AIAW Podcast
AIAW Podcast
E146 - AI and the Future of Healthcare - Carl Fredrik Swic and Kenneth Ilvall
In this episode, we sit down with Carl-Fredrik Swic, Founder & CEO of Mediqtech, and Kenneth Ilvall, Chief Medical Officer, to explore how AI is transforming healthcare from the inside out. They share Mediqtech’s mission to reduce administrative burdens with tools like Luna, which cuts down paperwork for doctors, and Maly, designed to detect journal discrepancies and improve medical record accuracy. We’ll dive into the biggest challenges facing AI adoption in healthcare—whether it’s technology, regulations, or resistance from professionals—and how Mediqtech is overcoming these hurdles. Plus, hear their take on the future of AI in diagnostics and predictive care, and how Mediqtech plans to scale internationally in the complex healthcare landscape. Tune in to Episode 146 for an inside look at the future of AI-driven healthcare!
Follow us on youtube: https://www.youtube.com/@aiawpodcast
And we've seen a lot of articles about the nomads that go around Europe and the world and just working on a constant basis and freelancing and stuff like that, and I was intrigued by that and so I can. Can I do this? Can I make a living from actually working remote? And I'm on the way I'm not really there yet, so I'm just taming to actually get that in place.
Anders Arpteg:What I think about working from home. All the time, though, my girlfriend is working from home as a daycare mom, and she is know is always at the house yeah and that becomes a bit tedious as well if you don't have any else to go yeah, but I I will.
Kenneth Ilvall:I'm accustomed to living at home and being at home and working from home as well from during the pandemic.
Henrik Göthberg:I love that that was my follow-on question. Was the pandemic a place where this thought started to grow from a small thought to something that you took action on?
Kenneth Ilvall:Yeah, actually it was. And of course there is a lot of discussion that, okay, can you actually work efficiently from home or do you just start a dishwasher, go out with the dog, rake the garden or whatever? But I'm really into work when I'm working, and then maybe of course I put on a dishwasher now and then, but I want to do the efficient work when I work.
Anders Arpteg:So you have no problem with self-discipline when working from home.
Kenneth Ilvall:No, because my work never ends when working from home. No, because my work never ends. I can still work when I'm sitting in the sofa, or I can um well, being out on the on a walk and still work, because so so diversified what I can do with my work because not every person can do that.
Anders Arpteg:You know, some people get distracted and turn on the tv show or whatnot.
Carl-Fredrik Swic:I'm super impressed, impressed by that, I need to be at the office.
Henrik Göthberg:Yeah, you do yeah. Yeah, I mean, this is some. I've heard this conversation as well. I want to the opposite, like I want to compartmentalize life here and work here. But on a deeper note, because I've been really thinking and dreaming about these topics, I have a. I lived in Australia for a couple of years and I love to surf.
Henrik Göthberg:We don't really surf here. So this dream of maybe I should have three houses and three offices and you know, and and just move with the. So that whole thing has led me down to a path where I really want to ask you now, have you kind of starting picking apart your process or your systems, or how can I make this work? I need to have a very good connection, I need to have the, because the technology is all there, but it's also to make it work. Have you thought about the ecosystem for this to be efficient?
Kenneth Ilvall:Yeah, I actually have the house that we are going to move into. It's an old house and actually a big hallway and there's a spot there that I can. Okay, here's my home office. But I can't use that home office if there's my little seven-year-old's home. So, okay, first an alternate office. So I have like a smaller guest house and okay, here's how I can have my other office yeah, exactly, and also some of the other living bedrooms that actually can have like a.
Henrik Göthberg:Okay, here is, I can do some some real work and and are you limited or are you needing of any technology in terms of really good connection? Yeah, because, yeah, video conference, yeah, video conferencing, yeah video conferencing.
Kenneth Ilvall:I need to have that, so, like a good Wi-Fi connection, it's fiber in the house. It's not that buggy of a problem now in Sweden.
Henrik Göthberg:Yeah, because you can take that whole thing to the extreme. How can we make our video conferencing be less stale and how can we get it more engaging? How can we go all the way into the technologies where we have? We sit in the room, you know. So there's a lot of technology and you know the whole metaverse conversation. You can go down this rabbit hole if you want to, but I don't know. For some people, or for me, I think it's to find something that really works, but then also to find where are the limitations when we're not connecting as humans, and how can we make those limitations smaller and smaller. That has been an intriguing thought process for me. Have you thought about this?
Kenneth Ilvall:Well, in a way, yeah, Because I'm always wanting the technology to work for me, not against me, and that's one way to do it so I can actually use the technology to actually make me work from home and have the internet connection actually working and the video conferencing to be. Well, you don't have maybe you don't have to have a background that's blurry and stuff like that. It's okay to have a real calm background that, well, it's neutral, and also see that you have a good. Well, the audio should be good and you should be comfortable in the situation. I see a lot of patients via video so that I can actually see them from all over Sweden or wherever they are. Of course, I have some patients that, okay, they went on a business trip to the US and okay, so the time difference, okay, we worked that out, and then we had a video conference. Do you know?
Kenneth Ilvall:the place called Misterhult. Misterhult, yeah.
Anders Arpteg:Yeah, okay, very close where I lived. Yeah, iult, mr Hult, yeah, yeah, okay, very close where I lived. Yeah, I did, but anyways, it sounds like a big life transition in some way to Västervik then for you. Yeah, yeah I think that sounds like exciting times. I think.
Carl-Fredrik Swic:Hopefully yeah, I didn't know you had this whole holistic side to you.
Kenneth Ilvall:Oh, you should be surprised.
Henrik Göthberg:Best of luck on that kind of transition and move towards the what we call the no, the metaverse. Living anywhere, countryside, lifestyle, working from home, the next generation of living and working. I think this is something that we will see more and more of, and finding this not only one way.
Anders Arpteg:And perhaps AI can enable this for more people going forward as well. Yeah, I can imagine. Anyway, with that, I'd love to welcome you both here First, carl Fredrik Zwick, is that the right way to?
Kenneth Ilvall:pronounce it yeah, perfectly done?
Anders Arpteg:Founder and CEO of a company called MedicTech. Is that right? Yes, yes. Awesome and also very welcome here, Kenneth Ilval.
Kenneth Ilvall:Close enough. Ilval. That's a terrible way to pronounce it because it's so hard, and even worse in writing.
Anders Arpteg:But you're also the chief medical officer, then yeah, and as you may then guess, we're here to speak about how AI potentially can revolutionize healthcare in different ways, and you have a startup here which have done some really cool stuff, and it's also a field that I know you, henrik, have a lot of passion too, and actually I've done a lot of medical projects as well. What was it project as well? What was it we did? Well, we heard on the news morning news recently about how AI can be used to diagnose breast cancer, for example, and I worked with a brain tumor case a couple of years back.
Henrik Göthberg:We actually did one thing where we tried to predict if they over-prescribe antibiotics in the dental care by looking at your hands at one point it's kind of kind of kind of fun and um I have had a ongoing conversation with my father, like you do for maybe 10, 15 years, on how to innovate the new state and ai um in the hardcore, in the hardcore profession, as a what do you call it? A children's surgeon, pediatrician.
Kenneth Ilvall:Pediatrician.
Henrik Göthberg:Yeah, and he was then at some point heading up the pediatrician surgery in Drottning Silvia. You know 200 staff, you know, and then dealing with the bosses and what that means, and dealing with the real operations and how to bring data and AI into this context, you know. So it's an ongoing conversation on the madness and the difficulties and what we should be doing and why is it not happening.
Anders Arpteg:Yeah, I mean AI can really revolutionize, I think, the healthcare. So I'm looking forward to hearing more about your approach to this and what you're doing there. But before that, perhaps, carl Fredrikssik, if you were to describe yourself, what's your background, your passion, and how did you get started with Meriktech?
Carl-Fredrik Swic:Oh yeah. Well, my initial background is economics. I studied international sales and marketing and international international management and was on a road to just work more with that. But from out of the blue during the summer a really good recruiter called me and pitched going into recruitment for going into recruitment for medical professionals, especially doctors. Medical doctors, yeah, medical doctors, the medical field in general, but specifically recruiting doctors and I got stuck. Sounds so bad but I was in that field for over six years.
Anders Arpteg:Recruiting for medical doctors basically.
Carl-Fredrik Swic:Yeah, recruiting If you have a hard time finding a doctor to your clinic or hospital. There was the part where I came in and tried to find the right person for you and in the process we talked to so many doctors and the uh, the bosses of doctors and like what can uh, what can improve, what is the reason people leave work and take work and what all these kind of things. So, uh, for six years I was basically the what do you call it? A person you empty out your work problems to A therapist.
Henrik Göthberg:You were the therapist, I know that yeah.
Carl-Fredrik Swic:So, and people who are unhappy about their work love to talk about it. So yeah, about their work, love to talk about it. All the problems they brought up during this time led to the start of Meditech with our colleague Jamal, who was in the same company as me, we started talking about all these things we hear. How can we take these problems and solve them? How can we take these problems and solve them with the race of the beginning of the AI bubble, or the AI boom rather?
Anders Arpteg:Cool, and why did you get started with Meditech? How was that? Specifically Because it's not really in recruitment that Meditech is working.
Carl-Fredrik Swic:No, it was just all the problems we heard, like all the problems regarding the administrative work that doctors have and nurses as well Well, all medical professionals. So much of their time is spent on doing non-patient work and things that can be done by AI, and Kenneth here knows that even better than I do. But there is this reason why a lot of people leave their work and why people leave. You can see there's a big escape from healthcare. A lot of doctors want to leave their work and a lot of young doctors in special leave the field and we were wondering like, why is that? And a lot of the research to why people leave is because they are overworked and it's not really what they expected when they got into medicine from the start. It's very administratively heavy and you spend more time sitting on a computer typing than meeting patients and doing a difference.
Anders Arpteg:My mom actually worked as a nurse as well and I remember when, you know, the whole digitalization of journals started and she was so frustrated and you know I thought it was horrible and you know she hates anything.
Anders Arpteg:You know having to type in this kind of horrible UIs that these kind of systems have. But let's not go into Meditech right now, but it's awesome to hear that that is something that you're trying to help with and partly revolution as well. Before that, then please, kenneth, if you were, can you just describe who is Kenneth? What's your background? How did he get into the field of medicine?
Kenneth Ilvall:What's your background, how did you get into the field of medicine? Yeah, so long story short, I was out on a round the world trip when I was 20, and I realized that I was actually good at massaging. So when I came back then I-.
Anders Arpteg:Where is this going, you'll see. You'll see.
Kenneth Ilvall:Yes, exactly You'll see that was a good trip. No, actually, when I get back to Stockholm after the trip, I decided I wanted to do something about it. I educated myself to be a massage therapist and actually have a small company with that and just went up to businesses and companies and also the employees which actually the employer paid for. So that's quite lucrative and I actually I realized I could do something good with my hands and the treatment there. And after a few years I decided, okay, I can do something more, I need to do something more about this.
Kenneth Ilvall:And then I went into med school and I still wanted to help people. That's the way I was. Well, that's the drive for me. And during that time I actually realized that, okay, I'm a bit of an entrepreneur as well, so I wanted to do something else, something new, and to start businesses in some way. And when I was almost ready as a specialist in general medicine, then okay, there is an opportunity that went up and so I can actually start off Healthcare Central Vårdcentral in Swedish which was quite lucrative, and we did. We opened another one and in that time I came across the video meet the crew, livi. Yeah, they approached us, our company, and and they wanted us to be the caregiver for them to actually have the medical thing.
Anders Arpteg:And that was a company that you had started at that point.
Kenneth Ilvall:No, we started 2Care.
Henrik Göthberg:That's our healthcare 2Care was actually the ward central. The ward central, yes, and here you have a collaboration now with Cree, who is the digital app. Yeah, where you basically, can the content they need for that app has to at some point connect with the Vårdcentral knowledge, Correct?
Kenneth Ilvall:correct, and I was the recruiter for doctors and just learning how, or teaching them how to actually do a video meet and what to not do and what to do, and also do all the medical background from them in the beginning. So I actually did that for two years and then I realized, okay, we have to do something more, and so I just skipped that and just went on with my general work and I did a lot of other things as well at the same time.
Henrik Göthberg:But you got a very detailed insight of one of the first sort of digitized startup ideas that actually has kind of worked, I mean, like what Cree is doing and Dr Prutesi and these guys are doing. They found their place, they found their niche and I would argue they are, by today, quite established.
Kenneth Ilvall:Yes, I would say so.
Henrik Göthberg:It's in several countries, in several countries in different ways, but you saw that business model and the back end of how this business model needs to work. Yeah, so that's interesting, right? That's the learning that you can bring with you now.
Kenneth Ilvall:Yeah, of course, and also the and other companies that actually went from from nothing. Uh, flash to froth, just a dummy, and up to meditech. I made a check. So that's the visible problem here confusion, confusion with them. Well, I'm referring to my compass that I've been working on, referring to my company that I've been working on, medicheck, that's a digital caretaker. That have been well, their business model and the vision was to offer specialized care to patients without any referrals, and they will skip the step of having to go to a healthcare center and get a referral to a specialist and then wait for a few months, or half a year or something, to see a specialist in, maybe a gastroenterologist or a skin doctor.
Henrik Göthberg:How to make the flow to the specialist way smoother.
Kenneth Ilvall:Yeah. So the founders of that company, they wanted to have a health plaza that you will connect as a doctor and then connect as a patient and just meet. And well, it worked and I did all the background for that and all the trip that the patient have to go the pathway, the customer journey through this patient journey. Exactly so. That's what my goal was in that company.
Anders Arpteg:Well, that leads us to, I think, the big question. Then, At some point, both of you or I'm not sure, was it you, Carl?
Henrik Göthberg:Yeah, how did it get?
Anders Arpteg:started Meditech, or how did it get started?
Carl-Fredrik Swic:Yeah, well, the initial start was me and our colleague, jamal sheriff um, working together as recruiters and, as I mentioned a little bit before, started to talk about, like all the things that the doctors were telling us. These are the problems they are facing every day, the reasons why they want to take an assignment, take a new job. Yeah, so all these problems that we hear, because a big part of the job was just to connect to doctors and speaking to 30, 50 doctors a day, you will hear a lot of problems. And yeah so Jamal, more than me, was the more tech savvy. We started talking about the initial boom of AI technologies and when ChatGPT launched, how can we use the power of AI to solve these problems?
Anders Arpteg:This was in what year? 2022. Right when ChatGPT basically launched.
Henrik Göthberg:Yes, so had you seen ChatGPT when you launched or did it come? No, we didn't. We were pre-ChatGPT, pre-chatgpt. It's also interesting because immediately you need to think about okay, how do I now understand the thesis? Could you summarize down what was your if you go back and now we know it's pre-Chachipiti, but what was your-?
Henrik Göthberg:By a few months perhaps, yeah, but still you know, in terms of how you have opened up your mind to, maybe what is the art of possible here. What was your founding thesis? What was the core problem? Or thinking that, because all this stuff shows there is an arena for improvement. But then how did you articulate in your head, you and your partner, when you started, we're going to go after this big hair problem. What was the founding thesis that you, or was that more involved? I mean, it could be that case as well, I guess.
Carl-Fredrik Swic:It's a really good question, but I would say like one of the big problems that we found that doctors have is finding information, and it's such an easy problem to fix.
Carl-Fredrik Swic:A lot of doctors spend a lot of time searching for information. Especially, we have a big, a lot of Swedish healthcare is now dependent on locum doctors, locum doctors I'm not sure Hydroläkare People traveling around Sweden where there's problems to recruit To travel from. If you have been working in Stockholm your whole life, going to another region, it's not a big country, but there's a lot of differences between the regions and there's used to be a lot before, especially when we started working, like six years ago, with this way more general systems and because now there's one who has basically taken over the whole of Sweden. But working from one system is are different and there's different regulations and there's different, uh just aspects of working in the regions and finding information is is hard and you're, you're, uh, you have all these um it's. You are expected to come into the um, to the, to the workplace, and just deliver from, from like the first thing you do from the start. How can we make it easier to find all this information? So knowledge discovery in?
Carl-Fredrik Swic:some way for, sorry, knowledge discovery or information retrieval kind of purpose to start with yeah, it's a specialist, like the word for, like what's it called in English, Like the Say it in Swedish yeah, all men läkare.
Anders Arpteg:General practitioner General practitioner yeah.
Kenneth Ilvall:GPs the.
Carl-Fredrik Swic:GPs. The word itself is kind of we talked about it before. It's kind of contradictory. To be a specialist in everything is hard, it's really hard, and you can tell us all about it.
Anders Arpteg:I think that it's a contradiction in terms to say you are a specialist, general practitioner. I mean that's kind of weird.
Henrik Göthberg:No, you're a specialist in how to be the most efficient in the GP role. Exactly, well put, well put. So you're a specialist in how to most. And if we flip that into, I would summarize that okay, so you want to. One of the founding thesis is how can we make a smooth onboarding of GPs, for instance, when they are in a higher contract? So how can I get the time to market for them from I hire them to they actually can do their job 100%, without asking stupid questions to everybody?
Kenneth Ilvall:along the way right.
Henrik Göthberg:So the onboarding process of a GP in a new setting, for instance Hyre Doctor, what is the English word Locum, locum, doctors? I think Locum New word for me yeah, like contract. But for me it's like the whole contracting around doctors.
Carl-Fredrik Swic:A different word for it. Yeah, consultant doctors. Yeah, and it's typical.
Henrik Göthberg:We have a lack of enough skilled resources in one region.
Kenneth Ilvall:Yeah.
Henrik Göthberg:And we have a really hard time getting permanent employees, so we need to solve it in some ways.
Anders Arpteg:Yeah, cool. Do you have like a mission or a vision for Meditech? What is?
Carl-Fredrik Swic:outspoken. Well, you are the wordsmith. Kenneth is incredible with summarizing.
Anders Arpteg:I would have a vision for you. After this, we should ask Chateau Petit otherwise.
Henrik Göthberg:We did, but it wasn't that good. How do you? How did you? Okay? So I'm just going to ask, Kenneth, how do you got intrigued and how did you? How did you guys meet? I was lured. Yeah, how were you lured? Tell me, tell me that anecdote. We're salespeople, we lure.
Kenneth Ilvall:Yeah, you lure. No, exactly I. I've been actually meeting up with Jamal in some other aspect and he said okay, you were like a front-end doctor and you really leaned forward to the new technology. How do you like to have lunch with me and Carl? Yeah, okay, sure, I'm game. So we had dinner. Was it dinner or lunch?
Carl-Fredrik Swic:Whatever, yeah, dinner. So we had dinner, it was dinner or lunch, whatever, yeah, dinner, yeah.
Kenneth Ilvall:So that was nice, in short. And then they discussed the problem that they saw in their former roles and I was intrigued by the idea of actually getting better information and better teaching possibilities for the doctors to actually, well, some of them, the locum doctors they were maybe right off the medical school and they don't have that much experience, or they lacked experience from that region that they were working in or that field, maybe lacked experience from that region that they were working in or that field maybe. So they needed some guidance and some help with tutoring. And I was intrigued by that because I worked as a studirektor that's some strange word for that in English it's not, well, something completely different, but that's actually someone who actually teaches not teaches, but helps the intern to actually get through the residency.
Kenneth Ilvall:The principal yeah, not the principal really, but anyway, associate principal there's a regional something, regional director, it's in English, regional director, anyway. So I have the knowledge of the problems that a doctor who's not really comfortable in their role. Yet maybe they need some more knowledge in some fields and they are looking, those guys are looking at a bridge to help those guys out there, and so I was intrigued by that we wanted someone to take all our crazy ideas and kind of be our, our filter.
Carl-Fredrik Swic:Yeah, so to say someone. Uh, we started talking about like a lot of different solutions. What are the most effective and what can we realize? And where's the technology and where is the? What is the actual value for the doctors? Um, so the pain points as well yeah, and the first, the first, uh, the mbo of meditech was okay. Okay, all these doctors out there, how can we test their knowledge Like they do for pilots, for example, in order to be a certified pilot, you need to constantly renew your pilot's license, and doctors don't.
Anders Arpteg:Was that the first use case, or what was the first product?
Carl-Fredrik Swic:you tried to develop that was the first use case. So was that the first use case or what was the first product you tried to develop? That was the first use case. How can we see if the doctor is at the knowledge base, where they should be, and not to get them in a gutsy moment, but more than to focus where can we? Because there was a lot of complaints about there's not enough time for foot building to Continuous learning. Yeah, continuous learning and further education. So how can we more effectively see, okay, where are the lacks in the knowledge and how can we pair them with the right courses in order to more effectively make the regions like, put their resources into the doctors Interesting?
Anders Arpteg:So more like lifetime learning and having them upskill themselves.
Kenneth Ilvall:Yeah, it's a lifetime learning. That's what we do.
Henrik Göthberg:But it's a lifetime learning, but it's also a little bit like okay, in order to put the plan together. When we onboard this doctor, we want to know his strengths and weaknesses and where is the most bang for buck If we give it the wish. Where should you go and read? Where should you go and study in order to excel in this region, in this context?
Kenneth Ilvall:Yeah, and I was also discussing about how the doctors then has been out there in the suburb or the remote areas in Sweden and how can the manager of that healthcare central really know that they have done their shit? Yeah, has they really done what they should do? What do they know, yeah, and what they know in advance and what have they done afterwards? So if there's anything that they have missed, that was also something that we discussed.
Anders Arpteg:So, like a support for them, Did you sell this somehow, or how did they? You know how did you go about.
Carl-Fredrik Swic:This was very early in the thought process and we discovered that doctors doesn't like to be evaluated.
Anders Arpteg:I can guess. So yes, I would take offense as well, yeah.
Carl-Fredrik Swic:It's interesting because, like, for example, you want your pilot to be up to date about the plane they're flying and all the rules and regulations, and we thought that, well, we were young and naive and we were thought that, yeah, so does doctors. They want to know where they lack. Not in Sweden. No, not in Sweden, doctors don't want to.
Henrik Göthberg:Maybe the regulation came first and then the culture changed.
Kenneth Ilvall:There's a regulation in Norway, for instance, that all the GPs have to renew their more or less renew their license every fifth year.
Carl-Fredrik Swic:Which is similar. Yeah, but in England as well.
Kenneth Ilvall:Yeah, a lot of countries in Holland as well, in the Netherlands, and then I think that sort of cultural blockage will go away. Yeah, but there is no, not sufficient political muscles or gut to get it.
Anders Arpteg:But then you move forward in some way, right, and you got started with some other use cases. Yeah Right, I heard something about Luna or Mali or which one, any of those that you'd like to go a bit more in depth with?
Carl-Fredrik Swic:I'm going to start a bit because when we started to figure out and because there were so many ideas, there's so many use cases, and when we got the help of Kenneth, like realizing and pointing out like what can be actual value, because you're not only a doctor, you have been running your own clinics and studio director and all these kind of things. You and running your own company, you have a wide set of skills that we saw as very useful for us and therefore the approach what's the word I'm looking for? Is it Luna?
Anders Arpteg:you're speaking about.
Carl-Fredrik Swic:Yeah, we wanted to see what products can actually be used, not just our young and naive thoughts about how it should be. What can actually be done. Luna was one of the first first products we came up with, and so what is luna? Yeah, it's information search, basically. So how can you more easily find the information you're looking for um in the region?
Anders Arpteg:lm rag solution in some way, or how would you describe it?
Carl-Fredrik Swic:when we when chat gT came, we explained it as ChatGPT for healthcare professionals, basically. So if you're a GP, how can you be updated in all these fields? And how can you, as a young doctor, as on the beginning of your journey, how can you be a less burden to your colleagues To not interrupt workflows, how can you be more efficient by yourself? And how can you without disturbing the other doctors? Because that's what one of the things we we saw doctors, they need to go and ask the other doctors to even more specialized people and the the there are three different stages of being a doctor. I guess at that time when you first first go to med school and then you just do your studies to become Internship, and then the residency yeah, and the residency and then you're a specialist in the training to become a specialist, and during this whole process, and even seasoned doctors need to go and ask other people hey, what is this? In order to more effectively to make the workflow more efficient.
Anders Arpteg:So what kind of information sources are you using to Everyone? Yeah, that sounds a lot.
Kenneth Ilvall:No, of course that's not. Decompose, it Decompose it yeah. So in every region you have something called beslutsstöd in Swedish, decision support, decision support, that's the direct translation, and I think that's what it's actually called.
Anders Arpteg:That's proper support.
Kenneth Ilvall:Yeah, and you know you need to know what the actual decision support is for every different diagnosis or every different symptoms, but you don't, so then you need to have someone or doctor's companion in the exclamation points that you actually could rely on that this is the most efficient and best information that I have at hand at this time and here maybe correct me if I'm wrong, but it used to be nuanced.
Henrik Göthberg:What is that decision support and what is the preferred best practice procedure for a certain? Yeah, it's different from region to region. It is Depending on where a certain hospital has gone and what they have trained on, and stuff like that. Yeah, so that's why this decision support needs to be contextualized with a regional context or the law of the situation of that region, exactly. And now we have an information problem.
Kenneth Ilvall:Yes, of course, and also there is something that's called natural guidance of a lot of things, and you have to merge those two together and then, okay, then I have the oh, this is the best practice for this patient in this region at this time, at this time, at this time and this changes, of course. So you can't really drag along with you a book. A Doctor's Companion was and is a little thick book, and then you have a smaller version that's this leaf thin pages With the changes every week or every month?
Kenneth Ilvall:Yeah, it should actually. That book actually changes every year, but it should be changed every month maybe, and this is hard to keep track on that. So that's why we created Luna, and so it will scrape down every other aspect that we have out there in the regions and you can pinpoint it through each region.
Henrik Göthberg:I fully see the point, because it's one thing not only to make these ones, but it's to do a framework or a template or the platform around it, so you then can fill it with your local, regional best practices.
Kenneth Ilvall:Yeah, and then it's have to learn from it as well.
Goran Cvetanovski:Yeah.
Anders Arpteg:So you fetch, scrape data from the web in some way, or where do you find information?
Carl-Fredrik Swic:Well, the regions have their sources. They want all the doctors to use. They practice in that region and in Sweden.
Henrik Göthberg:So basically that so when we say scraping, we're literally going to every single different department that has their policy and their practice and they're not always one-to-one connecting to each other, so they make sense together. One can be an anti-pattern for the other ones, but in practical terms, you will have practices around anesthesia, around the surgeon, blah, blah, blah, blah, blah, and then you will have practices around administration. You know administration, you know so. Then you need to go to another department to find their policies and stuff like that, and all that becomes a whole. That's the soup that the doctor needs to navigate from administrative best practices to treatment best practices in the context of a certain hospital, technology, etc. Yeah, so that's in a nutshell, in a nutshell, right. So this, then, is what scraping is all about to go to all these different sources and then mix it together and try to answer the question at hand.
Anders Arpteg:So do you train a model on that. Then Do you scrape the information, do you retrain or re-index in some way the information.
Kenneth Ilvall:It should be. Yeah, the work is ongoing and we need to make it even more efficient and more accurate, and presentation is everything. How do you present the information that you have gathered? Is there a decision support or a decision suggestion? So that's a big difference actually.
Anders Arpteg:Is it like a Chattipati interface, or how does it look right now?
Carl-Fredrik Swic:Yeah, basically that's how we wanted it to look, because people started using Chattipati to well first, people started using chat gbt to um well, first people started using doctors using google and uh, taking long like took them so long to look for symptoms and guidelines in all these different systems. So we wanted to take all these different systems and just combine them into one. Uh, for, because there was a lot of. We heard that a lot of patients were very upset about the google sitting, the doctor sitting there googling your symptoms uh, so having a system that looks more good yeah no, it's, it's a bit worrisome and uh, how can it look more professional and um, make this whole person so give it a concrete example.
Anders Arpteg:Is it when someone makes a diagnosis in some way and they need to basically confirm or give some kind of support for why they did this, or can you give some kind of concrete?
Kenneth Ilvall:example, maybe we have a patient that's tired and have a fatigue problem, and OK, so then we do some blood tests and have 10 to 15 markers that we need to check on and then the blood tests come back Okay. And then we see the blood tests and say, okay, b12 is low. Okay, so what is the decision making for this region of B12 deficiency? What's the treatment?
Henrik Göthberg:plan.
Kenneth Ilvall:The treatment plan. What's the next step? Do I need to have a gastroenterologist? Well, endoscopy, or do I need to just treat them at first? Or is there any other?
Henrik Göthberg:So there are two major steps. One step is to help establish diagnosis with the highest accuracy.
Anders Arpteg:Or is it mainly afterwards, or is it even before you do the diagnosis?
Kenneth Ilvall:Before, actually it's more before yeah well in well you. Maybe you don't have the diagnosis just to say they have to be 12 deficiency, but why?
Henrik Göthberg:yeah, this is what I mean. So so you have, you have someone is coming in and having a pain or something. Then you want to put the diagnosis in where you want to go to the root why. So this is the process what? What are our best practices on how you establish the why for the sickness and then, when you, when you think that we are now happy that this why is strong enough, what is our region's view on treatment plan from here on? What do we do first? Do we do more tests? Do we get them on these drugs? Do we refer them to another specialist? You know we don't have any money for specialists. We give them drugs at once, exactly.
Kenneth Ilvall:In some regions right, yeah, correct.
Henrik Göthberg:Which in other regions is like no, no, no, you go to the specialist now.
Kenneth Ilvall:Yeah.
Henrik Göthberg:That's the next step.
Kenneth Ilvall:This is the next step, yeah.
Henrik Göthberg:So you have the whole process of taking and setting a diagnosis is from original symptom to diagnosis. What is this best practice? And then you have the treatment plan according to the policies, the money and experts available in that region.
Carl-Fredrik Swic:Yeah, also. Icd codes, medicines, all these things that it's impossible for one person to have in their head. You don't need to have it in your head, you can have it in Luna.
Kenneth Ilvall:At your hands or your tongue.
Henrik Göthberg:But let me clarify one thing because, as I understand it, it's not that simple that you can use to load this up once and then go with it. So this has to be a very a platform where you dynamically can input the information that is the correct information to look at from this region and they have, and also adjust what treatment plans is the correct one. So you need to be able to interact with what this information. It's not as simple as just uploading everything once and then you're happy.
Kenneth Ilvall:It's very dynamic. It's dynamic work, and you we talked about the, the engineers, about this, and okay, how long did it take to scrape everything? Maybe an hour. Okay, so the algorithm was there, so okay, so you can do it actually every month. Yeah, every week, if you want to, so we can have the exact information, because then you get to the product view on this.
Henrik Göthberg:Because ultimately then, in order to install Luna, you need to install the practices of filling the data, in updating the data on the one hand side, and what is the UX for this and who should be doing that, versus the UX for the doctor that makes it usable and adoptable in his workflow?
Kenneth Ilvall:Yeah, well, yeah, are we getting it?
Henrik Göthberg:I think you're getting it. Yeah, it's a super good idea. Yeah, thank you. We think so it's an awesome idea.
Anders Arpteg:Is it used by anyone today, or is it more in pilot stage, or is it what's the?
Kenneth Ilvall:current in the learning stage so we have the, we have the template for it and the. We have trained it a bit and now it's in the working space that we want to do to put it to market in some way.
Anders Arpteg:Yeah, do you collaborate with any institutions or hospitals?
Carl-Fredrik Swic:We work with a few local private clinics that are able to test the things we want to.
Henrik Göthberg:Some sort of alpha beta client.
Kenneth Ilvall:Yes.
Henrik Göthberg:I think that would be smart right.
Kenneth Ilvall:Yeah, we have discussed it. Okay, so we can have one private driven healthcare central. We have some regional that actually have to get reimbursed from the region and also some larger hospitals and stuff like that. Can they use it as well?
Anders Arpteg:Do you foresee any challenges in putting it in production properly, so to speak?
Kenneth Ilvall:foresee any challenges in putting it in production properly, so to speak. Well, what you put in is what you get out. So if you ask a question that's not really on point, then you get a lot of answers that maybe are a bit diversive. But if you actually can make or learn how to really put the right questions, then of course the database will answer directly what you want to know.
Anders Arpteg:So you need to onboard or upskill or learn training sessions.
Kenneth Ilvall:Yeah, training sessions, that could be good, but I think you will learn efficiently how to ask the questions more properly and, depending on what you want to ask. If you like an ICD code, and okay, what's the ICD code for this? And then you have three different alternatives, so that's easy. But if there is a more complex diagnosis, then you have to maybe ask four or five questions to actually get to the point, and then okay, there's a more complex thing, but the the model will actually are capable of getting that to work. But we will see how it will in actual wildlife function.
Anders Arpteg:What's the next steps for Luka?
Henrik Göthberg:Before you go there. The challenges I think this is a good challenge. I can relate to selling this in relation to selling data and AI, for different reasons. One thing that you need to crack is that, obviously, then, for making this vision to work, you literally need to sell it to several different actors. You have several different change agents. You need to line up to get commitment with the guys having the right information that they think this is a good idea. Who is going to continuously dynamically update it and keeping it track, because they need to go 360 around the whole hospital. And then you need to convince the practitioner or the bosses of the practitioner. So you have, like, a user buyer, you have a technical buyer, you have an administrative sort of user buyer. You have a technical buyer, you have an administrative sort of. So it's so to crack the code on how to communicate and sell that and how to make it easy. I mean, like you can do something that looks great for the doctor but it's shit and creates a lot of work for everybody else.
Anders Arpteg:You see what I mean, right.
Kenneth Ilvall:So, this.
Henrik Göthberg:Am I on point here that this is a tricky?
Kenneth Ilvall:one. Yeah, that's the challenges that we see as well, Because when we went out and just look at the market and who's going to buy this? Who want this? Everybody wants it. It looks nice, but okay, have you tested it yet?
Henrik Göthberg:And who needs to be on board and do their part.
Carl-Fredrik Swic:But it's also about the value and also the perceived value of these kind of products, because there was a study in the US a while ago. I'm just reading it from a note and it's kind of what's the word, it's a kind of telling for the general perspective of doctors in what these kind of AI solutions can contribute, and it said the headline was that doctors have no use of chat GBT because the study showed that well, doctors who used chat GBT in the reasoning and diagnosis didn't perform better than the ones who didn't. But then it showed that chat tbt performed better without the doctor and I read that article.
Henrik Göthberg:There was a nuance on there that you know we have. This is how you look at bias and how you look at your old things. That actually, when you, when you, let it be all right, yeah and that's the thing.
Carl-Fredrik Swic:And the conclusion was as well that from the scientists or the researchers themselves that you need to optimize the interaction between man and technology, and that was the problem. They found that the prompting was mostly the problem not the knowledge of ChattyBT.
Anders Arpteg:I guess it's not only AI that needs to be trained, it's also the human that needs to be trained.
Carl-Fredrik Swic:But that's what I was talking about, about the perspective. Of course, if you don't use the tool right, the tool is not used to you. So you can always say, well, this is not going to help us, but well, if you're not using a hammer to hammer in the nail, if you try to screw something with it, but I get an epiphany listening to you guys now.
Henrik Göthberg:Okay, I think what you are trying to do puts a finger on a microcosmos of what is the real challenge on a macro level to get this to work with data and AI in healthcare in general. Because what you're describing now is that if we can get one plus one plus one plus one working together, that becomes 10. But if I can't line up these things, if I put one of them to zero, the product is zero. Yeah, so the core problem is like we're working with data and AI as a multiplier effect, but then you need to get the data right, the operations right, you need to line up all the ducks in a row and if one duck is not playing along, it falls apart.
Henrik Göthberg:And I think that's the challenge when we get into these topics of oh, we want to do AI in healthcare, and we take a very simplistic view of. Oh, we want to do AI in healthcare and we take a very simplistic view of oh, is the chat, gpt interface of the doctor, and we focus all on that, but we are not talking about the 360 view of what the other guys need to change for this to be so beautiful. No, so so did you see what I mean? Like so just studying your case and studying how do you sell this, how do you make the product out of that? That actually tells us a lot about why this is challenging on a macro scale. Do you agree with that? Yeah, exactly.
Henrik Göthberg:It's a microcosmos of the fundamental problem.
Kenneth Ilvall:The fundamental problem, that's the stiffness of healthcare, and that should be stiff, it should be non-fluent, it should be safe. It should be safe Iluent.
Henrik Göthberg:It should be safe.
Kenneth Ilvall:It should be safe. I think that's the general idea. So when we do changes and come up with new things, it has to be like a filter and a stop sign. Okay, is that really safe for my patient? Is that really safe for my employees to use? Is there going to be a problem? So I remember when we, in a way, we have the journal on net, so like, the journal is 1177. It was a lot of discussions before that came to play and I was on medical association and we had like a hearing and talked about this Okay, how can this work? And lots of people, okay, this is going to be a roar in the community, it's going to be the doctors will be totally swamped with questions from the patient when they read the journals.
Kenneth Ilvall:I said why? Because if you write down the journal as it was and you decided in coherence that, okay, this is what we have come to agreement on, that, okay, you have this problem and you do this, and then I do this, what's there to be alarmed about? But then I thought, okay, this is a fear of not doing the right thing at every time. Okay, so, when we go 360 on this, so when, when we do the the journal correct, and what will make the journal then so? So I was just saying that it all starts with the journal.
Kenneth Ilvall:So if you put in everything in the journal correctly, then hopefully as well that you have thought about it correctly and thought through every aspect of it. And then, when the patient reads it, okay, this is what we were talking about, so, and he got me right, and then he did this, or she did this and he got me right, and then he did this or she did this. Okay, so that's the one thing. And can we then help the doctors to put in everything right and also have a filter through it so it actually gets right every time? And also the information should then afterwards be available, yeah, available, yeah, yeah, and trustworthy, Trustworthy available in the right in the safe way.
Henrik Göthberg:Yeah, available, yeah, available, yeah, yeah, and trustworthy trustworthy, available in the right, in the safe way yeah, and then we have another thing that we wanted to do as well.
Kenneth Ilvall:We wanted to make the administration better, so we have some other tools in the line of getting that working even more efficiently, and also have the like a journal, journalgranskning, which way we have talking in English. Do you have another?
Carl-Fredrik Swic:word Journal review, journal scanning, journal scanning yeah, that's a good thing to say yeah.
Kenneth Ilvall:So journal scanning is the other way to actually get everything. So see to it that every doctor has done. Now you're getting into the next product, right? Yeah, before we go there, get everything, so see to it that every doctor has done.
Anders Arpteg:Well, it's now, you're getting into the next product right before we go there, perhaps just speaking a bit about you have a certain state for Luna right now. What's coming up next? What are you working?
Carl-Fredrik Swic:with for Luna specifically. I think the Luna as a product is pretty much done. It's, it's one of the the the first core aspect of speed.
Anders Arpteg:You know, I get really scared when I hear someone say a product is done yeah, but yeah, yeah, it's ready for market.
Carl-Fredrik Swic:Uh, I would like to say it's, it's under. It can always be better. The technology, always has been, has changed a lot since we first started. Yeah, just during these two years, yeah, from day to day More or less.
Carl-Fredrik Swic:Yeah. So, like a lot of companies, we wanted the platform. We saw so many ideas that we can help, saw so many ideas that we can help, and we like one thing leading to another and kenneth's um, um, put us. Obviously you were responsible for putting us, putting us in the direction of okay, what, what is the? What is the core um, the core problem and therefore the core solution.
Carl-Fredrik Swic:You started talking a lot about the, the of the journal. Where is the information about the doctor? The doctor writes down about the patient. There's so many things you can find and it's very important for the whole medical journey for that patient and therefore we saw that this is just a great product we can have as a part of the Varity's loader, the toolbox, and therefore we started talking about okay, so where is the other problem here? Where is one of the most effective way we can actually solve this? As we talked about before, the main mission from the start was how can we make doctors and healthcare professionals better? That was kind of the mission from the start. How can we help them be? I think you put it, you phrased it as the best self.
Kenneth Ilvall:Yeah, every day.
Carl-Fredrik Swic:How can we make the doctors the best selves every day? Improve productivity and quality in some sense. Yeah, um, there's a lot of problems in healthcare, mainly due to just being too few of you, too few doctors, and nurses.
Anders Arpteg:Just a quick question there, because you I think you mentioned journals as well and for luna, is that actually indexing the journals as well, or is it mainly the information about the policies that they have in their region? It's like the last part. Because otherwise you get into a lot of security concerns. Right, If you actually do index patient data as well?
Carl-Fredrik Swic:Yeah, there wasn't any patient data in that part For Luna, it was just information search, basically. So how can we, yeah, so how can we make the doctors more efficient in this particular aspect? So finding information, it was a big time thief, basically.
Anders Arpteg:But do you have something ready for a more upscaled? Is it likea pilot, or what's the pilot, or is it? What's the next step, potentially that you think Luna can be up for?
Carl-Fredrik Swic:No, it's, the product is done. It's ready to use.
Henrik Göthberg:So now it's in go-to-market phase. So now it's like how do we now sell it?
Kenneth Ilvall:No, it needs to be piloted. I think that's the next step.
Anders Arpteg:Because you can't say it's done without having tested it properly. No, it has to be tested in real life, and then it will continue to iterate, I think a number of times.
Henrik Göthberg:But you're mature for proper piloting? Yes, and in reality, if someone is listening, is intrigued, is intrigued. You know pilots wanted yes, yeah, of course, yeah, yeah.
Anders Arpteg:Should we move to the next one with the interesting name. You know I like Luna. It's actually the name of our dog at home, but I guess Luna is a very common name. But you have something called Molly. Yeah, what's that?
Carl-Fredrik Swic:But you have something called Mali. Yeah, what's that? Well, the name is just a very funny and creative part from my side, because we wanted to see how because we had this person in Swedish healthcare called Mal, or medically so the doctor, or also the equivalent nurse.
Carl-Fredrik Swic:So the person who is solely responsible for best practices being used at the hospital or yeah so it's a word play of mal, and Mal is making it a person, someone you can rely on, make it more tangible, make it more humanized, so we don't have this stiff sounding robot.
Henrik Göthberg:So here you have mall is for the mall. You're doing something for the mall.
Carl-Fredrik Swic:Yes, so the mall is responsible for the, that everything is done right medically and that's what we wanted to help that person with with malling. How can we help that kind of person in their process? Because the mall today has a and this is something we spoke to, spoke, spoke a lot about they have a task that's it can't be done. Basically, yeah, they are. I mean, you can, you can speak more about like what they do than I. Yeah, I can?
Kenneth Ilvall:yeah, because the responsibility for the MAL. Now they changed the tax to MLA, so Medically Leading Responsible Doctor, so MLA ledningstansvarig in Swedish and they have a responsibility to see that every doctor and every healthcare professional in that unit has done its stuff according to the regulations and the policies. And how do you do that? They have the journal editing, auditing, and they take like 10 journals for a month and check, but that's 10 of several hundreds of journals. So how do you know that this is the golden standard of their own practice?
Henrik Göthberg:So now you're doing some sort of spot checks.
Kenneth Ilvall:I would say stick proof.
Henrik Göthberg:Yes, Today we do. What's the English word for stick proof? Randomized?
Carl-Fredrik Swic:testing.
Henrik Göthberg:Randomized spot checks as a way to get a statistical view of? Are we? Every time I do my spot check, does it seem to be on?
Kenneth Ilvall:point yeah, on point, yeah. And the thing is that if you look into 99% of all the journals, you miss 1%. And that could be the 1% that actually they did something wrong or didn't do anything that they should.
Henrik Göthberg:In a way you would, with algorithms and statistics and other ways of doing it. You could actually, in a way you would, with algorithms and statistics and other ways of doing it, you could actually, in a way, have a signal system where things are out of sync and you could actually have this on every single journal.
Kenneth Ilvall:Yeah, so that's the, that's what's the general idea of it, so that we can if you can, actually look at every journal that this particular person has produced over one month, or what period that we choose, and then we can see if there's an abbreviation in some way.
Henrik Göthberg:An anomalous or deviations.
Kenneth Ilvall:Yeah, some deviation from the algorithm that we have put in this.
Anders Arpteg:So it's automated journal auditing. Journal auditing yeah.
Kenneth Ilvall:So journal auditing and also the presentation of if there's a pattern, okay. So every person with diabetes is sent off for a referral directly. No treatment, so referral, referral, okay.
Anders Arpteg:Ah, okay that person needs to be trained in diabetes care, so that's why we have so it's really to try to judge the medical doctor, to see if they write the journal and do the proper diagnosis in some way.
Kenneth Ilvall:Yeah, and the first case is just to flag up. Okay, is there anything that is patient hazardous? That's something that you need to address directly. For the sake of the patient yeah that's the first step, the first mission of it. Yeah, that's the first hand, the first mission of it.
Henrik Göthberg:Yeah, but it's very simple. Because you have someone who is according to law in Sweden and regulation, you need to appoint an MLA. Yeah, and this MLA has the impossible tasks to be accountable to make sure that, to govern that. We are yep, we are yep, we are within our practices and policies and we are following that to 100%. And in reality, they do stick proof as a way to. Yeah, I haven't found anything bad, so that means we're good. How can you basically make the work of the MLA better? You know and doing the right thing, but also helping them with the data, the whole data collection and the scanning, and it's an impossible task to do.
Kenneth Ilvall:Yeah.
Henrik Göthberg:At hand like manually. How?
Anders Arpteg:does it work?
Kenneth Ilvall:Well, the, the, the tool, go through every journal and see to it that, okay, has it the flow. So we have, like a problem firsthand. Then you have an anamnesis, that okay, some history of what. Okay, this is the problem that we have discussed. And also then we have some findings. Okay, they have a blood pressure. You listen to your heart and stuff. Like that day you have a blood pressure, you listen to your heart and stuff like that, and for certain uh questions or some missing reasons that they come, then you should do some sort of stuff and and whatever, whatever comes up with the history that you obviously, hopefully that you have written down perfectly, then you should have done this um test and then in the end, then you have the, then you should have done this test and then, in the end, then you have the evaluation of the problem and then you have the okay, this is what I want to do the measurements, that, okay, this is the next step. And if you have followed the pathway of the journal, magnificently.
Goran Cvetanovski:The discourse of the journal.
Kenneth Ilvall:Yeah, the discourse. So if you don't follow the red thread here, then there's a flag for that. Is there something lacking? You said that, okay, I'm going to write a referral, and there is no referral in the system. There's a flag.
Henrik Göthberg:Yeah, and it used to be sharp here. You could probably this auditing of the journal and the procedure check. You can do that completely anonymized. You don't even need, because you're looking at the procedure anomalies, you're not looking at the actual patient anomalies, so this is also something where someone's? Oh, this is sensitive data. It doesn't have to be. We are looking at the procedure and is the process followed.
Kenneth Ilvall:Yeah, exactly. And when you want to, then if you have procedure, is the process followed? Yeah, exactly, that's the yeah. And when you want to, then if you have 10 journals flagged from several hundred, okay, then you have time to actually go down deep in those 10 journals. Okay, let's see. Okay, this is an anomaly and here is some lacking or something, and here is just a lack. Well, they may be, they were, and then they do manual checking of that, or yeah, then they do manual checking of that. Yeah, they do manual checking.
Henrik Göthberg:So it's the typical data-intensive process to be able to scan a lot of information, to then focus the human labor where it makes sense to look deeper.
Carl-Fredrik Swic:Yeah, it's to point out where the potential problems are, instead of just randomizing and looking. You have a stack of paper and if you have a hundred referrals or journals, taking out two of them and looking at them and these two happens to be right, doesn't mean you have a hundred percent accuracy in the treatment of the patients.
Anders Arpteg:It's like a recommender system for bad journals.
Carl-Fredrik Swic:Yeah, you, you talked about all about candidates as well. There's so much information in the journal about the doctor. I think we as a society have this expectancy on doctors to be more than human sometimes, and as an AI system.
Henrik Göthberg:this is a very basic system of finding the outliers or finding the ones that doesn't fit, I don't know.
Anders Arpteg:I can just give you an anecdote. I worked with a similar kind of case in the past and that was more about trying to identify over-prescription of antibiotics in dental care in this case, and we found a number of journals that seemed wrong. They shouldn't really have prescribed antibiotics in this case, and then we tried to see, okay, who is the doctor of these journals that do the most wrongly potentially prescribed antibiotics? And it turned out to be a few selected doctors that did over-prescribe in some way and we thought, okay, these are bad doctors, potentially. But then it turned out when we asked, these are actually our top doctors, the best doctors we have. And then we thought more about it and it turned out like, okay, these doctors usually get the worst cases, the most difficult cases, which is really the reason for potential over-prescription. It's not really that the doctors are bad, it's simply that the worst cases they get more difficult cases to work with it would be.
Anders Arpteg:I mean, I can imagine similar kind of situation happening here as well.
Kenneth Ilvall:Yes, but then of course, you will find it and see to it. Okay, this is something that we can go forward on. So you can see that, okay, this anomalous, okay this doctor has a lot of other doctors that they come in and disrupt them in their work and so they missed out something. Okay, so maybe then should have a lesser workload so they can actually be interrupted from their younger doctors for questioning.
Carl-Fredrik Swic:And if someone keeps continuing making mistakes, for example in one certain area, help the doctor get that course. How can we make our doctors more efficient? How can we make them more knowledgeable in order to make less mistakes in the future?
Henrik Göthberg:But I think you can separate these into two problems, because it's the deeper problem, to go to the root or fix the whole problem. But if you look at this from a pure administration and manual work and we want to augment people's workflows, we can then basically say, like this first step I'm not trying to solve the last step now because that is maybe very, very tricky to solve but this first step we are tying down a person, an MLA, with very simply stupid work that is still not complete, it's just stick proof, so that part of the process we can truly augment, and then the entry point is completely different. And then you have another data and another AI problem. So I also think this is a point where when you try to fix everything at once with the end-to-end process, this is tricky, but to clearly carve out where AI is beautiful to do that, I think it makes sense how you augment the MLA workflow in a certain area.
Carl-Fredrik Swic:Yeah, because what we do. We don't do anything particularly different from the workflow that is done today.
Kenneth Ilvall:Or should have been done. It should have been done.
Carl-Fredrik Swic:For example, they are supposed to do this, and that's what I said before about the unrealistic expectation on doctors. You can't do it it's impossible.
Anders Arpteg:It's too much work. How do you try to evaluate or get them all to understand the flow or discourse of the journal? Have you, like, pre-annotated a set of really good ones and bad ones?
Kenneth Ilvall:yeah, I set pride in giving the best journal ever. No, just kidding. Your own, my own, no, we trained the tool for all my journals. That was some of those that I was particularly proud of and that has a real good flow, that this is something that I'm proud of. And also I have looked at some of my colleagues that I have a real good flow, that I will this is something that I'm proud of. And also I have looked at some of my colleagues that I'm okay, this is a really good doctor.
Anders Arpteg:So you're the good example and the colleagues are the bad example. No, no, no.
Kenneth Ilvall:I just have good colleagues, of course, but then I do a mock up with a journal that's lacking a big chunk of information, so you're removing some parts.
Anders Arpteg:Yeah.
Kenneth Ilvall:Okay, I see. So do they see the gaps?
Henrik Göthberg:Yeah yeah, yeah, of course, diffusion it's not diffusion.
Anders Arpteg:Potentially it's a data augmentation. I would call it.
Kenneth Ilvall:Yeah, but then we can make it in some way. So we have like a typical journal, can make it in some way. So we we have like a typical journal and then we make, make the, the model. Uh, look to a lot of just random journals that we don't have any any say in, and then okay, so here is something that okay, maybe they're good, maybe they are not. Uh, let's see how, how it will.
Kenneth Ilvall:Then you classify it to like good or bad yeah in several different ways, if there's something lacking or there's something needing to be done. I think they have a graph on this somewhere.
Carl-Fredrik Swic:A number of different classes. There's so many things to see and to evaluate in a journal as well, yeah.
Kenneth Ilvall:But if there's larger oversteps or wrong steps in the journal, then of course it's a big red flag and you can have to look at this as the system. But if there's just smaller, ones.
Henrik Göthberg:And then, how is this augmenting? What's your idea of how to get that back into the MLA workflow? What is the UX? So how would the MLA now use this tool? What's the thinking?
Kenneth Ilvall:Well, the thinking is that it should be okay the MLA has to work to see the already written journals, but if we can get the system to go closer, closer, closer to the actual meet of the patient Dr Meet, then we can have a proposition after the workday is done. Okay, so, I'm sorry, dr Kenneth, you're done for the day now. Yeah, okay, I'm done.
Henrik Göthberg:Okay, so maybe you should look into this and this journal because I think you're lacking somewhere something as you can in the end interactivity, proactivity, yes, catching it on the fly, on the fly. So mla, the mla workflow can go from a more uh, policing in the end to a coaching, yeah, proactive, uh, yeah, a different workflow in the end so, in in the best, there's no person that actually goes through and knocks me on my head.
Kenneth Ilvall:It's the system itself, the system itself.
Carl-Fredrik Swic:Well, that's the problem we saw as well. It was so randomized and there were so many journalists that didn't get the MLR to have a look at it and therefore the problem is already done Directly yeah it.
Kenneth Ilvall:And therefore the the problem is already done when, uh, yeah, so like, if you miss a serious symptom, you will find out by the patient coming back and screaming at you uh, later down the line, yeah, or when, when it's something that actually happened, and there's another doctor down the line that actually found my, my fault, my error, and report me, and then I have to stand up and defend myself what I've done wrongly.
Henrik Göthberg:So this is then, when you were really talking about data and AI allows us or gives us the opportunity to really reinvent the workflow. So the MLA workflow is there because we have a symptom in terms of how we need to treat data. Is there because we have a symptom in terms of how we need to treat data. If we reinvent it and can treat and manage and find and check it AI-wise on the fly, then the whole process is different and that whole role is different. It's very interesting.
Anders Arpteg:It's very telling.
Henrik Göthberg:Because I think a lot of times we are trying to fix an old analog process instead of reinventing how the workflow would be if you're AI and data or digital first. Like you just did, now as a good example. It would be in the system.
Kenneth Ilvall:It would help you, it would help you instantaneously. And there is some models that actually try to do that and some journal system that here's when you, you now you have prescribed this drug. Okay, do you know there's a interaction with this other drug. Oh, this is nice. Okay, yeah, I didn't know that. I didn't know that the patient actually had that drug. Okay, why not? Because, yeah, this is a different journal system. So, in the end, when all the the ai can actually find every information about one certain patient, okay. Then with other journal systems and other systems that is outside their firewalls, okay. And then we have a real good thing and the patient care or the patient security should actually be heightened a lot. But then, okay, there is a lot of obstacles on that way, but this remains to be foreseen.
Carl-Fredrik Swic:But it's also the kind of mantra of our companies of how does it fit into the process, Because we talked a lot about there's a lot of companies. That I think has damaged doctors' willingness or trust in tech companies, because there's often a lot of tech companies that come and say, well, we have this fantastic new product and it's going to revolutionize everything and you test it and it's not what you really expected. We had this really big case down a few years ago when this company promised, over-promised it cost the region I think it was Region Skåne.
Henrik Göthberg:We have the last one now in Västergötland region and. Borås. I'm from Borås, you know.
Anders Arpteg:Yeah, so that has been all over the news in the last year the millennium, yeah. The millennium, yeah, cool stuff.
Goran Cvetanovski:It's time for AI News brought to you by AI AW Podcast.
Anders Arpteg:Yes, so this is a small section in the middle of the podcast where we take a break from the discussion about how AI can revolutionize healthcare and speak about some exciting news in the world of AI that we have potentially seen. Anyone that wants to go first? Do you have any news that you'd like to share?
Henrik Göthberg:Something you followed up on or you read on.
Carl-Fredrik Swic:We talked a little bit about it before, but the articles I think that we wanted to focus on is the thoughts and the general expectation of AI in healthcare, and what are actually people saying? Why are they reacting to it? Why are they maybe not as open to AI in healthcare as it can be? I was wanting to see if I find the study I was looking for. I need to see if I find the study I was looking for. Yeah, it was an article about saying that AI is like an overambitious med student, which was kind of funny, I thought and it talks about why AI can't really do what the doctor does, because there's so many processes that the doctor do because of experience and all these kind of things, and I think it's kind of telling as well. We talked a little bit about that before. How can we integrate into the mindset of doctors to be open for the idea of opening their hearts to these systems that we see can help them by gaining their trust and showing them that it's maybe not the workflow that it is today, but we need to accumulate to the work cross with the technology.
Henrik Göthberg:So yeah, my take on that is that I think all over if I go to the industry and data innovation summit in the ambition to sell the tech, to make the license sell. If I'm really brutal, the vendor wants to. We put a lot of emphasis on the tech conversation and making it. Maybe maybe we do that really really well, but what you're highlighting now is that the the implications and knock-on effects of everybody needs to change, like what we talked about before doesn't make this as an easy tech buy that some technology department can buy and then the other ones can use. You need to line up all the different things in order to make this work, and I think in healthcare especially, I think that is how it should be and we really need to take a step back and stop selling in the wrong way or approaching it from the tech angle alone. We need to respect how difficult it is and then work the real process. Like we said, like with your Mac the problem, if you're going to succeed, you need to work the whole chain and I think that's what you.
Henrik Göthberg:When people are saying or shying away, I think they're doing it in a healthy way, because they can immediately spot that this sales process was not professionally done. It was not professionally done. It was not a 360 view. If I take the millennium example, I mean I have used Follett on afar, but it was clearly documented concern from many different angles, what was not clear, but where the technology department or the project leader decided not to listen to that. To me that is a one-sided question. In a 360 journey period, it's not sold in the right way.
Anders Arpteg:I think it also speaks to how hard it is to just find value from AI and what so many companies do so wrong. I mean they, for one, conflate what it means to build an AI product like Malia or something, or Millennium, to actually building a product that actually is something that is of value and can be used. But that's not enough even to have the product. You need to actually have an onboarding on that and being able to do the change management to make them use it. If you don't have all these kind of three phases, it won't actually be used and it won't bring value. And then so many people think, oh, now I have a working prototype, Okay, good, I'm done.
Henrik Göthberg:And that's not even a percent of the work I would say yeah, let's continue this rant, because one of the main reasons why we are getting into these problems is because the steering of how we source and procure.
Henrik Göthberg:So if we source and procure, where we are not allowed to talk with the right people but we are supposed to talk in a certain way with filters in around, and we, in terms of one objective we are trying to measure, actually puts ourselves in a situation where it's impossible to win it's for anyone. So then you need to reverse and you need to go back to what is it that we need to engineer for, or what's the objective function here? And therefore, how does procurement need to work? How does sourcing need to work If we have a cross-functional problem in order to solve here and our processes are one discipline at a time, a tech discipline procuring for doctors is madness, right. So here now we need to reverse engineer, okay, cross-functional in order to make that work. What does that mean in terms of how we do procurement and sourcing? So there's like the knock-on effect on the knock-on effect of an old world model, and when people then cry out and say this is not going to fly, I think they're right.
Kenneth Ilvall:I think you need to fix the right problems yeah yeah, to finger out what's the problem is in the first hand, of course which is the cross function.
Henrik Göthberg:I I argue then, like tech and domain working super close together, not in different silos.
Carl-Fredrik Swic:I would say, like for our company. We have had a very long research and development phase and I think the reason why we wanted to spend so much time on the products and involve and not be tech interested guys, we wanted doctors, we wanted people who actually use it to have a lot of say in how it develops. And, as I was touching a bit on before, I think we, as these kind of tech providers, need to regain the trust of doctors in a lot of sense, because there have been a lot of cases where a lot of things are just forced on doctors, and doctors and nurses and the healthcare profession in general are willful people, very strong-minded. They are very strong-minded are forced on doctors, and doctors and nurses and the healthcare profession in general are willful people, um, very strong-minded.
Anders Arpteg:They are very strong-minded, awesome any news from you, kenneth, or should be anything you picked up on?
Henrik Göthberg:no, not lately.
Kenneth Ilvall:Um, I'm, I'm was actually what the knee, uh on an affair in chista for a few, a few weeks last week, I think there was a lot of talk about the AI in healthcare.
Henrik Göthberg:Please, that's news. Which fair was it.
Kenneth Ilvall:The EHOR, EHOR yeah.
Henrik Göthberg:And what picked up. You know what is your key takeaways. If you want to share the news or the buzz from that fair the buzz.
Kenneth Ilvall:Well, I think the overall news on that was that, okay, there are a lot of apps and there are a lot of smart thing to make you live at home more and to be cared for at home, and also the workflow should actually be easier for the doctors. There are a lot of discussions about that and how you can manage to be a healthcare professional and also be an entrepreneur, so there was also a lot of discussions on that?
Henrik Göthberg:Did you feel that you were on point?
Kenneth Ilvall:Yeah, I was on the right place.
Henrik Göthberg:And what you have been thinking is on point where the conversation was going. Yes, that's nice.
Kenneth Ilvall:You have a lot of different pain points on the way of the journal for the patient, from sitting at home and at the kitchen table and discussing with their spouse or something, the spouse or something, and then in the end product and they have gone from the pharmacy with some prescription or just actually getting it into work. So on that way, and and also there was a lot of discussion as well that maybe not AI but yeah, well, it is there. Yeah, when you have a patient in the care home and then how to monitoring them? Okay, give them medicine, okay, who should be monitoring that and so they actually take the right medicine. That's one thing and the other thing is to to monitoring them in their own flats.
Kenneth Ilvall:If they actually go up to the loo or if they fall down during light. How can you see that is there is a pattern how they walk. So they have like monitoring that and okay, so last time she went up she went in this direction. Now she's going that direction. Okay, now maybe she's not going to the toilet, okay, so maybe she have to interact. So that one was really cool.
Henrik Göthberg:So much technology that could be augmenting and making our care so much better.
Kenneth Ilvall:Yeah, and easier and more safe. I would say More safe.
Anders Arpteg:Lots of potential, yeah, lots of opportunities for entrepreneurs. Yeah there it is.
Henrik Göthberg:Henrik, do you have anything? I'm sitting here with the open AI search mode and I'm sitting here with Gemini.
Anders Arpteg:Do you want to take that? Oh, okay.
Henrik Göthberg:I mean, like one of the things, if we go straight into the AI soap opera, they made search available quite freely, openai. So OpenAI has had, you know, they've launched you know in several steps how you can do search. So you know, openai, you know, has a fundamental search like competing straight with Google and this has gone, you know, from paid subscriber now to actually something you can, without having any subscription, you can start going into OpenAI as a search engine. That was one thing. Do you want to comment on that?
Anders Arpteg:We can continue on that, actually. So one news from just a few days ago was OpenAI released something called Deep Research, which is, you could argue, a continuation. We haven't even talked about that one yet. No, it wasn't released until a couple of days ago. I forgot.
Henrik Göthberg:Let's start there. This one is good. Yeah, start there, it I forgot.
Anders Arpteg:Let's start there. This one is good. Yeah, start there. It was released by Google Gemini for quite a while back, like a month ago but now OpenAI also released and even used the same name. They at least have different names, but they just copy from each other, you know, without any shame. Why should you? Yeah, well, anyway, it's a cool thing, and if we were to quickly describe what it does, I didn't have access to deep research because you have to have the pro version of ChatTPT, which costs $200 a month, but it's going to be released also for plus users, etc. Soon.
Anders Arpteg:But anyway, I did experiment a bit with Gemini's deep research, which is not as good as OpenAI's, but still and what it basically does is, if you take the normal web search that you spoke about, which has been available for some time, it's basically when you search for something and realize I don't have this in my built-in parameter knowledge, I need to go out on the web and search for some more real-time information. It does so, and then it compiles it and provides an answer. And then it compiles it and provides an answer. But if you were to do a more like, say, you do a market research, for how could I sell Mali to a lot of hospitals and then you want to figure out what's the best way to do that, and to do that it's not sufficient to do a single web search. For that you may need to go out and look for 50 websites and then you compile it together in some way and then in the end you may want to have a report. So this is, in a sense, doing what a lot of analysts are doing it's building a report and literally creating a PDF or a Word document with a lengthy report saying this is the conclusions for how you should do go to market with Mavi, for example.
Anders Arpteg:So I experimented a bit with it. I mean, this is also in the direction of agentic work. So, besides having a tool, the tool in this case is the web search. So you can augment the normal LLM or foundation model with a tool like doing web search, but it can do so in multiple steps. So then it takes like so in multiple steps. So then it takes a number of steps, it uses the tool time and time again and then it finds out ah, it's not enough information. Then I, without asking the user, take another decision oh, let's search for these websites as well, and when it's done with that and have completed the plan, first it actually sets up a plan and then it executes, and then it does another plan and then in the end okay, now I have enough information I compile it down to a concise and nicely written report. I did it. We have a thesis student at work and she's looking more into value capture from AI systems, chatbots and these kinds of things.
Anders Arpteg:So I simply told the Gemini deep search to say, okay, build a report in Norwegian this case because she's from Norway and in the Norwegian thesis master student report style and see what the challenges are to do value capture from AI, assistance and potential benefits of doing so and then, it looked up 41 websites, found similar kind of reports that have done similar kind of research in the past, compiled everything and present the report Actually Norwegian, which I don't understand, but still I had to ask them to translate to English, but still it was very nicely formatted in the style of a Norwegian master thesis report that you could submit and then the master thesis project is done. Actually it took 20 minutes, but still 20 minutes, okay for half a year of work.
Kenneth Ilvall:Okay, and 20 weeks yeah 20 minutes, I mean still.
Anders Arpteg:I hope you see the point. It does multi-step actions using tools like web search. That is moving in the direction of agentic, even though it's rather simplistic agentic work, but it does deep research.
Henrik Göthberg:Deep secondary research, or whatever you want to call it. Secondary, what do you mean? Primary research, setting up tests and experiments yourself? Secondary research compiling what other research is already out there?
Anders Arpteg:It doesn't do any experiments, so it's just compiling, yeah. Background research literature.
Henrik Göthberg:Literature research. I've seen the word secondary research used for that.
Anders Arpteg:But kind of cool.
Henrik Göthberg:Yeah, yeah.
Anders Arpteg:Could be useful for you as well, potentially, I guess.
Henrik Göthberg:But okay, so that was a fairly big one. You know what was the sort of case that used to go into the sort of soap opera. So what was the general reflections that you picked up on when they launched research? What was the general buzz?
Anders Arpteg:I wasn't super impressed. I mean, it's rather simplistic and obvious kind of thing. You take web search, which is a tool, you use your LEMs, you build up a plan and build a report. It's something that is not that super complicated, it's just adding like a gigantic framework around it. I saw that Hugging Phase is a big AI platform that have a lot of models, open source etc. They tried to reproduce the work of deep research in 24 hours. So they tried to build their own one, just put an agentic framework around it and use the web search, and they got some rather far in 24 hours in reproducing it.
Anders Arpteg:It's not on the same level as OpenAI, of course, but still it says something. But it's not super high innovation height, so to speak, in this one.
Henrik Göthberg:Okay, but let's go into Google. Now we have Google 2.0 Flash, Google 2.0 Pro Experimental Flashlight. Do we have any view of what this is? This is fresh off the press, the latest release from Google we go into smaller models and more efficient.
Anders Arpteg:And I think everyone realized that the super big models of trillions of parameters is too expensive when you have hundreds of millions of users using them and then feeling a bit of a threat from deep seek and they have to have some smaller models that are cheap and and uh and uh, not too expensive to run when you otherwise can just download an open source model and run it for free.
Henrik Göthberg:I mean it's, it's uh yeah, but and bottom line is used at an accelerated pace, tweaks and you know feature, you know it, we are. We are literally the devop cycle now of any software that gets better and better and better and more advanced. Yeah, and the only way to really keep track of it all is when you're in such a tool and really use it extensively to understand what's happening. You, you need to get your hands dirty, is my point to follow this.
Kenneth Ilvall:But what do your take on the deep seek? Then you want to go there.
Henrik Göthberg:We can go there, because because there I mean, like I think we we talked a little bit about it on this podcast.
Anders Arpteg:In the last two weeks we spoke about it, but I'm getting increasingly tired of it.
Henrik Göthberg:But yes, let's go there. I think it's important because I think you said as well. But if you're looking with the real people, who has the deeper knowledge of this, it's more media hype than what it really is.
Anders Arpteg:It's even worse than that it's fake media news.
Henrik Göthberg:Yeah, there are many. We are comparing apples with pears Like a simple topic. We build something for 5 million and they build it for 100 million. It's not even true, because then you need to look into Antropic what was their original model training? What was their inference training? Yada, yada, yada. So you need to compare apples with apples. We're already there. That is completely wrong. And then you need to look at the starting point for DeepSeq. They didn't start from scratch. Most likely, we know, think that they have started out of something they have started on. The baseline of OpenAI is the general consensus. So they've been cheating? Oh, they're cheaters.
Anders Arpteg:No, and I would actually if we take the pros and cons, very, very briefly. For one, they did distillation from OpenAI, meaning they didn't train from real data. They train from really high quality data from another frontier model, and they were examples like if you ask DeepSeq, what's your name of the model, they said OpenAI, so very badly distilled and stealing data from other frontier models, which of course makes the data size much smaller and more efficient to train. They also didn't have any vision support, which both GPT-4.0 and 0.1 has, which they don't, and that reduces the size of the model significantly.
Anders Arpteg:And they didn't spend only five, six million dollars. They spent hundreds of millions of dollars. But still they did some innovation. I wouldn't take it away. I think they had actually some awesome stuff with the way they did reinforcement learning, the way they used Python code to validate through the steps if this is an accurate solution or not, and they actually did a lot of innovative stuff, really low, very close to the hardware, so to speak, in how to make this really efficient. And that also tells me it's perhaps a bit more people involved in this than people think.
Henrik Göthberg:Okay, but what does it all mean? And here we have a debate, but my take, where I completely shared and me and Goran shared this view, is that what is very interesting and positive about this is that it truly shows a way. There is a game to play, this position to play, even if you yourself cannot master to build a frontier model. So it's a little bit like oh, what should europe do? What should sweden do? We're screwed right. We can never getting you do what open ai is doing or stuff like that. And now, from a media perspective and getting buzz, someone was doing something that we now have picked apart. It was way simpler and actually most of the research they did it was already out in open papers.
Henrik Göthberg:So, then I get fuck. Man. You can make a big splash here with what we know in Sweden and with the competence we have in Sweden If you use smart enough, savvy enough with your marketing, cynical enough to steal with pride.
Anders Arpteg:You know, instead of us saying that- but this is actually literally illegal because it goes against the terms of service. But you know China wouldn't try-.
Henrik Göthberg:Yeah, but I think it's completely bullshit when one guy who has been stealing data from all the other guys is pointing finger that you are stealing my data, but you couldn't argue the efficiency point. I think it's bullshit that Sam Altman can say you have been stealing my IP.
Anders Arpteg:That is the dumbest thing he could ever say but the efficiency argument doesn't hold anymore, so they're not more efficient because they steal from really high quality data from someone else that trains them.
Henrik Göthberg:So we're getting into a deep argument on how big was the? You know they shouldn't have done that. And I and I say look beyond that. Just look at this fundamentals building standing on the shoulders of giants stealing smartly or even legally.
Anders Arpteg:The big innovations here is really you know that they did put it out open source, which is an interesting take, and you can think about why. And then you get into more geopolitical kind of reasons for this. You know, I tried myself, of course, asking you know what happened in Tiananmen Square and of course it censors that and there could be some other perhaps reasons for them putting this out there to actually influence a lot of opinions, but also to gather data or whatnot. So I mean we have to be a bit more.
Kenneth Ilvall:Yeah, that's what I'm thinking about, it as well Of course.
Henrik Göthberg:And then of course, it's always the whole game. You want to destroy the moat that the other guys are playing. So if you're number one, you want to be proprietary, you want to lock it down. If you really want to take away the competitive advantage of OpenAI, make it open source, you know, and then compete on another level.
Anders Arpteg:So it's a lot more than media portrays about this and it's a bit annoying to see how much misinformation and disinformation that is being spread about this. But generally.
Henrik Göthberg:I find it super positive because it shows a position where we can play. I find it super positive because it shows a position where we can play. You can play really really well. Don't just try to go out and do this. Don't try a game that you can't play. Find the game that you can play the new space race it was called.
Carl-Fredrik Swic:It was called the new space race of AI.
Anders Arpteg:It's been around for a long time. It's a big AI race for many years now, I think yeah.
Henrik Göthberg:Well, what was your core? Did you have?
Goran Cvetanovski:oh yeah but, um, yeah, there are a few actually that I think these kids. One is peter salim. So, um, there's a couple of news actually which we are missing. On top of this, we know that Google released Gemini 2, then in Peter Salin, from Scylo AI, basically is going to lead the open source AI initiative in Europe for 56 million. So Europe is investing 56 million on top of all the AI factories I was actually speaking with. It doesn't matter today, but it was interesting that Finland is investing far more, almost 10 times more, than Sweden is due to invest in AI which is very bad for us, and the interesting thing is now, in this open AI Kardashian drama, that we are on to still.
Goran Cvetanovski:So, sam Altman, as you know, they have started discussing like, hey, shit, man, we have been a little bit wrong in not being on this open source race. We should have done that in the beginning and stuff like that. And then, of course, he was involved in making a 500 million billion project, right, where Oracle, him and a couple of others will dedicate 500 million, right, great. So if you have those money, why are you asking another 40 million in funding? They're just basically talking about taking a 400 million in funding, potentially rising valuation to 340 million. That billion. That is how you get your extra hundred that you want to invest. So it's very interesting.
Goran Cvetanovski:And then it was another thing that was very. Let's leave it like that. We, I think it's. You know, we are looking at the race from a fraud perspective and I think that there are very big moves that are happening on a very bird eye perspective and we are not looking at it. We are just jumping as a frog from the ground. So that is from the news from my side. We talk a little bit more next week.
Henrik Göthberg:I think we'll leave it any comments last finish.
Anders Arpteg:I mean they're not going to pay the 500 billion for the stargate.
Goran Cvetanovski:They have a lot of customers.
Anders Arpteg:They make 3 billion less here in europe but the the inference cost that they have is insane. They're not making a profit, right?
Goran Cvetanovski:yeah, I don't need a microphone for this. I heard that. I heard a very, very nice saying. It is like it has always been the case that the second mover has the bigger advantage than the first mover, and I don't understand why people are so angry at, basically, that DeepSeek and many others that will follow on this. They will capitalize on what the first ones did, right, obviously. Why so it's? This is the opportunity for us in Europe, in Sweden, in the Nordics, to be the second movers and not look from a prospecting. It doesn't matter. Let them build the foundation models. We will build vertical AI on top of it.
Anders Arpteg:Yes, that's great.
Henrik Göthberg:You can see how we're getting into the debate. Yeah, it's really sorry that they did an innovation or did anything you know, very spectacular just build the top of an existing one.
Anders Arpteg:So that's what media gets wrong. Then if you should do it, I don't care about that, but don't say that they were as good, or something no right, but we still can innovate on top of innovation, and that's what we should do. Yes, yes, and that's, I wish, what eu will do and what p Peter would do, and not do the idiotic thing of training something from scratch again. That would be. Don't do that again. That's the stupidest thing you can do, right, yeah?
Kenneth Ilvall:Okay, so they're going to invent the wheel again, right?
Anders Arpteg:There's no point. I mean, building a foundation is not actually that hard. I wish they spend money on, you know, not actually that hard. I wish they'd spend money on what is hard, which is really how do you adapt it to your needs? How do you do the proper continued pre-training? Or how do you do the supervised fine-tuning on the data you have? Or how do you preference reinforcement learning on top of this? Or how do you do the reward-based reinforcement learning to have your kind of use or your kind of data in the way that you want?
Henrik Göthberg:This is much more advanced and much more useful to spend money on, and there is a bigger debate in Sweden and there's still data from models. The bigger debate in Sweden is how should we spend our money? Oh, we need to build our own frontier model. Really, is that what we're going to spend our money on? Or are we going to build baby models based on someone else's frontier model? Are we going to fine tune it for different verticals and stuff like this?
Anders Arpteg:I think, in your case as well. I mean, if you were to use the coming which there are coming really big models, that is going to be super expensive and if you put them in production, it will be super expensive to use and obviously that would not be, I think, for you the right way forward. You would have a more specialized model that is tuned for your needs and at a suitable size and at a suitable size, and this is what most products and applications will be in the future and that is what we should be good at. If we are great at that, that would be an awesome future, I think, for Sweden.
Goran Cvetanovski:Yeah, so this whole topic becomes like…. Another interesting thing for you is that this was an open source, so you can download it right. Which one, deepseq, deepseq, you can download it and edit it and make your own virtual one.
Anders Arpteg:Yeah, fine tune it. Yes, yeah.
Goran Cvetanovski:But if OpenAI kills, their service will DeepSeq work.
Anders Arpteg:No, they couldn't retrain it anymore.
Goran Cvetanovski:Yeah, but will this version work?
Anders Arpteg:Yeah, yeah, yeah, I mean, they still have the knowledge in it, but they can't update it.
Goran Cvetanovski:So let OpenAI burn their servers and you make innovation on top of them. Let them burn the money, let them build foundation models. We put them on top of it. Yeah, I mean, that's how we should do it.
Anders Arpteg:It will be like a handful of frontier models in the future, like five of them or something, and then we distill data from that to train other models and fine-tune it for our needs. That, I think, what the future will be. We've been saying this in the future a couple of times in the past. You know it will be a few frontier models, but thousands and millions of smaller models that this actually be possible to use.
Henrik Göthberg:And in practical terms, the cost to run the frontier model on a microservice will most likely be too high anyway. So you're doing that in order to build your model, and then you can use an open source framework, you can use LAMA, whatever and then you get to a point where you do the beta model fine tune for a very specific purpose that you can then run at a drastically lower cost. So the trajectory where this is going, I think it's fairly obvious, because you get to a point where people are just staring at the frontier models and it's like there's no practical use in a device. Okay, good.
Henrik Göthberg:Sorry, that's a rabbit hole today.
Anders Arpteg:I thought it was going to be short news. You opened a rabbit hole there. Yeah, sorry, that's not big Zeke.
Henrik Göthberg:I wanted to have the opportunity he didn't know where he's getting himself.
Kenneth Ilvall:I realized that after a while. Anyway, well, it's good. Thank you.
Anders Arpteg:Did you yeah?
Henrik Göthberg:Yeah, I learned a lot. Did you learn some? Yeah, yeah.
Anders Arpteg:Carl and Kenneth. Working in the medical field with AI and tech in general is hard, I guess, and what would you say? That the biggest challenge is to really do? Yeah, I hear something.
Goran Cvetanovski:There is a phone that is on voicemail going around.
Henrik Göthberg:That was my fault. Sorry, I didn't hear it because I'm in the headphone.
Anders Arpteg:Yeah, anyway, there are a lot of challenges, I guess, to truly put AI systems, or tech systems in general or IT systems in general, in the medical field. What do you say the biggest challenges are what have you seen so far and what's your thought about really making sure that you find value, that they are being used in healthcare Well in general?
Kenneth Ilvall:the largest is the tech, not awareness, so to speak.
Henrik Göthberg:The lack of awareness, the illiteracy, the resistance.
Kenneth Ilvall:The resistance, but the resistance is based on fear. Yeah, illiteracy, illiteracy and fear and they go hand in hand. It's something that you are afraid of, you don't want to actually.
Anders Arpteg:Why do you think they're afraid about AI? Is it because they think they're losing their job? Or why would they be afraid about?
Kenneth Ilvall:this. There's a lot of discussion of that. Okay, in some way in the future we will be out of a job. But then, okay, some voices raised in the no, you have to have someone who actually see if this process is hallucinating or is something, to actually take a hand on the patient, and so you have something to guide them through this. So maybe our role will be a bit more of a guidance, um, counselor or whatever, but we, our use in the medical society, will be there. I'm I'm positive of that and and if, if we.
Henrik Göthberg:Really, how can we even think like that when we, at the same time talking about that, we are swamped in administrative work?
Henrik Göthberg:exactly so how can you take those two arguments at the same time? I really think it's we. We are really complex people, no, but I mean like so. And we are talking about the whole. If I go to the public sector and elderly care and all that, we can see a shortage of, you know, 30-40% of the workforce going in retirement and we are not seeing the same numbers in new recruitments. No, not at all. So we need to literally flip 30-40% shortage to tech.
Kenneth Ilvall:Yes, that's why don't we do it?
Henrik Göthberg:So why don't we do it then? And then you say that the major hurdle, one of the major ones, it really starts with education or learning to get it into your comfort zone.
Kenneth Ilvall:Yes, to get it into comfort zone and have used it and people who have used some of the tools that we have now on the market. They don't want to go back.
Henrik Göthberg:No.
Kenneth Ilvall:So why should I write my own journal again?
Henrik Göthberg:Yeah, and I think you're on the money here, because if you have knowledge on the deeper level, then that knowledge disseminates out into the different disciplines that needs to work on this the regulatory dimension, the safety dimension, the adoption dimension, the engineering dimension. But it really starts, in my opinion, if I take that as truth, this is a literacy problem. What is then the real problem? Money, money. Then the real problem Money, money.
Carl-Fredrik Swic:Is it really money, time, fun and time yeah.
Kenneth Ilvall:It takes time and money, of course as well, to actually prove a point. If you have a proof of concept that you want to engage in some way, okay, this is something that I'm burning for this. This is something that actually will renew the work that we have here on this premises, okay, and how much will it cost? That's the question, not what would the benefit be? That's not the question. So they put it over the thing. The manager of the healthcare center says okay, how much will that cost a month for me? Okay, not the benefit it will make.
Henrik Göthberg:But I want to test something with you. Yeah, sure, and you will follow me when I say this, because when I think of this knowledge gap and the way I've been failing to communicate my way through this, I come to one single conclusion. There is only one way, and this is learn by doing, and ultimately, we get to the famous Spotify quote, which would be the fastest?
Anders Arpteg:Which would be the company that fails the fastest.
Henrik Göthberg:Spotify, tony Leakes says it and of course it's about fail fast to learn faster, right but you can't really compare that to medical, because we can't fail at all. Yeah, you can, but then if you understand that, the main problem is how.
Anders Arpteg:I'm on board but but not in general, and do human trials at the pace that I think no one else is really doing, even in this kind of very difficult field. So I mean I think there are ways to do this.
Henrik Göthberg:So what we're talking about here is to be precise on what is the problem to be solved, because I think it starts with acknowledging that it's an education or it's a literacy gap, and then, if we have been in the field, we have understood, all of us, that there is only one way to do it, and this is experiment.
Kenneth Ilvall:Yeah, I'm on board with you guys. That's why I'm here. No, no.
Henrik Göthberg:So what I'm saying is then, if we're going to talk to our politicians, if we're going to talk about how to invest money in this, I would be very, very scared if someone did a huge AI investment program at this point in time. That looks like this as a kind of big bang approach. I would rather want to see investing heavily in how do we do safe experimentation, how do we ease our way into this of spending a lot of money, but in the right way? That's what I'm trying to get to.
Anders Arpteg:But we also have, I think. I'm not sure who said it, but I think he spoke about the opportunity cost of not using AI, and he phrased it in terms of healthcare and for administrative tasks, which I think you can use as well. He said one way to phrase this yeah, we're going to make medical doctors 10% faster and they're going to do less boring work. Okay, that's a safe speech Not a very good one, potentially but if you instead say we are, by doing this, not going to have 10 people killed or be dead because we can actually spend time on patients instead of in general systems, that is a much stronger statement. Right? So you speak about the opportunity cost.
Carl-Fredrik Swic:you know, if you have to spend so much time doing administrative tasks instead of helping patients, people are going to die yeah, I think a lot of doctors are just also tired of being told to be more effective because you can only beat the mule until it dies.
Carl-Fredrik Swic:Yeah, and that's I think we need to. We need to. We would like doctors uh in in medical professions to be more open to ai, but they are skeptical. And where we need to find out, why are they so skeptical? And they have been burned so many times with the next new thing and and that's why I think it's, it's very hurtful when a lot of companies just rushes in with new products and doesn't do the homework like in the way it should have been done, the change management that needs to be, yes, yes so let me try another one, because I'm working on exactly the same problem, but in enterprise, and we are talking companies like Scania or Tetra Pak or something like this Having the same problem of adoption why doesn't it happen?
Henrik Göthberg:And a few people knows about it and sees it and is on board. And here I can kind of pinpoint another problem. I think it stems from education. I think it stems from education, but it moves into a very strong topic of the way we have organized our competencies has been in separate disciplines rather than cross-functional teams. So if you go and look in detail, how is Spotify organized? How is Klan organized? How is Digital Native organized? So, even if they are Digital Native and it's all tech, when you really scratch the surface they are very clear cross-functional teams.
Henrik Göthberg:This is the data scientist, this is the data engineer, this is the behavioral science psychologist, this is the UX person and this is the product owner. So you have a completely different way of how you have organized the competences. So you, you, you build away these, uh, communication gaps or lingo gaps, so and so I think it stems from competence. But that competence then takes the one drawing the org shot, the one setting up the operating model, the one making the budget between tech and business. It needs to change. So all that thing is blocking.
Kenneth Ilvall:I have another take on the problem, if you may.
Henrik Göthberg:Yeah, please please.
Kenneth Ilvall:Yeah, I think this also is the lack of time. If you have the time to actually test something, if you have the time and energy to actually well, this is something new. I want to test this. Okay, I take the day off now and just dig into this. But what about the patients? Who will take care of my patients that day? Well, no money. Who will take care of my patients that day? Well, no money. Okay then, because I can't afford to have a temp that day, so that we have a problem. So we need to have people who have the time, or make time, to actually look into something new, and that's a bit rare in this slimmed organization. So we need more money to actually put off the-.
Anders Arpteg:Is it money? I mean, it sounds like they need to take a leadership course and think about what the growth mindset is versus the fixed mindset.
Henrik Göthberg:Yeah, but I can-. Let's run with your comment, and a nice word to describe this is using the word slack. If you use the word slack, you need to have in order to innovate, there needs to be slack in the system, and if you use that word, if you Google slack, there's so much research on this. Another key topic that we are describing, which is different in a digital native, you have the development and operations in one, so there's slack in the system to innovate, to develop and to operate. What we have been trying to do in the medical field or in the old analog places is like oh, this is the line organization and here is no slack. Now we put our programs and projects on the side. Here we have billions of money. We are investing in the Millennium Project. That, then, is trying to move in that way, and so it's a completely different way of spending. The same money could be reallocated as slack and investing into the core business or the core of our segment. So it's also. This is a reallocation of steering of resources, of innovation.
Anders Arpteg:Still, I worked a bit with medical fields as well, and I think there are a lot of doctors that are very, very interested in this.
Kenneth Ilvall:Yeah there is yeah, fortunately Very interesting so.
Anders Arpteg:I think there are examples of this.
Henrik Göthberg:But they're working against the system. They don't have any slack, they don't have any money. They need to talk. If they want to do tech, there's someone else they need to talk to. So everything is stacked against the innovation.
Anders Arpteg:That's what I'm trying to get. You haven't mentioned regulation, by the way, no, that that's what I also top my mind now.
Kenneth Ilvall:it's a regulation you, you are bound to do certain stuff and you have, you, you're not allowed to do other stuff, because if you do that, then the regulation will prevent you from actually having your core business working at all.
Anders Arpteg:So time is flying away here. We could speak a lot about regulation. I think we've done it so much, I don't think we need to go into it, and I think everyone knows that regulation in the medical field is even harder than in the normal field. It's not just GDPR anymore that you have to follow, but a lot of other regulations as well, which is tough, right.
Anders Arpteg:Anyway, I think it can be done. I think it's, with the right leadership in place, we could actually make this happen, and I've seen it happen in places. So I think it can. I hope you're right With that, perhaps thinking and moving a bit to more closer to the end kind of topic. But what's next for Medic Tech? Can you speak a bit about what the coming roadmap for 2025 would be, or what's going to come up now, coming months and years?
Carl-Fredrik Swic:We have a lot of interesting projects going on. It's unfortunate to be a lot of things going on. It's unfortunate A lot of things the customers we have talked about, we have signed NDAs, so we can't really go into too much about it. And there's a few things we're actually doing now that we haven't seen anyone else done or figure out. So we don't want to go too much in detail and and spell out the secret sauce, so to say. But, um, I think what we're we're trying to do now is just looking at the market, because in, in, uh what is the market?
Anders Arpteg:by the way, is it sweden mainly, or is it other?
Carl-Fredrik Swic:places. I think we uh, yeah, sweden is, sweden is the main market and I think you can always in the beginning, uh, you play best at home field. Yeah, um. So like perfecting and getting our products in a way that we really think that doctors can use, because if you see how the doctors react to a lot of products that have been pushed to the market, they are, they are relentless in their criticism, and I saw the comment about a big, um, a big product being pushed to the market now and the the doctor's comments were something in the line of the more a person needs to use ai, the more the lack of natural intelligence. That's so stupid, yeah, and so it's those kind of hurdles I think we need to work around and I keep circling back to getting the trust of the people using it and the mantra of Meditech, but finding our place in the workflow.
Kenneth Ilvall:You need to have ambassadors on those places.
Anders Arpteg:I think that's that's change management.
Kenneth Ilvall:Yeah, change management is one thing and you have to have someone who actually have used it and just stand on the forepost. Okay, here I've tried it and I'm a doctor and you know me, and I think this is working.
Anders Arpteg:Get on board. And speak about opportunity costs perhaps, and say you know and I think this is working Get on board and speak about opportunity cost perhaps, and say you know, who do you think in five years will be winning? Is it the one using AI or is it the one not using AI? I think the answer is very simple. Yeah, of course, it's a very simple answer to that right.
Carl-Fredrik Swic:Yeah, but the thing is we have Twice the many patients. We've been speaking to a lot of regions and we were so close in getting our product into one of the regions I don't want to shame them, but not yet, Maybe one day but everyone was so on board with the product we were clear for the lawyers to sign off on it. The doctor said they want it. For the lawyers signed off on it. The doctor said they want it. And then when we wanted to push into the region, they said well, this kind of product can air out our dirty laundry and show all the mistakes we're doing. We don't want that to be out there For our legacy?
Carl-Fredrik Swic:Yes, and we were like you. Serious is this? Is this why you're stopping the product? Because you see the value, you can see the potential of saving lives and uh that's actually a very horrible concern yes and this reason to stop.
Henrik Göthberg:Yeah, that is really cynical.
Carl-Fredrik Swic:yeah, we, we were kind of like, uh, that dropped our jaw a little bit.
Anders Arpteg:That was the reason, and perhaps you can, you know, rephrase the way they do it and make sure it is there for helping people or medical doctors as well. It's not about you know finding problems, it's about helping them.
Henrik Göthberg:But it makes it so clear navigating. But then it makes it so clear navigating the messaging and taking away their concerns so they don't end up in the tabloids. If that's the biggest concern then this is fixable, of course.
Carl-Fredrik Swic:Yeah, and that's the thing, and that's something we need to that we have seen a lot and we need to focus on that part of it. We need to focus on that part of it. There's very few benefits of being first and being innovative in healthcare. We see that because there's so many things that can go wrong and there's very little upside to you of being first, because you get a little pat on the back and the risk is getting fired and being shamed. So I think one way to push innovation forward in the medical field is for someone to benefit from it.
Henrik Göthberg:Find value. Yeah, Find value and lower the threshold, lower the risks.
Carl-Fredrik Swic:Get the right people, because there's a lot of people saying things and are the voice of the AI from the healthcare perspective, that have no clue what AI is about. They think they're AI experts because they attended a three-month course.
Anders Arpteg:Yeah, I've seen that Been there, done that yeah.
Henrik Göthberg:There's quite a bit of that. Is that a problem in itself? That we have now all of a sudden a lot of self-proclaimed AI experts and they have opinions that they've done good work and in very much things it's great because we're building momentum, but we the risk, I guess, can be that we're getting on the wrong train, so we're doing it in the wrong way, like you said. So we're doing it in the wrong way, like you said, like someone is pushing product in the wrong way, or I mean like we have another example you know your CEO goes to a course and someone shows how easy it is to put together a chat, gpt setup for your own personal productivity, to make invitation for your children's birthday right, and then the core problem with that is how hard can it be? And they think that it's a quick fix to get it done in the enterprise world or in the real world, when in reality you need the data.
Henrik Göthberg:All the stuff that you guys are working on in the back end is almost unseen and all of a sudden we get to a very dysfunctional perception of what you should be working on and what you should be doing. So this is what I think is really trickiness when so many people are also standing on the hype but not really knowing the full length of what they should be doing. We should be doing, but not really knowing the full length of what we should be doing, but I'm glad.
Anders Arpteg:I'm sorry that you have NDAs and can't speak about it, but it sounds good that you have a lot of plans at least coming year?
Anders Arpteg:I guess Awesome. And to end with our standard question and being a bit more philosophical as well do you believe AGI will come? By the way, let's say we define AGI as the time when an AI system can be equally good as an average co-worker medical doctor or nurse or whatnot, meaning literally not only understanding text, but actually being able to take action and do the work that the human could do, which they vary from from today, I would say. But still, if we assume agi means being on the level as an average co-worker, do you think it will come?
Carl-Fredrik Swic:I think it's, it's unavoidable at some. That's, to some stage, like I am an AI optimist, of course. Yeah, well, I wouldn't be here if I wasn't. I wouldn't try to push it on these medical professionals if I wasn't. But there is this concern. It feels like that we are all going to be this fear of being replaced. And you have this one, one group of people saying that, well, if all the boring jobs and all the things we as humans don't need to do anymore are taking over, there's so much more time for us to be our true selves and we can go hike every day and we can do all these things that we need to fulfill our lives. But what are the purpose, then, of human existence? If you do like just I I'm, if we go into my personal beliefs, it's, it's just to strive to something that makes you to go forward, like what makes a person want to wake up every day.
Anders Arpteg:It's to, it's to develop and do things that are important and and hard yes, let's not go that too, because that's a follow-up question, basically, okay, do you also cannot believe that it will come at some point at?
Kenneth Ilvall:some point, yes, but as a medical doctor in in general practitioner's office, then there's a lot of levels of that knowledge and it's not. I mean you need to have like a robot doing the things. Yeah, so you have to have the physical person that can actually walk around and do something. Yes, and then okay, then we have it. Then. Then there's, of course there's yeah, well, in some so many levels that Thinking about when do people accept a robot to do this? Yeah, exactly, to pat on the shoulder and maybe do an exam of the belly or do a pelvic exam.
Henrik Göthberg:But I really like Anders. You have cooked together some sort of different levels based on whatever other people. There are definitions of levels of agi, but I really like the ones we've have.
Henrik Göthberg:We've been discussing where we talked about three ones yeah, the three or four, maybe with societal, maybe the top one. But you have said about first we have the knowledge, the fundamental knowledge, oh, you mean the opening eyes, levels of, or you said, yeah, maybe it is, but you would you talk about. I think sometimes about first we have used cognition. Ah, yeah, yeah, okay, and you know that we can ask, and actually the cognitive level is as any co-worker, yeah, and then you get to maybe to digital, administrative, digital work. So digital work where you're at the level of a co-worker, of actually taking action, but your main interaction today is a screen. And then you get into, you know, so you can do stuff right, but you can only do it in the digital space. And then you have the third level where we have this sort of the real actuations you know physical AGI Robots.
Henrik Göthberg:Robots yeah, and then you actually mentioned the next level, from the robots to societal AGI, where it's accepted, yeah, disruption in society. So, and of course, here we have the Mac. You know, this is the first level is almost there. The second level, not so far away. The third level, with robots, oh, it's pretty far away. And the societal, uh, it's generations away, most likely not generations but no, I think so because it I?
Henrik Göthberg:I think it's about um. The newer generations will not. This will not be a problem for the ipad general. The ipad babies won't have the same problem with that in 20 years.
Anders Arpteg:But it's also that organizations are very slow. So even if we have the technology there, it can work both in the perception space, in the digital space and physical space. It still wouldn't change how Karolinska Institute works. Potentially it will take a lot of time for them to change the way they're working, and that will take five, 10 years.
Henrik Göthberg:So the societal topic is that, then we have fundamentally reinvented everything, and Karl-Heinrich Svanberg talked quite well about this when we have also on data innovation summit. He was arguing that when he was the CEO for Ericsson, you know we had a fairly good idea of how we thought this would play out with IoT or digital blah, blah, blah, blah blah, and we were envisioning not too far away from where we are today. It's the same, like internet and all that. We could see it in 99, but it took us 20 years for the fundamentals to have reshuffled, so to speak, and this is what we're talking about again.
Anders Arpteg:Okay, anyway, assume at some point X number of years in the future we will have AGI. We can imagine that becoming in the two extremes and a spectrum between but one extreme, of course, is the matrix and terminators of the world and robots going rogue and killing all the humans and whatever. And then the other extreme is more toward what Nick Bostrom now wrote in his latest book, deep Utopia, and we move into a utopian future where AI have solved cancer and AI have solved climate change and energy needs and we have fusion energy and it's free for everyone and we basically have a world of abundance where goods and services are free to use. That could be a utopian, at least for some people, I think. Where do you think we'll end up? What's the most probable place if you rate it from 1 to 10?
Henrik Göthberg:dystopian to the real utopian. Dystopian to utopian.
Kenneth Ilvall:I can start. I think I'm liking the idea of utopia, so I'm leaning on utopia.
Anders Arpteg:For just because you have the hope or because you really think it's going to happen.
Kenneth Ilvall:That way, I'm really hoping More than actually believe it in my core, because people are a bit, people are weird and people are mean and the world is going in a strange direction. Yeah, it is in the real strange direction. So maybe if we really put our mind into it and really have the path lined up to us, okay, so then I will gather we might end up in utopia eventually.
Anders Arpteg:So 70%, 90%, yeah, 70%, 70% Awesome.
Carl-Fredrik Swic:What about you? A few weeks ago I went to a seminar by Hans Rosling's son. His son's name escapes me now, but we are so programmed to think that everything is going to be so bad all the time and there's the human bias of everything is going to shit, but we have survived.
Henrik Göthberg:Actually, someone said that is how we actually survived, that we always are thinking about the risks for the mammoths. Yes, Ola Rosling.
Carl-Fredrik Swic:Yes, Ola, yeah, that's his name and they have this. They pull down the pants on everyone at the lecture and saying, because everyone took a test, like how is the world going right now? Is it more wars, are people dying? All these kind of very interesting kind of examples, and people still, even people who work in med tech and I think are very educated on the business of how the world is going and developing, still are very negative on how the world actually looks. Um, so, yeah, I, I have a hopes, I have a high hopes for the future. I think we will get into the good for you yeah, that's good, yeah, awesome but I I think I mean like the.
Henrik Göthberg:the tricky summary of this is, I think, the the the book is not written, so it's all about how we as society, innovators, politicians, that we take charge of where we want it to go. So sometimes I also think the debate is a little bit like everybody makes themselves into a victim. We should forge this destiny and I think that's the main. We should make our own destiny right isn't that the saying?
Anders Arpteg:you know, the best way to predict the future is to build it. To be it awesome, yeah, awesome. I hope you can stay on for some more after after work and off camera discussion and going even more into more philosophical topics. But thank you so much. Carl Fredrik Sick and Kenneth Illval for coming here and best of luck with your very societal benefits. You know work that you're doing, trying to improve and revolutionize healthcare with AI.
Henrik Göthberg:So basic, good ideas.
Kenneth Ilvall:Okay, true heroes.
Anders Arpteg:True heroes here, here. Thank you, thank you.
