Doctors are using ChatGPT in clinic and not all care about privacy (Health.Tech 2026)

[00:00:00] Dear listeners, welcome to Faces of Digital Health, a podcast about digital health and how healthcare systems around the world adopt technology with me, Tjasa Zajc. Doctors are using ChatGPT in clinics right now and some of them are also putting identifiable personal information in these models.

[00:00:23] This was one of the topics that we addressed at Health.Tech in Basel in a live Faces of Digital Health recording where we unpacked the convergence reshaping clinical software.

[00:00:37] We discussed ambient AI scribes, the current state of agentic AI in healthcare, how is the field developing with on-device LLMs and the regulatory drag, especially in Europe where providers need to take into account the Medical Device Regulation, EU AI Act, EHTS and more. This is widening the gap between what clinicians actually use and what hospitals are allowed to buy.

[00:01:06] So on the stage in Basel, I spoke with Jonathan Bringas, CEO and founder of LabSea Health, a provider of Keiku, an FDA-cleared AI stethoscope, ambient scribe and clinical assistant in one device. The second speaker was Blas Striglau, CEO of Mediatli, a drug information platform that has over 1 million healthcare provider users across Europe.

[00:01:34] And Amanda Herbrandt, clinical data modeling consultant, formerly working at the University Hospital in Basel as a data architect. What you're going to hear in the upcoming discussion is why is EHR data fragmentation the precondition that AI hasn't solved yet.

[00:01:59] We discussed the challenges with shadow AI, so clinicians that trust chat GPT more than enterprise tool, the conversions of various tools and how do we measure ROI of AI in healthcare and much, much more. So enjoy the discussion and also check out our newsletter, which you can find at FODH.substack.com. That's FODH.substack.com.

[00:02:26] And you can also see a video recording of this session by going to YouTube and looking for our Faces of Digital Health YouTube channel. Now let's dive in today's discussion.

[00:02:40] A special session on what the doctors want and how is AI changing the game. The trends that we've seen happening in the last year actually, or maybe a year and a half, are quite significant. AI agents are entering healthcare.

[00:03:10] It's becoming easier and easier to create apps. Some people are even saying that we're entering the trend of one user per app because with wipe coding, things are becoming much easier. And I'm here to discuss these trends with three experts. First of all, we've got Jonathan Bringas, the CEO and founder of LabSea Health. LabSea Health is the creator of Keiku. Keiku is a digital stethoscope. That's how it started.

[00:03:40] But today, this is actually a device that's also a scribe for clinicians and an AI assistant. So welcome, Jonathan. Thank you so much. We've got Blas Striglau, the CEO of Mediatli, a company providing drug information to clinicians across Europe with more than one million users across the continent.

[00:04:00] And then we've got Amanda Herbrandt from Basel, a clinical specialist, clinical data modeling specialist who was the team lead in the university hospital in Basel and is now working as a consultant. If you know anything about data in healthcare or EHR systems, you will know that it's complex. It's talking about data standards is messy.

[00:04:27] So when we think about agentic AI and thinking that AI is going to solve everything from an IT perspective, that doesn't seem as a reality we will see very soon. So Amanda, how are you observing the latest trends with AI based on what you know about EHR systems and their architecture?

[00:04:50] So I think the fundamental problem that we have in the landscape of data in the healthcare sector is the fragmentation of the data, being it within one healthcare supplier as well as between different healthcare providers. So the question is, can AI solve that? Or isn't it actually the base for AI? So do we need to solve the problem of fragmented data first before we can actually implement reliable AI, be it agentic or LLMs or whatever?

[00:05:19] Jonathan, you started your product or your company with the thought in mind that clinicians care about the user experience. We know that burnout is a big issue. How did you start knowing that digital status scopes were available on the market already? And what are you observing in the last two years in terms of the expectations around UX for clinicians with AI?

[00:05:48] I think that the UX conversation is really important because we're all building a lot of different technologies. And the way we present them to the physicians or to the care teams is probably as important as the technology itself. Otherwise, adoption doesn't really happen. If you're not really having pleasure in technology nowadays, it's just not going to happen.

[00:06:08] And when we built Keiku and we said, okay, we have this incredible machine that has the potential to really do a lot of functions, have a lot of functions in the clinical space and be able to be a diagnostic tool. How can we then bridge the other part, which is also very important, which is the expansion of large language models and generative AI? How can we also bring that into a context of one device, solving multiple questions or multiple problems in healthcare?

[00:06:34] And how can we bring all this technology and convert it into a workflow? So what we thought is a device that you're carrying with you all the time, like a doctor's carrying their stethoscope all the time around their neck. So if we replace it with this device, you carry it all the time with you. So ideally, you're also going to be able to transcribe every single conversation you have yourself. You don't have to take out your phone. You don't have to go and look for your laptop or whatever. You can just basically deploy generative AI, ambient AI.

[00:07:02] And then we thought, OK, so if we can do that and we have that context, we have that transcription context and we have the clinical context, then we might as well also be able to respond to questions like an evidence tool, but also contextualize on clinical and diagnostic events. And then we can generate a machine that has more usability for a clinician than the fragmented tools that are right now exposed to them into one convergence of technologies.

[00:07:30] And what we did, and what we see is that doctors are loving it. They are understanding the workflow. We built it really simple. This one app, you log in. Now we're going to deploy voice commands to the device. So you don't even have to use your app or your computer and it will become a medical assistant that you can just have in your pocket. I feel that is the way we go now with technology, just deploying physical AI devices and letting them solve issues for the consumers.

[00:07:57] So in essence, in your particular case, AI significantly transformed and helped you kind of grow the product and improve the product quite faster. 100%. Blas, what about in your case? As I mentioned in the beginning, Mediatly is providing information about drug specifics to clinicians. And now we've got LLMs.

[00:08:22] We have open evidence, the top startup in the medical space valued at 12 billion US dollars, giving open access to clinicians to the information that's backed by peer-reviewed clinical articles and scientific research. What does that mean for you? How are you adapting to AI and all the competition that has come to the market with AI? Yeah, it's definitely an interesting space.

[00:08:50] Much more interesting than when we entered it almost 14 years ago. Some of the things that we learned back then are coming in really handy now. So for example, if you look at open evidence and the approach of the tool is free for everyone. Everybody can use it and we just focus on getting directly to the doctor, right? That was a lot of what was behind our success when we started in 2011 and how we got to 1 million HCPs relatively quickly, right? What I think is the really interesting aspect and what Jonathan mentioned plays a little bit into that.

[00:09:21] Companies or products who started as maybe stethoscopes are going into ambient scribing. Ambient scribes are going into EHR integrations and clinical decision support. Clinical decision support is going into ambient scribes. Us as drug information support are going into expanding, again, thinking about ambient scribes, thinking about integration.

[00:09:38] So the space is really seeing not just an increase in competition because of AI, it's also seeing an increase of convergence among every player that was put previously in different parts of this ecosystem. Now just being in almost every part because a lot of these pieces became so much easier to execute with AI, right? And in our specific case, part was because we talked to physicians so much, we were already at the end of 2023.

[00:10:05] 2023, we saw, okay, physicians are already trying AI a lot. They don't trust it, but they're trying it a lot. They're trying it to the point where they're throwing in patient data. They don't care about privacy. And this wasn't like, oh, great. This was like, oh, this is both a warning. But at the same time, if they're willing to risk that much to use this technology, imagine when it actually gets better. Because at the end of 2023, it wasn't there. Now, two years later, it's way ahead and it's just proven now to be this extremely fast growth curve.

[00:10:35] And we're trying to ride that wave rather successfully so far. Amanda and Jonathan, you both are clinicians by background. So I really want to hear what does your doctor's head say on what we see now? The fact that, as Blas mentioned, clinicians are using so-called shadow AI, just chat GPT tools that we use as patients as well.

[00:10:59] And at the same time, are potentially reluctant to use AI solutions that would be introduced top down from the enterprise perspective that are safe, but also give more transparency to what the doctors are doing, which is not necessarily always desired. So how do you look at that and how we manage both ends?

[00:11:23] So I think the interesting part that we can take from this is actually the question, why do doctors like to use AI by themselves, but not like to use it top down? So I'm wondering if this is actually the participation aspect of it. So when they go and use AI on the Internet for themselves, it's something that they do actively. So there's a much bigger part of them contributing to find a solution than when they've been given a tool from top down.

[00:11:47] So I'm wondering if we shouldn't use this motivation and this kind of also problem solving skills to engage them in the overall process of building a solution together with the clinicians. Jonathan? Yeah, I think that's definitely one of the reasons for sure, if not the most important one. And maybe also the fact that they are, the solutions that we have are very limited for clinicians. So what we realize is that even open evidence is very limited.

[00:12:17] If you look into ChatGPT, you can just generate images, build your PowerPoints, videos, right? You can also ask about like a cooking recipe in the middle of your questions, right? So believe it or not, we need a lot of flexibility for technology to be able to really succeed in the space of user experience. And what we did is we were trying to create a flexibility layer there on our technology just to be able to give that opportunity to clinicians to do other things with it aside from just one intended use.

[00:12:48] I feel that and maybe that is the very sad word of medical devices or medical technologies is intended use. We should have something that is just general use, in my opinion, that would really engage the clinicians. Speaking of engagements, and Amanda, you mentioned that clinicians should be involved in the start. One of the big questions at the moment for hospitals and healthcare providers looking at implementing or buying solutions is like, how do I measure ROI? What should be the priority?

[00:13:18] Should it be the clinical side? Should it be the efficiency? Should it be the financial side? And I actually ran a workshop yesterday about what should be the ROI, who should be involved in the assessment, and what should the end users be told so they can actually trust solutions that they are given to use. And the conclusion was that the user should be involved from the start so you know that you're actually addressing an unmet need

[00:13:47] and not applying a solution to a problem or just looking for a problem for your solution. So, Blas, let's start with you. Immediately, with all the users that you have, you also do annual surveys to measure trends. What are you observing? What kind of feedback are you getting in the last few years about doctors and technology? And then we'll go to Amanda and Jonathan as well on this. Sure, yeah.

[00:14:15] Like I said, very early on, back in 2023, we usually run the regular surveys, which are around 10,000 physicians. So, it's a pretty good sample to understand what they're actually thinking, what tools they're using. And then we can see trends. Okay, this is something new. This is a big change coming. And what we see is what everybody is observing. Very rapid expansion of AI tool usage across the board. It's lost as one tool, a very rapid expansion of usage, very rapid decrease in the lack of trust.

[00:14:45] So, initially, the lack of trust was very high. Usage was high, but also lack of trust was very high. Now, the trust is dropping. The usage is going even higher. And yeah, it's just that the amount of use cases that doctors use these tools for is really much more than people imagine, right? This is not to say all of these use cases are super valid and super accurate and that the tools today and like the LLM models today are good enough at solving them. But you will have physicians and not the young physician. You will have quite old physician coming up and say,

[00:15:14] I've never used the MoMo and F because this is too much techie for me. But now I do everything through ChatGBT, which is both, of course, a major red flag in a sense, but also testament of what the UX of no UX. So, we can just do, I can just type anything. What is it actually doing? And it's the first time that physicians have a UX that's so simple that they feel comfortable using, even just off the bat, not knowing anything about the technology. Jonathan, what does that mean for your product development? ROI.

[00:15:44] I love that question because you were saying we want to save time, but we're also saying we want to accumulate time usage in our platforms, right? Which is like a contradiction in itself, right? So, do we really want to decrease time? That's a question that I'm still wondering. Financial ROI, that's for sure. Especially if we're selling to enterprise, we have to have an ROI model.

[00:16:10] And then there is the other ROI, which is clinical accuracy, which it used to be the ROI five years ago. And apparently everyone is forgetting about it nowadays. So, we're talking a lot about time and money, and we're not talking about clinical accuracy. And that is sad for the patients. So, ideally, you want to have a product that has the three of them embedded, right? ROI for financial, for time, and for clinical accuracy.

[00:16:38] I hope that we can eventually become that platform. And then the other thing is about no UX. I love that. I agree with that. I want our next goal is to be able to just have voice commands to this device. So, there's actually no UX whatsoever. Yeah, yeah, yeah. Amanda? I would actually like to add a fourth dimension to the ROI, which is user satisfaction. Because even if I spend the same amount of time documenting, interacting with the computer,

[00:17:07] being it with an AI or the typical system, it might be that even if it's the same amount of time, I might be much less frustrated with it. Because it's not a death by clicking, but it might be something that feels more natural to me by interacting with it, like you would do with another human being, putting all, let's say, safety aspects aside. But just the pure form of interaction, changing that, I think is a really important point as well, especially when you think about clinician burnout and so on.

[00:17:36] Stat News actually published an article recently about research with time savings with scribes, especially in the US market where they're very prevalent. And as it turns out, they don't actually save that much time. It's maybe 20 minutes a day. But clinicians are so much more satisfied. And from the healthcare system perspective, that's really important because many clinicians left clinical practice because of burnout and the unfriendliness of technologies.

[00:18:04] Which also brings us to the next question. We've got AI. Technology is developing much faster. How can healthcare systems, hospitals, healthcare providers, know what to choose, given how easy it is to design new solutions? I had a discussion, informal discussion earlier, with a hospital leader who said, I just don't have the capacity, I don't have the people that could assess all the apps

[00:18:33] or all the solutions that are coming to me and compare which one to choose. Maybe Amanda, since you're now working as a consultant, what would you advise? How would you go about choosing the right solutions? Because also from the company perspective, as a startup or as just an established company, you don't want to spend one year on a pilot that goes nowhere. That's a really good question. So I'm thinking towards a more modular way of solution building.

[00:19:03] So if you would have, let's say, described modular functionalities, if you say, okay, this is the functions I wanted to have, that you can really clearly phrase your requirements, that might help you to find a solution that is possible. That still requires, of course, some input and work, especially from an IT department's point of view. But I think in this environment and times that we're living now, it's really time for the hospitals to step up and let's say emancipate themselves from a pure consumer

[00:19:32] to an actual agency as a customer in the tech world and grab that responsibility and build their own knowledge and portfolio to make these kind of decisions. Blush, Jonathan, do you have anything to add? You are selling your solution. So what kind of arguments or approach do you take to make sure that you're seen as a competitive player or a reliable partner in healthcare delivery?

[00:20:01] We are using our moats. So we're FDA cleared. We have launched our entire platform for AIs embedded in a regulated medical device. And that gives us an aura of trust. When we go to any hospital or enterprise partner and they're like, oh, but how do I know that? I'm like, is it cyber secure? We have FDA rigor. We're a class two medical device. So there's no other company in the space of LLMs

[00:20:30] for hospitals that has the rigor that we have. Period. So that helps a lot in the way of selling. Ideally, I would love to see that. I know that in a very little future, all the EHRs are going to develop their own. So there's going to be like a very big consolidation in that space, unfortunately, but it's true. So where we are preparing, how we're preparing our companies to be able to orchestrate those LLMs coming from external vendors through our hardware

[00:20:59] to be able to prepare ourselves for that moment of consolidation. Yeah, a bit of a different kind of market than Jonathan's. Because we don't sell to hospital systems, we basically always went directly to the HCPs just from the perspective that it seemed so hard to get into hospital systems, get pilots, get traction and so on, that by the time that we'll be done with one pilot, within one year, we had 80% of physicians in one country using us. By that time,

[00:21:29] we'd only be starting a pilot in one hospital, in one corner of our country, right? But again, we're in different parts of the value chain in regards to that. So we're continuing down that path, basically betting on not being tied to any hospital system for the value that we provide is going to enable us to be faster, more user-focused, less burdened by tremendous changes in the model qualities that are coming, and just being able to iterate in the product way faster than somebody who's working directly within a hospital system.

[00:21:58] Yeah, absolutely. The sales cycles are one to two years long when it comes to hospital. If we talk about EHR or IT systems, the implementation cycle can then take another year or two. So if we look from the broader perspective of what's happening in Europe, we're building the European health data space. And Amanda, you as a data modeling expert,

[00:22:26] how do you see the two trends? On the one hand, EHDs is a huge endeavor. It's about data standards. It's about a common language and exchange across Europe. And then at the same time, we have this rapid speed of AI development. Do you think that is going to have any impact on standard development? We are saying that with AI, it's easier to structure unstructured data.

[00:22:52] But is that just something that you don't see that will be very impactful or will it be a complete game changer? For me, I find it very hard to make an estimate, to be honest. So I'm wondering if the connector is missing. So we have on one hand, we have the AI on the data entry side at the moment mostly. I'm not sure how much is being used in actually structuring the data and making it into a reusable format. So, and then the connection to the EHDs

[00:23:22] would be then the outbound stream. So I'm wondering what's happening in the middle. And I think this is where hospitals need to be, yeah, putting their effort in so that this also becomes not like everyone's doing their own effort, but maybe there can be a schema developed where everyone is doing the same thing so that you can share the effort and the work and the costs in the end. Before we continue, I would just like to invite the audience again that if you scan the QR code,

[00:23:51] you can add a question if you have one for our speakers. And with that, I will move to the topic of vibe coding. So today, you can download an app, Codex, something else. You can say what kind of app you want to have, plain English, bam, there's the app. Blush, how do you see the impact of vibe coding? Is it just going to create a bigger mess?

[00:24:19] Or is it going to solve the issue of missing healthcare solutions? I think it's going to do both. I think it's going to... So I always say that doctors were the original vibe coders. If you go into any hospital system, let's say with a few hundred clinicians, and this is what a company in Slovenia did, basically they mapped out all of the internal tools that doctors have built, and it was around 2,000 Excel spreadsheets

[00:24:48] with various calculations and various formulas. The lack of vibe coding solutions didn't stop them before. It's only going to make them more aggressive now that basically you can build a full UI and full interface and all of that. So I think on the one hand, it's going to really lead to an explosion of even more tools the doctors built for themselves. And on the second hand, it's going to create an even bigger headache of how none of this is going to work together because the doctor is going and does this talk to any other tool? Not really.

[00:25:18] But I honestly don't see an easy solution to that part because you will not stop this tool building. You couldn't stop it when it was Excel. You're not going to stop it now that it's these development tools, AI development tools. That doesn't sound very promising. Jonathan, are you already thinking about this? Are you going to have to create an integration engine for the solutions that Lea has built to be connected to the data and the platform that you're also building? If the regulators allow me, but they won't. Why?

[00:25:48] Why wouldn't they allow you to do that? Because it's part of the whole framework of regulation. You can't be so variable and just bring things within the framework of a medical device. So that puts a lot of constraints to companies. Otherwise, yeah, I would love to see the doctors just Vive coding whatever they want in my platform. I love Vive coding. I've been using it a lot. I actually, I basically coded the new website of Keiko completely unlovable. And then I exported to GitHub and have to do a lot of stuff

[00:26:18] to make it productive because it's basically a very nice MVP. But when you want to convert it into a product, it becomes super complicated. So what I think is that I would compare Vive coding with personal printers. Like, nobody built a book at home using a personal printer. But everyone has the capacity to print. And I think it's the same as Vive coding. You can all build apps that look nice

[00:26:47] and maybe some websites that would look like okay, right? Looks very okay. Way better than a WordPress website. But then if you want to convert that into a product with outstanding UX, responsiveness, that actually makes sense, that the code makes sense, and you can integrate with a back-end database, and you can export clients into a database and do payments and multi-payment support, and then it starts to become an actual product, it's almost impossible to build right now

[00:27:17] with Vive coding. Maybe at some point that would be the possibility. But right now, it isn't. I still feel that it's going to be more like a microwavable dish. It is almost there. Regulation is a very important part for healthcare. It's a risk-averse industry. Yet, at the same time, as we discussed before, when tools are there, we use them. It's not like clinicians are avoiding LLMs. But, Amanda,

[00:27:47] because of regulation, what do you think this is going to mean for the way medicine is delivered? Are we going to be in these two parallel worlds? One is the regulated one that's really progressing super slowly because you have to do MDR, AI, you AI act, then now there's EHDS, et cetera, et cetera. And then, the consumer tools that can also be used for healthcare. So, how do you see

[00:28:17] these two speeds of development? So, I think, because we're not talking about products that manage your personal grocery list, but it's medical products, right? So, it's life and death decisions. So, I'm afraid, also, when looking back, like, without AI and live coding and everything, how little innovation really happened in the last 10 years in the space of medical software. And I'm afraid that the regulation part will hinder any kind of development, being it traditional, low-coded,

[00:28:46] or AI-driven. So, I think, if we really would like to support, help to provide us with new solutions, maybe this will be a point where something needs to be done. Because I don't think the regulations will change. And then, what happens on the side? It's like the Wild West, right? Even if it's Excel, or if it's local tools, or if it's live-coded apps, I think it will develop in parallel unless something will be done on the regulation level to make these two worlds collide. But what about EHDS?

[00:29:16] One of the aims of EHDS is that after 2031, we're going to have this large pool of data on the European level for secondary use for the development of new solutions. And EHDS is a regulation, so can't regulation also be a catalyst for positive change? Do you have any hopes about that? Jonathan, you're smiling. I'm laughing. Why? Why? Why? You don't believe that EHDS is going to happen? No, I don't believe that regulation is a catalyst for innovation. Okay.

[00:29:45] I am a physician and I have been living in constant constraint by regulation. I studied in Latin America and I moved to one country first and I had to regulate myself there and then I moved to the Netherlands and I had to certify myself again. And I've always been living in this regulation nightmare, you would say. The problem with regulation is that they're not really addressing the problems of today. They've been built, the majority of norms have been built 10 years ago and we still have to live

[00:30:15] with the advancement of technologies. Things that are no longer a risk are a risk under regulatory frameworks and that's a mistake. For instance, when we were trying to regulate this device, we were using all anti-allergic plastics and medically certified products. However, we still needed to do a cytotoxicity test for a product that demonstrates no cytotoxicity. Why? Because somebody needs to do this and that is thousands of euros

[00:30:44] spent by a company that could have been poured into innovation, into growth, into teamwork, enhancement and so on just because the regulation is still contemplating materials that are not certified for human exposure. But given that you're on the US market, Blas and Amanda need, we're all in the European context more than maybe you because you're targeting the US market. US market and the regulation is basically being dismantled

[00:31:14] at the moment. FDA is using AI big time. They're changing the rules. I think in Utah now they've actually implemented that AI agents or AI can do repeat prescriptions for patients so they're like on a different speed when it comes to regulation compared to Europe. Isn't it easier there? Faster there? What are your observations trying to get into this market? 100% is easier. There's a lot more competition because of that.

[00:31:43] Everyone wants to be in the US because it's a free market and compared to Europe so 100% it's more difficult. But I would say if you look in the regulatory lens I could name you a country that is really interesting which is Chile in South America. It's a country that has a framework for regulatory in which the only thing you need is to have an FDA clearance and then you can enter. They don't regulate. So you can just basically import and immediately take something out of port and send it to a patient's home put it into work

[00:32:13] in a hospital. They don't regulate they just say one regulatory framework is enough for us if somebody has vetted for it that's good for us. They have the lowest mortality rate in Latin America and all the other countries are heavily regulated. So is regulation and that's my question is it really addressing security for patients or addressing better outcomes or addressing any type of ROI and then is it actually hindering innovation? I think it is. So in my opinion I would just

[00:32:42] and I think that the UK announced it last year that they would accept FDA regulation for MHRA UKCA applications. I think that is the way to go. We should try to consolidate frameworks that would actually allow us to scale technology way better. It's not going to happen in Europe. Easier said yeah easier said than done exactly because in Europe even with EU regulation you've got different interpretations and country levels so you need to adapt based on markets but this is where

[00:33:12] authentic AI comes in right because when we talk about regulation the challenge is that there's not enough assessors so you create a waiting time and it takes forever to get through MDR because there's so many applications that a single person needs to view blush with AI agents is it utopian to think that things are gonna get easier and faster because agentic AI

[00:33:41] could help us with some of the assessments and basically speed up regulatory processes I think agentic AI is going to stumble into the same issues that everything stumbles in in regards to regulation is that before this gets approved before this gets integrated and done in the way that regulation would allow this to happen we're going to be five years ahead of where we are today and the world is going to look completely different right and I think companies

[00:34:11] I think the safer bet not saying it's the healthier bet for healthcare systems overall but the safer bet would be that we will come into a reality world and LLM can read your screen can read an epic EHR system can read whatever other system you're using compile that data into your own random solution that you've cobbled up which exactly fits what you want to do as a physician before anything on the regulatory side catches up

[00:34:40] to what's happening in the world again not saying this is ideal but the more the capabilities of AI overall and then later on agentic systems and AI go ahead the harder it's going to be for hospital systems to keep up for regulators to keep up because if you have a wonderful tool to use and there's almost no way for somebody to stop you from using it you're going to use it there's no written ways above that's an interesting thought because I know

[00:35:10] that for example for the EU tenders and the funding that you can apply for I'm pretty sure that applications are supported by AI but assessors as far as my information goes aren't allowed to use LLMs to just scan through the applications so definitely an asymmetry speed making things potentially even slower when you've got more applications and still the same way of doing things and even today

[00:35:39] it's 24 months to get certified as a medical device imagine if one side gets five times higher in terms of applications but this side doesn't use any tools to automate that it's got three four years before you get to a market with a medical device three years ago chatGBT fairly came out four years ago it didn't exist what is going to be in three to four years from today so let's talk about the future the speed of change in the last two three years is

[00:36:08] anxiety inducing to put it mildly it's so hard to follow all the development and take time to use and get used to the tools that are available for us so how are you looking at the developments in the next five years let's say or just the future what are you excited about is there anything that you're worried about in the areas that you are working on maybe Jonathan let's start with you Amanda will finish with you I'm loving

[00:36:38] computational chips just loving to be able to run AI on device my vision of the future is living devices and devices that can basically talk to you and interact with you and have a specific oh there's a picture devices that have a specific intention that have a specific purpose in your life not generalistic physical AI devices because those haven't really succeeded so far we started with the humane pin

[00:37:07] they didn't succeed and the Rabbit R1 didn't succeed that much but I feel the devices that have intention and they are living devices they will succeed so I'm excited about computing chips and the possibility of running AI inside of them another thing that I'm excited about is when we're going to be arriving to the first layout of artificial general intelligence at some point that would be really cool to see what an intelligence thinks about my own

[00:37:37] sensors and how can I enhance them and amplify them for specific purposes that I haven't even thought about with my own technology I would love to be able to have a chat with that intelligence so those are the things that are really making me very excited I am burned out completely on GPT topics too many LinkedIn is plagued everyone is developing the same thing and there's a lot of different launches and hundreds of millions of dollars just poured into companies so yeah I'm trying to stay

[00:38:07] away from those kinds of news and looking more into what's happening in the next 5 to 10 years especially in my field computing chips Lars similar to Jonathan stake out teams the really interesting part especially for healthcare because I do think regulators are going to eventually catch up and probably do put a lot of limitations on what's happening right now but something that could avoid a lot of those limitations is on device AI basically local AI that doesn't go to a server somewhere but still offers you

[00:38:36] a good level of intelligence to solve problems for your patients and in connection to that is similarly just a lot of what we've built

[00:38:45] in technology over the past 20-30 years has been taking us away from looking into each other's eyes and talking to one another and so on especially in the healthcare setting right I remember when we started out there were one of the first smartphone apps for drug information and doctors had this really big worry because they don't want to take out their phone in front of the patient to look up a piece of information about the therapy that they're receiving because it feels offensive. That's completely changed now in the last 10-14 years

[00:39:13] that's no longer unnatural but it's still unnatural to the patient because the patient still perceives it as are you even paying attention to the conversation we're having. So basically technology moving back into the background where it's somehow ambient and still providing a lot of value which is what ambient scribes are doing with a lot of this technology is doing.

[00:39:32] I think that's probably the most hopeful vision of the future that AI can bring also instead of staring down and like this on laptop screens you move your eyes back up on people in the world a bit more. I love the optimism. Amanda?

[00:39:46] So what I would be really excited about is that AI could be a really good tool for patients to understand that disease is better because I feel like we often have a mismatch between the let's say the professional speech of doctors and then the language that most patients can actually understand.

[00:40:02] And I feel like AI could be a really good translator between that like a secret doctor that you could ask the questions that you might be actually ashamed to ask a real person that you haven't understood about your disease or that you are worried about your disease to close the gap on let's say patient engagement and also education in a reliable and safe way. This would be something that I would find really cool.

[00:40:24] What I'm worried about with the coming up of AI in medicine is that we humans we are quite lazy right and we do everything to outsource our tasks our jobs and so on and so forth and what I'm a bit afraid of is that we will lose know-how and we will lose experience and doctors. The same way that nowadays almost nobody can navigate with a paper map anymore but you just take out your phone and you look at Google Maps.

[00:40:51] I'm a bit afraid that we will have the same kind of development in medicine where we will lose the experience and the knowledge because we outsource it. That's like how our brain works and I don't think anyone is safe from that. Yeah. Yeah. We will see what the future brings. For today I would like to thank the three of you for joining me for this discussion. Thank you to the audience that stayed with us. That's it from us and I can always say follow Faces of Digital Health for more great content. Thank you.

[00:41:25] You've been listening to Faces of Digital Health, a proud member of the Health Podcast Network. If you enjoyed the show, do leave a rating or a review wherever you get your podcast, subscribe to the show or follow us on LinkedIn. Additionally, check out our newsletter. You can find it at fodh.substack.com. That's fodh.substack.com. Stay tuned.