Latest / The Unpack Show with Anandh / EP-7: Senthil Kumaran - CIO,MNGI Digestive Health
Transcript
- speaker-0: you speaker-1: Hey, in today's episode we are going to talk about the hottest topic that everyone is talking about today. I have a special guest with me. I've known him for years. He's a good friend of mine. It's going to be an interesting conversation because I love how he presents at the conference and the way he delivers the speeches. So it's going to be exciting for me. I hope it will be exciting for you as well. So the topic for today is unpacking the human plus AI collaboration, right? How AI is acting as a copilot. I think we've been talking about, we imagined the AI is going to replace humans, but the story is unfolding in a way today is horror. more human, right? AI is not taking over. It is teaming up with human, right? That is how I see it. We are entering into an interesting era where AI acts as a co-pilot. It's not the co-pilot that you're thinking of, but in a co-pilot in many ways, like, you know, it's a personal assistant, personal co-pilot, a partner that amplifies human creativity, if you will, right? Speed ups decision-making process, removes repetitive work that you are doing right now so that you can focus on the high value thinking and thought process work. So this collaboration is making people to be more effective and efficient. So we're going to unpack this whole standpoint of AI as a co-pilot. So in this episode, we're going to explore what real collaboration between humans and AI looks like, the wins, the tensions, the lessons learned from leaders who are navigating the ship, right? I have discussed Senthil. I want him to introduce himself and then we can go into the topics. ⁓ speaker-0: Thank you so much Alan, it's a pleasure to be here. Thank you for posting me. About me, I've been in Twin Cities for more than 27-28 years now. I've been in healthcare technology for the past two decades. I started my career as a software engineer, did a bunch of websites, right from deluxe checks to Colson Wagon, the travel mini.coms in Twin Cities, my stamping on it. Then I shifted to healthcare. was part of telehealth company like... Co-founded, I call myself the technical co-founder. So was the CTO of the company for the past 15 years and for the past six months I'm at Minji, Minji Minnesota Gastroenterology at the law. This private GI company and as I'm the CIO for Minji. So most of my examples today will be healthcare. Unfortunately, may I, we'll talk excited to be here man. Thank you. speaker-1: Yeah, thank you. Thank you for the introduction. Welcome to the show. Yeah. It will be interesting to relate AI to healthcare, right? It will be interesting and looking forward to it. I want to dive into this concept of AI being teammate or a partner center with you. So for a long time, the conversation around technology is focused heavily on automation, right? How machines could take over tasks to things, to do things faster or more efficient. But the story is evolving right now. It's not just about automation, right? You're not just looking for a repetitive task and automate the repetitive task. It's about collaboration. It's about Asking someone, it's like talking to someone. AI is starting to feel less like a tool we use than more like a teammate right now, right? So I use it every day. I'm sure you're using it every day. We stopped going to multiple places instead of like, know, going to multiple places, we are going to one place. So a co-pilot is a classic example that helps us to think, create and make decisions. I'm curious to know, how do you personally think about AI as a co-pilot in your own work or personally shaping up yourself as a person? speaker-0: makes life lot easier. Isn't it? wish we had these kind of tools 20, 25 years ago when we are struggling. This kind of liberates, isn't it? Like it becomes conversation. Let's not simple search. Perplexity, for example, it's verified and you can take it accountable to it. It's more like your research assistant and it's a partner. I think that's the right word today. It's definitely an amplifier is how I call it. Each tool sits on pluses and minuses. It's, I always bring this example. If you're a carpenter, 50 years ago, you had one saw. And at the manual saw, today you have circular saw, miter saw, jig saw. There are like 70, 80 different saws, but obviously you're not using all the 70 for a task. You probably use the five or six. You should know which one to use to do which task. And it makes your life a lot, I mean, it makes it a lot faster. can get to your destination a lot more ease. That's, that's how I look at it. And it's, definitely a partner. I'll give you an example in healthcare. instance, company I'm working for right now, we are using an ambient AI tool. It listens into the patient position. Transcription is 10 years old. Now it does like in many different languages, the transcribing part is where it started, but now it's moving from transcription to creating notes based on my own notes. So physicians train this tool and then it learns from their notes and it starts to create the patient notes. What the treatment plan is not looked like, what are the orders that it needs to generate? So it saves them a lot of time. This is one tool which amplifies their productivity instead of spending like 20 minutes after every visit are 15 minutes in documenting. This is what we do at our telehealth company at the beginning of time, 2010, when we launched this telehealth company. Our goal was to minimize our physician, patient interaction and make it lot easier for our physicians to make decisions, to make the treatment plan with a higher degree of accuracy. And it did definitely work, but it took some years. Today you can like... from one place to other in like a minute or two. Yeah. That's how I look at it. It is happening. speaker-1: Yeah, it is true. think that's a classy example. Like every one of us would have had this real war moment, right? Like, you know, we might be thinking something cool, but it might be like, you know, giving you surprises, right? For sure. Everyone, I'm sure everyone had their war moment. Like for me, the war moment was when I was like thinking about starting my own podcast. I was like thinking about, okay, I don't have time to do this editing. I don't have time to like select on this topics and you know, how do I cut it? How do I make the sound more elegant or the video more crisp right and I found this tool that lot of tools available in the market and the tools are like helping me in a tremendous way all I need to do is just spend like 15 minutes of time right and you can sample it right that was a warm moment for me right so you don't need to spend too much time in doing editing like wise yeah we can try speaker-0: translate this whole audio podcast into 80 different languages in the next eight minutes. You can live stream it in like 10 languages. It is all possible today. And we can hyper personalize the whole podcast after. So there's a lot of volume. Eight, nine years ago, was at a conference where this presenter was presenting about the AI tool to predict diabetic retinopathy and the FDA approved it. This is not even augmented AI. This is like a complete autonomous AI solution. So that was the one moment like this, was before pandemic. And said this, we are proving an algorithm, takes months and years. And said this company was. talking about, we've got the autonomous AI. So autonomous in the sense for, for viewers who are viewing and you don't even have to see a physician and ophthalmologist and like five other people before you get your treatment plan prescription for diabetic retinopathy. This one medical device does this for them. It just kind of the barriers and the bait time and also makes it available at a lot cheaper cost for everybody. That's where we're towards. speaker-1: Bye. with Comet browser from Proplexity is making my life easier much more better. I don't even log in at 10 a.m. to post exactly like I just need to schedule it and I need to tell Comet that hey, post this video with this context, with this message at this time. Boom, right? happens. I want to probably shift gears. This is an interesting topic. know you talked about this in one of the speaking opportunity that you had. think AI today can do so many things. we once thought what humans could do like it could be like recognizing patterns, making predictions or even generating text, right? But there's something unique that human only can do but cannot do which is like the empathy, emotional intelligence, intuition and deeper understanding, right? And then how you can relate to that. I remember a phrase, I don't know where it's from, but it said AI can tell you what the data says, but only humans can feel what that data means, right? Yes, that's stuck in my mind. I'm curious, Santala, you you teach AI stuff to students and you are implementing it in your work. What do you think the shift is going? Like human and AI working closely with each other. What do you think or what do believe the humans will continue to do the best that you're working with doctors, you're working with students, even as AI keeps getting smarter and smarter every day? speaker-0: Correct. Correct. First of all, this is what we need to remember. Repetitive tasks, you leave it to AI. That automation is a lot easier, isn't it? So, so that you can redeploy human to higher value. I personally believe goal definition is where humans will excel. Like you maybe do the strategic guidance to this tool. You will probably course them and you put the ethical and as you said, what does it really mean? And then. do the judgment arbitration. So all of that is us as humans, are going to, but we leave these conversational and verifiable answers part, all of that to some kind of an AI engine. That's how I believe the future is going to be. It's a complex analysis, which we cannot even do exponential calculations. that for many of us calculus is super complex. So leave all of that, like the complex design and just do the higher order thinking. I believe that teams which excel in taking AI as a partner is going to succeed in the future. That's how I look at it. Like whoever is not, this is the same as like 25 years ago. I'll give you an example. when internet just started to come about, every company wanted to get their dot-com presence. Senate said this is year 97, 98. Some of my buddies who were working in power builder in this team, they were talking about, my God, internet is just a fad. We are printing paper. selling checks and all that. Why do we need to be in a .com? Now, if you don't have a mobile app or a mobile friendly website, you're out of the game. That's like 2011, 12 times after iPhone 2, 3 came out. Today that is shifting to that. you've got to constantly evolve and just move on. speaker-1: Yeah, that's very true. You have to probably assess and adjust and adopt, right? Those are the things that you need to do as a human being. Yeah. I wanted to switch gears on- speaker-0: are going to stay, isn't it? That's why Netflix is here. Netflix approached Blockbuster three times. I'm sure you've heard this story many times. You know where Blockbuster and Sears and Aces are. it's whoever are adaptable. So next, next gen, whoever is going to be adaptable to this AI. speaker-1: Right. Yes, yes. know we talked about this when we met, when we talked last time, like, you okay, we should probably find some time to constantly keep ourselves up to date, right? It's changing every minute, every second, right? It's hard to keep up with. Correct, correct. Yeah. Yeah, yeah, speaker-0: Imagine this, I want to throw another medical exam, healthcare example here. I was at a conference recently and this doc was presenting like from 90s to 2000. It took 50 years for the medical knowledge to double. Okay. So published art, then I will paint. And then from 2000 to 2010, it doubled in 10 years. Can you guess like what's the doubling period right now? Just like what I was thinking. I was thinking like maybe in a year, but right now the day is like 74 days. every 74 days. speaker-1: Within months? Within months? speaker-0: So imagine if you're publishing a book today, it's already out by the time it gets published and gets out. So we actually, at our current company, we work with open evidence. That's like perplexity, but medical field. Okay. So open evidence. There is another one called Doximity, but most of our docs use open evidence. And that brings this, all these behind the paywall medical knowledge into like one place. It's like walking into a buffet and you get all. 100 or 200 items and all of it is given to you in like no time. And it's all verifiable. All the links are there. So our folks use it for creating the notes and giving it to the health plan. It's like, for instance, you want the prayer out for a complex procedure for a patient. You want all the bulleted points, but you also want all these links where you got this information from. Imagine this, took like five, 10, verse 20 hours to create this thing before. Now it's like. five, 10 minutes. That's why I'm saying it's a multi- speaker-1: I think it's freeing up lot of the time where we are spending most of our time with the repetitive tasks. you know, I can't wait for a day where you just like send only commands, right? To the missions and everything will be done. Right. I think we are, getting there. I want to shift gears a little bit. central to say, we talked about all this goodness, right? It's evolving. There's a lot of new stuff that are coming into play, but sometimes it feels like it is a little unpredictable, right? Because you have to really have human in the loop to understand the context, right? Not every time it's going to give the right response, right? There's a lot of interpretation that it does, right? Because of the prompts that you use, because of the way you ask questions. You know, you are in a healthcare space, right? What is this partnership with AI looks like for you when it when it's not going as expected because it's a life or a death situation, right? speaker-0: Yeah. So the way I look at it as yes, three, four years ago, even like look at before coming to the healthcare field, I GPT one, which is like what five years ago or four years ago, they are on the test. Now it tells you that I don't know the answer to this four years ago. will hallucinate or even a couple of years ago, will hallucinate and make up things. And today that the training has changed. So the model training has changed. So it doesn't know it now comes back and says like I am. going to disappoint you. don't know the answer. So I'll go very fairly. This is a phenomenal step that the AI is making. So that's why we use open evidence. As I mentioned earlier, the healthcare example. Yes, it's a research assistant has a lot of verifiable answers, but we can also ask our doctors before you publish this, you just read through, you have to spend your time, but to collect everything, to put it in one place. This is like the taste testing going back to the Dubai. Airport example, like, okay, you have all the 200 plus minutes, you know, which one is going to work for you. You will just taste it a little bit and then, okay, this is what I'm going to put it in my plate. So that's what you need to do as a human. You're not going to take all the 200 minutes and then you visit us in a GI clinic. So I don't want that. So the AI definitely is getting better and it's, making these very favorable answers. It's not dominating the stock. Unfortunately, that's what people relate to, but over the years, I think this will get better. Right. speaker-1: Right. Yeah. I think looking at it, like you said, it's going to get much better and better and we can see that like every second every minute, every week, right? think going back to your examples about like 74 days, like that's fascinating to me. But when you look ahead, right, probably two years from now, five years from now, I think it's hard to predict. I was talking to Manjit on a similar topic, right? Yeah, How AI is going to help in the... universities right you know he quote a lot of examples I know you teach students too but I think the one thing is very clear right the relationship between human and AI is going to get closer and closer right I think it's not going to change I think we are moving away from using these tools as an experimentation to kind of like using this as part of our day-to-day life right what do you think is going to happen right you know how is this human plus AI relationship is going to evolve in next two to three or five years speaker-0: It's going to get even closer. think for folks who don't want to listen to that, maybe you got to go back and read. I've been talking to our students. The goal is be adaptable, be flexible. I'll give you another example. Like 10 years ago, FDA approval for AI algorithms was like, 10 a year, today it's exponentially growing. Last year alone was like 500 plus. So it's growing, isn't it? So this is mostly augmented AI solutions on the radiology and pathology, but it's also on cardiology, ophthalmology and neurology. So all fields are getting disrupted in healthcare, which is the super regulated industry. So the example why I'm talking about healthcare is if it's like the most complex regulated industry is getting some AI implementation. It's already happening. Robotic surgery is already happening for the past decade or so. It's going to change and you will be partnering with them. All I'm saying is you as a human, if you are adaptable and if you augment this AI tool skill with you, then it makes you much more employable. Yes. It's like, okay, I'm not going to use calculator in medicine. Today, I don't know whether you need to know all these complex math problems. It's like, you just pick up your phone and do all of these calculations a lot easier. This is back 1970s about like, the biggest job at that time was what? Telex operator or FAS email operator or something like that. A hundred thousand plus jobs today, obviously, that is not up to it. So we do not know what is going to happen in the next. five years on the new jobs. So maybe it's an AI augmenter. There might be some new jobs which are going to come about, which we do not know. And that's where we are marching towards. And people are all worried about, I know even at our company, we always talk about, ⁓ my God, is my job going to be a place? No. As long as we are growing, we make it operationally efficient. We can move into this higher order tasks. are like a lot of fish to fry and we're like working on repetitive tasks. speaker-1: Yes, yes, yeah, yeah, I think I can imagine it. Good, good. speaker-0: Yeah, maybe a decade or maybe five years. Who knows? It's changing. speaker-1: I think I can imagine you will have like a teammate whom you're going to talk to, right? For simple, maybe like code approval, right? It could be anything. Or can he schedule this meeting, right? It is on, you have. speaker-0: happening on it. The Bincode tools are going to be your assistant, co-pilot. Like imagine life without co-pilot inside Outlook now, or the Gmail automatic responses that it automatically generates, it's a gemini integration. makes, I personally feel it saves a lot of time. I always think about, wish I had five years ago or 10 years ago. Wish I hadn't dictated things like 10 years ago, I would not have had the carpal tunnel. That's what I always think. speaker-1: I feel like it's a luxury, right? It's a luxury, but so that we can focus on the high valued items. Again, thank you for taking the time. I know we can talk about this topic for hours and hours. I know we've been talking about it. I've been bugging you lately. Like, you hey, what do we do this? What if scenarios, right? It's always interesting to talk about someone who's actually implementing it, right? Especially in the healthcare space. Thank you so much for taking the time. Thanks for being part of the show, Santosh. speaker-0: Thank you so much for having me Anand. Thank you so much. Wish you good luck on this show. I hope your audience base increases in the next few weeks. Hopefully in AI algorithm that you find R2 the main goal is a lot easier. Good luck. Thank you so much for having me. speaker-1: Thank you.