence Podcast, where we make AI simple, practical, and accessible for small business owners and leaders. Forget the complicated T talk or expensive consultants. This is where you'll learn how to implement AI strategies that are easy to understand and can make a big impact for your business. The Artificial Intelligence Podcast is brought to you by fraction a IO. The trusted partner for AI Digital Transformation at fraction A IO, we help small and medium sized businesses boost revenue by eliminating time wasting non-revenue generating tasks that frustrate your team. With our custom AI bots, tools and automations, we make it easy to shift your team's focus to the tasks that matter most. Driving growth and results, we guide you through a smooth. Seamless transition to ai ensuring you avoid costly mistakes and invest in the tools that truly deliver value. Don't get left behind. Let fraction aio o help you stay ahead in today's AI driven world. Learn more. Get started. Fraction aio.com. Now I'm really excited to have you here because usually healthcare moves very slowly. Because there's a lot more red tape, because there's so many things that are important, like patient confidentiality, and we require longer studies. And what we've seen is that some companies just jump right into AI super fast, but in healthcare especially, we have some more protections that kind of require a little bit of a slower transition. So do you think that's a benefit for like hospitals and doctors that they can make a little bit more cautious decisions or do you think it's gonna leave them behind? I. Jonathan, thank you for inviting me. Things designed for a reason, and I know we like to blame a lot of healthcare participants in a slow rate of adoption, but in healthcare, we cannot move fast and break things because breaking things here means a lot of negative things On the same time, there is always a way to move faster, right? So can we move as far as the front end frameworks? No. We cannot, can we move faster than the glacier pace we've been using in a lot of industry? Yes. So I think the question is not. Is healthcare capable of moving faster? Healthcare and pretty much any innovation you look at, it's definitely capable of moving faster and putting more efforts in a very targeted discoveries. Very often in the healthcare especially, things you talk about integrating the data with legal frameworks. It's coming not from, inability of a new technology to implement or ideas how to do it safely. It's coming from, analysis paralysis, pitis where you're trying too many things and cannot come up with, there are things, or frankly, regulatory confusion because the reaction when you don't really know what to do, right? Even if you know how to do it safely. Is either not do anything, which is, most common reaction or something that's actually help us to propel to the current state is design it as a clinical study. Something that we can do under, in, internal board review and do it this way. Which is great. It's a great way to create evidence and do the real science in implementation world when you have to build, previous technology that's been already proven or maybe technology that may improve your experience and not necessary, make like a clinical impact. A lot of time this is maybe significant overhead that slow things down. One thing that's. Always fascinated me is that in medical field, the buying cycle's often very long. If there's a new machine, whether it's a CAT scan or an MRI, it's often a three year cycle between when you first start thinking about it and when it shows up in the hospital. And when I think about all those patients that maybe needed that particular scan during that time, why is the buying cycle so long and is three years accurate or is that just the government. I, again, if you think three three years of buying cycles, it's usually shorter, right? If we talk about proven equipment, to be honest, because there is no reason for it to be so long. The government procurement is notoriously slow, but and not necessary in years. Years. Think about how many years does it take for this kind of machines to get into the. Because I think the sourcing and procurement from the proper vendor and registering in the vendor system and selecting the one is one thing, but decades and decades that I take from research, I. And proven machine to actually get in the market. I think that's the real problem because we always can fix supply chain if we truly have the need and if people trying to get it, there are way to do it. For example, even for MRI Machine that you cited, yes, it may take very long for you to actually get one in your hospital, and a lot of this is a combination of you financing, installing, finding the right . A place to install it, right? Preparing your hospital. But there are way to address it. You can put, external machine to start using it immediately and bringing it to your patient if you if you want. But if new type of AI driven technology that's implanted in your, cardiac device, for example, trying to enter the market. And right now, for example, there is a not very clear pathway for the non-deterministic, generative AI driven devices. That's maybe a problem. That's, that definitely will make things longer and more complicated. One of the things that I sometimes think about is you have confidentiality with a lawyer, but if another person's in the room, the confidentiality is broken, you don't have it anymore because there's someone who is not part of the confidentiality nest who's there. And when you're having AI in the room, we haven't really figured out where an AI fits on the spectrum. With HIPAA and with the importance of kind of patient records and that there's this worry that the AI will remember or will learn from things it's not supposed to listen to, and that even when you promise all these big companies go, we promise the AI's not listening in these situations. And it's thank you for your promise. But it's like I sometimes think about when there's like a court case and they say to the jury, forget that he said that. And it's now they're definitely not going to forget that, right? Because it's like the most exciting thing that's happened all day. It's you can't, how are you gonna make me forget it? It's the first time you started smashing the gavel down and you're saying to forget it. Now I'm thinking about, it's like when I say don't think about a pink elephant, always about speak elephant. So I just wonder if that's one of the areas that causes hesitation. There's this. Fear of the incidental HIPAA violation or kind of this hesitation because it's now, it's a learning machine and it changes the dynamic because it listens and remembers everything. So machine by itself cannot do anything right now. We're still in 2025. I know we are very concerned about Skynet like technology, but let's get things straight. Okay. Our current generation is not ent. It cannot do anything without humans doing it. I'm not sure what promise truly means in this case, right? Like you, you have a legal obligation you need to follow and you need to have a disclosure that you need to do and actually enforce. For example from the MBN documentation, like the data collection. For example, when the doctor putting a technology in the room, when, I'm seeing you as a patient or someone doing it and this AI tool doing my notes, right? It's helping me to document faster, which by the way is a great thing because it's helping so many clinicians not to stay longer at the job and cutting by times. The way system implemented, it's used for one particular documentation parts, and then as soon as the visit is done, it's removed, right? If I have to reuse it for anything else, for example, model improvement, I need to have explicit. Agreement. Which is not promise. It's a legal agreement with the vendor to actually purge the data. And I, when I buy it as a software have to review, it is actually part of my responsibility as a clinical leader and IT leader to ensure that a, I understand how my data being collected, reused. Outside of intended use case, even in an anonymized perspective, I think that's probably the biggest fear you verbalize because unfortunately we have a lot of instances where vendors may get out of agreements with a healthcare providers or maybe technology company anonymize the data and then start reselling this for the outcomes. A lot of things need to stop and frankly, some of the delay in bringing smaller vendors that may allow you to actually do fast experimentation is related to that exact point. For example, every time I bring a vendor that's new in doing something, invasive from a data perspective, I need to ensure that the vendor a, not just have it in writing, but truly knows what they're talking about. Because it's pretty easy to use chat GPT or standard legal templates and say, oh, we are doing all of the above without really doing it. Because a lot of us do not have mechanism of auditing every vendor. We have to believe in third party certification. And that's the beauty of a good certification. Unfortunately, when you talk about the very small vendors pilot level. A lot of them may not have it yet. And say, we gonna do it when we reach a certain point. Those risks are unacceptable for many enterprises and they never do. The pilots, some other groups, the, especially the one with innovation hubs going and spending time. But in the core of this is actually outdated legislation that we have right now because HIPAA was created in a world where we had facts. HIPAA was not created for the web world. HIPAA was not created for, the world where we potentially have an incredibly efficient systems that can collect and analyze gigabytes of information a minute and pull from multiple data sources. So we need to rethink a lot of the time. And unfortunately the current model of incentives still incentivize actors to source the data and sell the data. It's true in social media. Which is unfortunately following us, right? And I think you've seen a lot of this unfortunate reports where ad networks were following your location even without your explicit permission, which is unacceptable. And similar things happen in some of the tracking, unfortunately, from the healthcare system. Because if I'm not a healthcare entity if I'm not unsure, if I'm not a clinician. I actually don't have to follow. So if I'm just a random, step tracker that's saying Jonathan, can you please give me access to your health records right on Apple or Google Fit? And you give all of that and they pull your medical information, you're actually not protected by hipaa. So don't, you don't even have to bring it. They can actually anonymize and do anything they want. Inside what they describe in their user agreement and those user agreements usually very vague. Yeah, I've seen that. It's the, it's like the patients who are the wildest with their data where they say things like, I saw there was a campaign on Twitter to upload your x-rays or upload your scans. And I'm sure plenty of people I know we do this where we'll get a result from the doctor and x-ray and take a picture and have the AI scan it and get feedback, but. Try not to certainly not your entire set of medical records, right? So sometimes it really helps to understand something or you get a prescription, it's what does this drug do? What does this drug do? So there can be some usefulness there, but I certainly, there are people who maybe they do upload their entire record or to a social media platform, which obviously is like the wildest decision you can make. But we see more and more where people, like they share more and more of their private lives via social media. Like they wanna be an influencer and they share like stuff that you would never talk about or places you would think are super private. Now they're doing dances and more and more private information in the background of these videos and the AI can scan everything. So we're starting to get into this world where data is just everywhere and. I'm really fascinated by the thing with 23 and me, right? Everyone gave 23 and me their most intimate data, right? Their genome, and then it's the board of directors resigned. They, their idea of like DNA health never worked, or they have the research part never worked out yet, and the running outta money, and if someone else buys all of that data. They could do whatever they want with it, right? Because now it's even, it's not even limited by whatever kind of terms of service or conditions or promises that the first company made because the second company bought it without making any of those promises. I. So I'll start with social media, Jonathan first. We live in a free society. As long as I'm informed and I am adult, I can do anything I want with my data period. Like I'm not here to lecture you on what you do. If you want to start publicly uploading your medical data, your sequencing data and everything. God bless you. That's your choice. And there are certain benefit to a lot of that, right? Like co-discovery, especially for rare diseases or some of the thing that's maybe harder to manage and you are learning from community. So there are certainly, benefit of sharing this information as long as you are informed and it's your conscious decision. I, I think the biggest challenge is happening where a lot of this happens under the umbrella of safety and protection. I'm going into the some app where I expect some degree of privacy, right? Just call some kind of a doctor or maybe some intelligent chat and instead of, just giving me an answer because misunderstand, understanding of a clinical information is a real thing. So people are seeking a lot of help into all of this. So instead of just answers, I'm getting, so far maybe incorrect hallucinated answers and unfortunately harvested my information. I think that's the real with. With X. It's an interesting part because if you upload your x-ray to x I don't think you have any expectation of privacy like that. To me, that's the first point where like I am pretty aggressively informed that this is social media people will use all of it. So as long as you realize that, that's not everything is coming. Please do. Helping us to build bigger multi-model or any kind of AI models is very helpful for the progress of medical technology, right? So if everyone, will come up with a good way to publicly or privately preferably start sharing information that will help us to get faster to a lot of the, there is no doubt around that. 23 and me is a very PE case and I wanna see where it's. Will end up right. Like I, I don't think we have a clear answer. Will they sell a lot of this? And will they sell and all of this information will suddenly become available to private equity. Or something else that's maybe used not only for good but for something that's may end up being challenging for people who paid in this. Yet to be seen, right? The reality is you do not own your data. E even if you pay for that, because a lot of this was subsidized the agreements change . So if your data retained and if you're not the one controlling your data you have to be ready. That rules may change. Cost may change. Ultimately, so far the data is still the biggest assets you have, right? And from economic perspective, if that the biggest assets you have, like what else do you expect someone, especially when the company did not have a better way to monetize? We'll try to monetize their biggest assets as clear as they, no matter what they say, unfortunately, and I. I don't wanna be pessimist or start alarm bell, but I think this is the reality we'll have to deal with for the, next decade possibly, where until we have a super clear regulation about what you cannot do and start pushing in that direction about your data ownership, retention, and, data reused, which already getting better, to be honest. Still in this dark, so think. So carefully select people, right? And technology and the company you trust. And we were talking in February of 2025, right? So deep seek part is and hype and scare is still here and high is. Maybe the company that's even giving you things for free and claiming we're following, particular country law. Not other countries law. Not the exactly. Application. You should trust all of your data, even if it's gonna give you, some of the helpful feedback. Yeah. If it's like that saying, if you're playing poker and you dunno who the sucker is, it's probably you. If it's free, if you think something's free, it's probably not it just seems like it's free, but there's gonna be a surprise at the end and. Too good to be true. Usually it's good to be true. I think that it's a really interesting time where I can see why with so many data things happening, why people are hesitant to jump in because there are different models all the time. Oh, chat GBT, that was so last week. You need to be using Claude. No. Now it's about Deep sea what Deepsea is over. It's about Deep mind and it's very hard to keep up with even when it's my full-time job. Even when I was reviewing Deeps Seek this weekend, there's such a difference between Deeps Seek R one and Deeps Seq V three. Completely different models. They sound like they're the same thing., the names are so similar. So I can understand why in the medical field people are hesitant for the doctors out there and people in the medical field. What do you think is the right approach? What should they be doing as they look to use AI specifically?'cause what every doctor really wants to do is. Give a better experience to their patients, whether it's more accurate diagnosis or a better bedside manner, or just a shorter visit. What do you think is the direction to go in and where's the future going? I. Great question. Frankly, we can spend the whole hour talking about just that because we are in such an inflection point where healthcare is poised to change drastically that it's important to actually work in, educate on both clinical side and the patient side because a lot of this progress requires you to rethink the way you deliver care and. One of the core thesis of realign healthcare, by the way, is identifying the weaknesses we currently have in our incentive structures and the way we deliver healthcare, and ensuring that we address it first before we fully allocate, because there is absolutely no societal or individual care benefits of automated. But systemically racist, biased economically harmful systems, but that's now been powered by AI and being encoded by ai. So from a technology perspective, like if you look at technology that's been wildly successful in healthcare, one of this technology that's been around and refuse to die is facts. And I would say you take any generation of innovator for the last time, 30 years, they'll say, you know what, we should address the fax problem because fax is anything, right? It's not secure. Before it was encrypted and digital fax, right? It's flat files. You cannot extract the data. E everything is wrong, but fax is still here. Why? Because it's easy. It's reliable. Everyone can figure it out. And a lot of those same things need to apply to technology. We bring R'S Medical Records is a great technology for data storing and you'll find a lot of the benefits and a lot of challenges in literature, but it's anything but easy. It has a pretty steep learning curve. You need to learn the way right, the epic way of doing it, Cerner way of doing it, Meditech way of doing it. And that's limits your benefits while you're transitioning. And that's make a lot of the basic things hard and that's increased a lot of our administrative over. Because you started to, generate more of the notes and then you hire more administrators to do some of this additional parts, and then you put administrator on top of administrators, right? And create this, human human flow of a technology and now you have an, a lot of AI entering the picture that can help you with, and now you have to balance, excitement of, oh, I, I have for the first time, I have this incredible technology that can work with on structured data and sell me all of those things with the fear that you realize because you work with a model is. This is non-deterministic technology, and very often if you ask the same questions four or five times, you will get different variation of your answers like you depending on your temperature setting, but still it's not gonna be precise healthcare like precision. That's why we like facts. That's why we like actually forms with checklists. Because very often there is not enough variation, especially when you are talking about particular conditions, right? Or particular real. So the way you have to start look at that and deploy is to find a scenario where your human, feedback, right? Where your reinforcement loops are strong enough for you to capture those tales of errors that the system. That's why people started working well with ambient documentation because ultimately it's working almost like smart intern. It's entering the room with you, right? It's helping not only to listen for your discussion with the patient, but it's also pulling some of the medical records that help you to, hand your orders. It's help you to do some of the tasks that's you're gonna do. But AI is assisting you, but ultimately it's your responsibility and it's expected that you as a clinician will review every single line, every single order and place it. So from a technology perspective, you may have incredible incentive on can we just automatically deliver care? How can we start doing this integrated healthcare pos that you are walking in, you're getting a full assessment, AI doing incredible things, and then writing you an order or you go to full triage in getting a treatment. And there is a reason why unfortunately, a lot of those companies are, not around, right? Or bankrupt or restarting once again because. Even though from a side healthcare may look out what can go wrong, it's actually a specialty right in the medical field with incredibly long tails. And AI can do relatively well, right within a few deviation of the norm. But all those tales of patience, of condition, of information, all of those tales with unstructured data that you cannot present well because AI doing very well in the widget style things. Are very hard to deal with a current generation. I'm not saying it's impossible, right? Like we, we likely will get progressively better, but we need to ensure that we innovate within few standard deviation of things. AI can reliably do well and do not burn the bridges with regulators, with patient trust, with clinician trust by doing that, because if we start doing too much too fast, I. We definitely will have negative outcomes, which lead to more regulation, which slow down innovation. Yeah, I think you're exactly right that it's that finding that path that's not too fast and we make mistakes and not so slowly to get left behind. That's the trick. I really do. Appreci outta your time. I think this is really valuable for people who are, 'cause more and more startups and more and more things are happening in AI and medical space and it's one of the places where. People just like when PubMed and WebMD first came out and you could suddenly look up what you have online. People just begin self diagnosing more and more. And so we do also have to find that you. Area where it stays useful without kind of indulging.'cause if you ask in the right way, the AI will say, oh yeah, you definitely have what you think you have. And it starts to create a cycle. So I think this is really useful to help people have a more methodical approach. What's the best place where people can see what you're working on now and find you online if they want to hear and learn more about your projects and amazing things that you're doing. First please stop by to realign healthcare the podcast that's, we started in the fall with few of my colleagues, talking with the industry leaders around the way we need to realign the incentives in our current healthcare system. To help actually lead to better automation. LinkedIn and Twitter I'm pretty active on those. Please connect. And happy to engage in discussion. And from innovation perspective, Jonathan the healthcare is such a great field for innovation because all of us have personal stories. This possibly is one of those fields where in contrast to, you know, nuclear physics, you will have at some point in your life, personal experience and you will have a strong opinion of what can be. Your, the way your schedule, your appointment, the way you know your family member was treated the way unfortunately, maybe you've been treated or anything else, you will have an experience with that. And what a lot of those tools are doing right now, they are democratizing access not only to, the pub met like information, but to very specific domain knowledge, for example. When, 10 years ago, we were working with startups and the startup will come and say, you know what? I don't really know how to integrate with EHR because I never heard of fire, or I never heard of, a mob mapping, so I don't really know. We would say, yeah, it's happened, right? This relatively small domain. Right now, there is absolutely no excuse. Because if you go and you ask Chad GPT free one, you don't really need a fancy model and say, how do I integrate with a healthcare system? And you say, I want to read blood pressure and write blood pressure. You will get good enough description and resources. And I think that's part of a power. You don't really have to do a very fancy. New bleeding edge, clinical decision support, innovation, medical robotics, but this power of many of democratizing like individual groups and developers with the access to the very specific domain knowledge and based on infrastructure we already have, at least in United States, like you already based on 21st Century Cures Act, can access your information. From your, providers, from your insurance. You can request that. You can analyze this and you can build the system a lot. This part is very powerful. Amazing. Got me excited for the future. Thank you so much for being here today for another amazing episode of the Artificial Intelligence Podcast. Thank you, John. Thank you for listening to this week's episode of the Artificial Intelligence Podcast. Make sure to subscribe so you never miss another episode. We'll be back next Monday with more tips and strategies on how to leverage AI to grow your business and achieve better results. In the meantime, if you're curious about how AI can boost your business' revenue, head over to artificial intelligence pod.com. Slash calculator, use our AI revenue calculator to discover the potential impact AI can have on your bottom line. It's quick, easy, and might just change the way. Think about your business while you're there. Catch up on past episodes. Leave a review and check out our socials.