[MUSIC] Hello and welcome to the next episode of Transpurged Lifesight Talks. I'm your host, Mark Wade. I'm the practice leader here at Transpurged Life Sciences. Today, I'm joined by RZ Kenzor. He's the Chief Operating Officer and Co-Founder of Estena Solutions. And Standard is an AI software organization that lives in digital health. And actually, before I even butcher more, why don't you introduce your organization because you'll do it better than me? Yeah, thank you, Mark. I appreciate the opportunity to chat today. Standard is a company I co-founded way back in 2003. And we do focus on digital health software development, data, and AI more so now than everybody. It's just what everybody wants to talk about and focus on these days is the idea of AI. When I say digital health that encompasses a large volume of potential projects and potential opportunities, we really tend to focus on the R&D and the clinical trial side of things. Helping companies develop experimental solutions, really trying to push the boundaries to improve patient outcomes. Health and wellness, if we can develop a solution so that people can live longer, healthier lives, we're much happier. I'm glad that you said it because to me, digital health or DH is just this enormous animal, this enormous behemoth. And we talk about the entire patient journey. So everything from home health and the technology we give them at home health and the technology we give them in clinical trials and the technology we give them in sites. It is a massive thing. And you speak to all of that life cycle and you are looking for innovative solutions. Yes, we really focus on that patient side of things in that journey of helping them improve. So it might be a clinician focused on something that a clinician is using to better understand data or the situation. Something a patient is using to improve their own health and wellness. Or even when we work with medical device companies so that they can better understand how their device is being used out in the real world to improve that patient experience. That's interesting. Because in my world I see AI is an overhyped, overworked phrase. And I think it's really proven. We've proven it's worth in imaging. There's no doubt about that. In C-A-Way where I live, AI is still a gray area because AI is so data-hungry. And by the nature of the B-C-A-Way is not data-hungry. So do you see that as an opportunity for AI in C-A-Way or do you see it as is more of a clinical trial larger enterprise? I think there's tremendous opportunity for what we can use AI for. There are areas we haven't even thought about yet and how we can apply it into better understanding a patient situation. Either during an actual clinical trial or just out there in the real world. Because it enables new paradigms that we as humans can't get at. When you do start talking about large volumes of data, we can't process it and understand it intuitively as an AI can't. Just drop a large data set and you start looking at it. You see a bunch of numbers. What's the first thing you have to do? Well, you need to graph it and chart it to start understanding that. But if you ask the AI to start teasing out these ideas and find sub-phenotypes in the data, there's a lot of opportunity there just examining the data. Let alone looking at different paradigms which we haven't really thought of just yet. Think of in medical fields, one of the biggest use cases now in the hospital, the clinical encounters, ambient listening, listening to the encounter to take notes. What if you take that technology and then listen to the emotion of the user to better understand where they are in their journey? Do you feel that AI can actually parse the emotional piece out? Yes. There are various solutions that are out there that are doing these today and very experimental early methodologies of trying to do that. But imagine in a clinical trial where you're going through how the trial is going to work and the informed consent. Is that person really understanding it or not? Can the AI start detecting notes of hesitation in the voice in notation of that individual that you're working with? You may not pick it up, but the AI has a pretty good shot of doing stuff like this is fascinating because I know about bringing it back to my world, but in my world, we do a lot of cognitive briefing where we're interviewing patients on how they perceive a particular instrument. And I've always been an advocate for the face to face cognitive briefing because the tradition will be very sensitive to the body language, the nuances, the timber, the tone of the person, the patient. This is kind of like out there. You're saying that AI can actually substitute that cognitive briefing interviewer and parts out those particular identifiers. I find that fascinating. It's one of those use cases in its early stages of understanding the experience of humans and not even in terms of clinical trials and medical encounters, but understanding where people are in the journey. It can pick up those little things. You can understand your heart rate. You know, we're not talking about chat GPT. You know, for general use cases, there's a lot of these AI's that are more specialized for these specific use cases, but as they do get more and more, what's called multimodal can interact with different instead of just text when you, you know, chat GPT, you're sitting there typing and interacting with it. Now it does have access to video. Like it wreck when I use it on my phone, it recognizes me. It recognizes my wife as scanning around the room and it can make different negotiations in what it's doing. Talked me through, because I hear all these things, but not every AI engine is the same. Some are very different. Some like you so much, chat GPT is quite functional. What you're talking about here is much more intuitive. So can you talk about around what are the different engines and why would that engine be so unique to your average engine? What's fascinating is these models, large language models, they're leapfrogging each other as we go month to month and sometimes even week to week. Think of this last two weeks ago as we're recording this, chat GPT and some of the Google models, incorporated the ability to create images through that natural language processing, where there was a lot more specialized systems out there. It was very focused on just creating images. That's what it was meant to do. Now chat GPT has that capability built into it. So it's marrying together that intersection of that language processing and understanding what you're looking for with the ability to create an image. So it's fascinating. You can upload a picture of yourself. I've been playing this a lot myself. And it's like creating image of me surfing on a lava across the ocean. And there's a picture of me surfing across a lava on a ocean and it's fascinating what it can do. So it's marrying these two things together. What's going to be possible next week, next month in these different models. They're just continually leapfrogging each other. It's interesting because once one of these vendors comes out with a particular approach, a particular solution, then all the other ones start jumping on the bandwagon kind of thing and see who can. It just shows how nice and this whole technology is. And you actually bring up a good point there. So there's you surfing on a lava. If you like it, right? One of my worries. I'm sure a lot of you of the audience can worry about these things. All these like defects. Once we start you know, imitating your voice, imitating your mannerisms, the AI learns you. And suddenly you are interviewing me naked, you know, doing something. You know what I mean? Like there is no limit to it. That's concerns me. It is 100% a concern. I live in New Jersey and they just passed legislation today. Outlawing the idea of deep fakes. It is a growing problem. You might remember the one couple months ago back of the Pope wearing a large puffy white jacket. Everyone really knew this wasn't real. But it was definitely amusing. You're like, "Hey, wait a minute." And that technology, that ability is just getting better and better. For my job, I do a lot of interviewing of candidates. In my interviewing a real person or not or even the other scenes. Are they using AI to be able to do better on that interview? But at the same time, it's an opportunity for me. How can I make our interview process better? So sort of bringing this back to the idea of the clinical trial. How do you create that digital avatar to be able to interview a thousand people in a week? In my world, in terms of hiring people, I posted a job a couple days ago and got like 3,000 resumes. I don't have the resources to interview 3,000 people. In a clinical trial, you want to interview a thousand people. You know, you as an individual or get five individuals. It's going to take a long time to process that through. But if you can create a digital avatar to now interview these people, then you're going to have a recording of that transcript. Can use the AI to listen to the language of the patient in the interaction. It's a powerful potential to scale. Yet all that being said, that's wonderful. However, I'm very content of AI hallucinations are very content that it.
does introduce a bias, whether we like it or not. We saw that last year, didn't we? We saw that there was a bias introduced in a very large, I don't imagine the name of a very large data set that was published and it was had a massive bias. And my feeling is we are running a tiny bit too fast and right now we do need a human in the mix. Your point is well taken that your AI engine could infuse thousand candidates. But at some point it's got to be a human in the mix. And my worry is that Johnny who's a good candidate for your AI has a bias and says no, Johnny's not suitable for the job. And Johnny actually is very qualified for the job and he's very, very good for it. So I think we can't be on all the parts. But right now I just think that there's got to be some sort of guardrails and some sort of like human in the loop definitely. - I agree with that. And we haven't made that leap to the AI first round interview just yet. We're still using humans for that. But that first round interview is verifying information that that candidate has submitted to us to make sure it's accurate that they understand the job they're applying for. So it's very basic hallucinations. Yes, it's a challenge. It's a problem. But there are known methodologies to implement to help alleviate those. And what I find fascinating and all of the AI research and everything that's doing out there is this continuing, we're raising the bar of what does it need to be able to do for us to find and to actually use it. Humans make mistakes all the time because we're human. We don't call them hallucinations. And I remember what that number was. So I made up a number, I fudged it or I generalized the number kind of thing. Humans do it all the time. But when we talk about these AI, I don't know, that's not acceptable. There are clearly situations where you have to be absolutely certain in terms of medical treatments and things like that where it's life or death. And you can't have any hallucinations. And even with humans, you're going to have multiple checks and balances. But implement those same checks and balances with the AI. And it's all something the other day that was an article and it wasn't peer-reviewed at the point that I read it. But it was looking at AI's doing PhD level math competition problems. And the AI didn't score that well. What was interesting was the AI was getting the right solution but wasn't presenting its proof in the way that the examiner was wanted it so it was scoring low. And that was a criticism of the AI. Well, how many people need the AI to do? PhD level math competitions. I think you need to start using it and playing with it today. I'm going to push back slightly here because you talked about like humans make mistakes. And we absolutely do. But the problem is that the average human mistake is not as huge as a machine mistake. Let me give you a real-world example. An AI was asked recently about how to clean or how to do a gut cleansing thing. You need these health and it's gut cleansing, whatever. And the AI said, yes, use bleach. Drink a pint of bleach. And it will clean out your intestine kind of thing. Now, a human would never make that mistake. I actually know. No, no. I don't believe in that. It's not as you preach. And I won't name any politicians. But-- [LAUGHTER] OK, that's how it is. So-- Yes, there are clear examples where the AI goes off the rails and you're like, no, eating rocks. And there was an example of just eating rest. Yeah, you eat three rocks a day. You do have to apply some of your own human intelligence. And it's getting better and better every day. These things are fading off into the background. And there are checks and balances to put in there. If you're using it for medical purposes, you are going to put those appropriate checks and balances in there. But in terms of a second opinion, in terms of understanding the challenges, again, sort of bringing back to clinical trials in your world, as you're developing protocols for a study, running against the AI. And you can create personas in the AI. So you can tell the AI to act like the person that you are targeting in this particular study to see how it interacts and how it responds to your study protocol, to your own-- I actually-- --to get really good feedback in those early stages. I agree with you, by the way. And we had another person on the podcast who their company actually does AI protocol iterations. And that person talked about that. So thanks for the plug. There you go. [LAUGHTER] Runs off. I agree. I mean, it's definitely-- I think one thing I will say is like anything. Do you remember like a world enough to remember the Intel law? What was the founder of Intel? His law? More, more's law. More's law. Everything would double coming. And we're definitely-- we're in that world where AI is leaping forward in capability, exponentially. As you rightly said, they're leapfrogging each other. Someone comes out with a new capability. Literally, a day later, someone else comes out with something very similar, if not better, a better strap. So I agree with you. It's exponential. I live in the technical world. Right here right now, I think we need a human in the mix. I think we're not there yet. But I totally take your point of view that we'll get there. And we'll get there faster than ever. Yeah. I agree with that. So one huge proponent of AI. Yes, a human still has to be in the mix. It's an 80% solution. I think of it more as a really smart intern to the first year employee kind of thing. It's my make mistakes. And you have to give it really good instructions to get what you want out of it. In an example, if you remember Alton Brown, who did a lot of TV shows and what a huge network on. I went to one of his presentation variety show that he did. And he talked about AI. And he's like, I was skeptical. So I gave it bunch of ingredients that were in my refrigerator and told it to make something. And then he made this on stage kind of thing. And it turned out to be disgusting. Well, you gave it a list of disgusting ingredients and just told it to create a recipe. And it did what you told it. You didn't tell it to create something that tasted good, in which case it might come back and say, these things don't go together. Yeah. I did a presentation on using AI and software development and trying to gather requirements and figure out requirements. And I had this idea of creating an app to help people when the fear of peanut butter getting stuck to the roof of your mouth, something bizarre, unusual. It came back with one word, really? It just floored me. It was understood that I was like, well, can this is a little tongue and cheek? And you really want to do this? And-- Hey, I get the snarky. Yeah. And I said, yes, this is what I want to do. And it created a whole bunch of requirements and user interface guidelines for doing just that. But yeah, it's very first response was really-- Well, it's just kind of interesting what you said. Programming, in my world years and years and years and years ago, computers only do what you tell them to do. They don't do what you want them to do. So you got to cook that shaft for putting in those ingredients. You get horrible ingredients. It's going to make a horrible-- you're dead right. If it gave a good ingredient, it said, I wonder what it would have come up with. We could have gone through a database that what do we normally do with chicken and tomato sauce and et cetera, yada yada. It's very interesting that machines only do what you tell them. They will not do what you want. AI potentially could bridge that gap and parse out what you actually want. And I think the really snarkiness, that snarkiness, is kind of the next grade, isn't it? It's just fascinating to see those examples. For me, anyway, I see the huge benefit in my real world is in how it can get huge bowls of data and cross compare. I hear a great example. You look at compounds and you look at, like, contraindications and you look at compounds, reactival compounds in a patient's body. So the huge amount of data that you need to go through, and AI could do that so readily. We see that before we read a paper, it's their last year around five-moira algebra and how it reacts with certain compounds. And in a literature search, no one found this. AI found this particularly in one paper. We yes, that it does react with these certain ACE inhibitors and it does give us this reaction. One paper. It was fascinating. I think AI is great. I just did a lit review for a paper that I'm writing and we use AI for that lit review. And it literally did it for us in five minutes where we would have taken us how long. It literally did it for us in moments. And it came out with the right, I mean, so I see the value with you using large data sets. I do see the huge value. There's a huge efficiency component to what is capable of doing in your example of the drug interaction is one of those. I as a human cannot keep up with the literature or what's published out there. Even if I could read everything, I'm not going to retain everything. And the computer and the technology has that capability. So marrying these things together to make our lives easier, it's here today. It's capable of doing that today. What's the future? If I gave you a mind to go on right now, what do you see the future of AI in digital health for patients? I've been quoting Bill Gates a lot. And Bill Gates once said, as humans, we tend to overestimate what's possible in two years and underestimate in 10 years. I don't think we truly know what this is going to do in the future at the macro.
level. Like this is game changing. I think of it as the printing press, electricity, computers, smartphones, the internet. These things have been game changing technologies to society. They've all along the way came with their challenges. The new challenge of AI is it's happening much faster. So we're really struggling with adapting to it at the quick rate it's coming. But when the printing press came around and it became much easier to read books, people were fearful of reading books. And we'll we ever be able to remember anything anymore because there's books. Well, that hasn't happened. Calculators. Oh, we never have to teach math. When we still teach math, we've changed how we teach math. And so now the AI is going to change what we're doing. But it is, where is this going? I wish I had that crystal ball to be able to say, I don't think I really know where it is going. I'd love to think that it's going to make our lives easier. The thing Bill Gates recently said a couple of weeks ago is we're looking at that 10 year timeframe, but two day work week. How can we make an actual difference in the world from that perspective of not having to work five days, seven days a week? One of the interesting things about what AI is doing is it's taking that and what is really good at today is those creativity type of elements. These are the things I as a human want to do. I want it to go off and do the dishes. So I can go and play with the kids in the back yard kind of thing. Oh, it's not quite there. But yeah, I think as we become more and more familiar with it. And as people start using an experiment with it and they are going through the school system now, we are going to realize all sorts of new potential, new ideas, new ways of explaining things that we could never think of before. One of those examples too, in terms of a clinical trial and understanding symptoms you don't think about. I saw something where typing on a keyboard, if that skill degrades over time, I'm not going to recognize it. You looking at me aren't going to recognize it, but the computer can recognize it over time, your ability to type in the same fashion and degrading, which could be an indicator of Parkinson's disease. So how can we use this technology to find those hidden signals to be able to enter in sooner and make more of a difference? Everything, I guess the biggest thing A.I. is going to accelerate everything. Everything. Yeah. That's the challenge of it. Everything we do is going to be accelerated. And that's the biggest challenge. The biggest benefit and the potential of the biggest fallback. Interesting. I just hope that it doesn't do what the next day I'll say it's going to do and destroy the world. I don't think I do not. Anyway, Robert, thank you so much for that. I really enjoyed this. Today I was talking to RJ Kizorra from A Standout Solutions. He's a co-founder and CEO. I'm Mark Wade, TransPurple of Life Sciences. Please join us for the next episode. If you want to be on the podcast, DM me or email
[email protected]. Thanks a lot. Bye-bye. Thanks, RJ.
PodcastTranscript AI