Dr. Mittinty Explores Biostatistics, AI, and Mathematics in Healthcare on Indian Community #19

Exploring Biostatistics, AI in Healthcare, and the Power of Mathematics with Dr. Moorthy Mittinty

In this episode of the Indian Community Podcast, hosts Rahul Mehra and Amit Gupta interview Dr. Murthy Mittinty, an esteemed associate professor in Biostatistics at Flinders University, Australia. Dr. Mittinty discusses his journey from a curiosity in mathematics to pioneering research in biostatistics and its applications in health data science, including oral health and pain medicine. He shares insights into the importance of asking the right research questions, the role of AI and machine learning in modern healthcare, and the ethics of data usage. Dr. Mittinty also emphasizes the value of interdisciplinary learning, the language of mathematics in interpreting the world, and his belief in contributing back to the community. The conversation illuminates the critical aspects of biostatistics, causal inference, and the future of AI in healthcare.

00:00 Unlocking the Mysteries of Math and Its Real-World Applications
00:32 Diving Into the World of AI and Machine Learning
00:58 Spotlight on Dr. Murthy Mittinty: A Journey Through Biostatistics
03:18 The Serendipitous Path to Biostatistics
11:32 Exploring the Depths of Causal Inference in Health
19:55 Navigating the Complexities of Response Bias and Uncertainty
25:17 Unveiling the Truth: The Complex World of Data Analysis
26:04 Exploring the Intricacies of Oral Health and Pain Medicine
30:25 The Real-World Impact of Biostatistical Research
34:58 Navigating the Ethical Landscape of AI and Machine Learning in Health
45:08 Seeking the Ideal Student: Beyond Academics
48:51 Community Engagement and the Value of Giving Back

Join and Connect with us:
Indian Community: https://connect.indian.community/
Instagram: https://www.instagram.com/bharatiya.community/
Facebook: https://www.facebook.com/bharatiyacommunity

Dr. Murthy Mittinty at Flinders University : https://www.flinders.edu.au/people/murthy.mittinty

Some Keywords:
Murthy mittinty interview on indian community,
flinders university murthy mittinty,
Murthy mittinty,
new podcast on indian community
murthy mittinty podcast with amit gupta and rahul mehra
biostatistics professor murthy mittinty
mathematics application in real life
biostatistics
biostatistics professor interview
indian professor living in austraila

Transcript

Amit Gupta: 0:00

one thing that I really liked about this is understanding the application of math.

Murthy: 0:05

that question is still valid even today for me. Where is this used? How is this used? But I'll tell you my experience and where it cracked. if you're a carpenter, you use maths. That maths is very basic maths. maths is taught in a incomplete manner.

Amit Gupta: 0:24

I'm sure after listening to this explanation there's going to be a very different perception about mathematics and statistics the most buzzword in the industry today is AI and machine learning. Almost every sector or every aspect of life has been touched by

Murthy: 0:43

AI is tremendously useful in the digital age.

Rahul Mehra: 0:48

For the students that we would like to know top three things.

Amit Gupta: 0:50

beyond academics, what are those traits that you would look for?

Murthy: 0:55

Academics are secondary for me

Track 1: 0:58

Welcome to the Indian Community Podcast I am Rahul Mehra and with me is my co host Amit Gupta. And today we are thrilled to spotlight Dr. Moorthy Mittinty, an esteemed associate professor in biostatistics at Flinders University, Australia. With a rich academic lineage from Agra University to the University of Canterbury. In an impressive journey through prestigious research positions across New Zealand and Australia, Dr. Mittinty stands at the forefront of biostatistics and health data science. Dr. Moorthy has a trail of accolades, including the prestigious President's Award from the Statistical Society of Australia. Dr Murthy stands at the forefront of research. Empowers transparent and replicable science. His work, which spans from mediation analysis to machine learning applications in health showcases his commitment to advancing statistical methods for better health outcomes. An inspiring mentor and educator, Dr Mittinty is not only shaping the future of health data analysis, but also guiding the next generation of scholars in their pursuit for impactful research.

Rahul Mehra: 2:19

Welcome to the show.

Murthy: 2:19

Dr. Murthy Thank you very much. Rahul. Thanks for the generous introduction. And I think maybe you have done a lot of research on the trail of my website and the moments. Oh yeah. We are

Amit Gupta: 2:35

like the CIDs.

Murthy: 2:36

No it's been a long journey for many years. And yeah, I'm very happy to have come to this position today. Where I am, and and the enriching learning and knowledgeable trail of research as well as students meeting different students and You know, varied opportunities that I got on the way. I'm very thankful for where I am. And I'm also very thankful for my mentors and students, basically, who have taught me more than mentors.

Rahul Mehra: 3:16

Welcome, sir. Welcome. Dr. Moorthy, let us start from how did You know, you have done your mathematics and statistics, you know, you were a mathematics and statistics wizard. And how did you get introduced to

Murthy: 3:29

biostatistics? Yeah, that's a very serendipitous moment when I got introduced to biostatistics because even though I had a deep interest in math, I always was curious about, you know, where is this applied? And how is this applied and how is this useful for real world? Because that was not clear to me. Like, sometimes you have this puzzle inside your head. Why am I studying this? What is the use of this? What is the purpose of this? And then when I came to year three of my undergraduate. I saw some applied statistics like time series analysis or sampling survey sampling and agricultural bias, agricultural statistics. So when I started to read something about agricultural statistics, then I thought, like, oh, wow. This is applied in agriculture and quantification and experimentation for growing you know, good quality rice or wheat, then that must, then there should be certainly, there should be some applications of, This in health and human population. So that led me to you know, my masters and during masters again got introduced to some very generous and knowledgeable people. Who have done work in demography and being educated at International Institute for Population Science in Bombay. So I, you know, being inquisitive, I asked them what would be the procedures to get into these top institutes in India. And I was lucky to get the UN fellowship to get there and then go to IAPS in Mumbai. And it is there I found one excellent professor. Professor Arvind Pandey and he was the one who taught us a subject like a topic within the demographic studies on biostatistics. And when I was reading this topic. You know, it still was unclear, but like there was one one chapter in that topic, which is called life table analysis. And that really opened my eyes and that really showed how statistics, how math. Is used for improving health and that introduced me to cause specific that all this kind of you know, diseases and how are diseases measured? How, like how does statistic help understand? the important predictors for various diseases and things like that. So that led me to you know, biostatistics and then working with the Tata Cancer Memorial Research on cancer research that further enhanced my inquisitiveness to work more in this area. So that is the short answer to you know, to a wonderful question of what is the journey to biostatistics. Thank

Rahul Mehra: 6:36

you, sir. Thank you. That was really a long journey. You know, a lot of thoughts coming. And,

Amit Gupta: 6:42

you know, one thing that I really liked about this is understanding the application of math. Just this afternoon my, my daughter was describing a math question about you know, how do you measure the area of a circle inside of a triangle? And she asked me, Dad, where will I ever use this? Did not know the answer I told your dad is not a mathematician you know, we'll have to figure this out. Maybe ask this question to your teacher at school. But I, in after a few minutes, I was thinking about that. And I was like, this is important for students to really understand. What we are studying in how it's going to be applied. And once you are able to connect that dot, that's when you will really be able to appreciate that knowledge that you're receiving.

Murthy: 7:36

Absolutely. Because I'll tell you for sure, out of my this, you know, 25 years of experience, we study undergraduate degree. We can't correct that question. We study masters. We can't correct that question. We do PhD. We can't correct that question. It's only when you are teaching and even then when you're teaching, you should have a little bit of thinking outside the box. And that question is still valid even today for me. Where is this used? How is this used? But for me, I'll tell you my experience and where it cracked. I was teaching the biostatistics collaboration in Australia. I was teaching into that program. And then when I started looking at all this You know, when I started teaching the mathematics again, the calculus part to biostatisticians, that is where I figured out, like, why is this useful? How is this useful? It's because you know, when we measure, like, if you're a carpenter, you use maths. That maths is very basic maths. Because that's counting, measuring, and that's all. But when you have to predict something, when you have to you know, estimate something, then the concepts are completely different to measuring things. You know, you knew you need tools that would allow you to predict or that would allow you to estimate something. When you want to measure, you just need a scale. But when you want to predict and estimate, that is where maths comes into play. Because if you look at the history of development of maths, Right from in the 3rd century BC, when people were stargazing, all they did was collect that data, and they were not interested in prediction or estimation, but they were just collecting the trail of the planetary movement. And they enjoyed that, but it is later on when they started to think, how can I make a prediction of a lunar eclipse? How can I make something? That is when maths grew and maths became more and more abstract. So the, again, the answer to your question is when you are interested in prediction and estimation, That is where maths learning comes to play, but actually maths is taught in a incomplete manner. When I say incomplete manner, they teach mathematics as if it is a tool. Right. No, it is not a tool. It is a language. It is the language of creativity. It is the language of abstractness. It is the language of uncertainty. So when you want to measure uncertainty, when you want to think creatively, this is the tool that allows you to do that. But it is actually a language. It is a language of uncertainty. Patterns that occur in nature. It is a language of patterns that occur in behaviors. It is a language of patterns that occur in every randomness. That is beautiful. In

Amit Gupta: 11:07

fact, a lot of us don't look at math. In that way, so I'm sure after listening to this explanation there's going to be a very different perception about mathematics and statistics at least for layman like me and speaking about layman I noticed that Your research interests are across a wide range of different subjects, like missing data analysis, machine learning. And there was one term that I was very curious about, and that was causal inference. Right? And because understand it, but I still don't understand it, right? So I understand, okay, there is cause and then there's inference, but when these come together and what is biostatistics got to do with this and how is that, changing the world, if I may say that. So how, and how does it apply to our daily life? So could you shed some light on, on that subject of causal inference and maybe some examples

Murthy: 12:03

that we can relate with? Again, this is a very million dollar question you have asked me because people need to know this. This is a very fundamental thing. Causal inference we are very familiar with in our daily life. For example, we always ask a question after something has happened to us, or had I done this in a different way? You know, maybe the result would be different. Right. What exactly is that? That is counterfactual thinking. Okay, counter to your fact, your thinking. But we can't go back in time and change any of these things. Because we don't have that time machine to go back in time. So, now let's translate this to health data. Let's translate this to, you know, inference or whatever that we want to make a decision on a health policy relevant question, let's say that. So, let's say I want to answer the question that whether if I take you know, a Neuropen or a Panadol. Would it reduce my headache? So that is the question that we want to answer. So usually when you do a trial on people, you give a drug to a few people, you don't give a drug to a few people, and then you say, okay, all these had headache and you gave a drug and you didn't give a drug and you control the whole environment. Other. the noise from the other aspects. And then what you test is whether the drug worked or not. So if in those patients who, whom you gave the drug, you can see some people took that drug and then for them, the headache reduced or they don't suffer from headache any longer. So that is our causal inference because that's what you said. Okay. People have a headache. So the outcome should be headache should reduce after you give a drug. Yeah, so that is the cause and effect thing that we are talking about. But now this is a controlled trial that we are doing. So you can restrict the environment and you can keep the populations, which are exchangeable in the sense like all people have a disease. And you give a treatment to some, you don't give a treatment to some, and then you observe what is happening. So once you get the result that people have taken the drug, people adhere to your restrictions, and then you can see the drug is cured by giving the disease is cured by giving a drug. So that means for those people who did not get the drug, now you can give the drug and you can see a reduction in the disease. So that is your causal inference for reducing the drug by giving some intervention. So you can do this in a randomized trial. Or you can do this in observational data. Observational data in the sense that people use or people collect information for administrative purposes. So there is no Randomized controlled trial per se. So you collect information, you go to a hospital, they get your vitals, in a sense, like you get your age, gender you know, education, blah, blah, blah. And then they have information on your disease. And then they give you some drug, which is tested and trialed. And now you have an outcome. But how do I replicate or how do I emulate a trial in this observational data to see similar effects? Okay. So then we use the statistical model to create two different scenarios. One is a factual what you have observed and another one is a counterfactual where everyone in the population is not given a drug. And the second option is where everyone in the population is given a drug. By using that observational data, we are creating or we are emulating these two scenarios, counterfactual scenarios. So that data then allows us to make the causal inference, which is same as a randomized trial. So that is what is all about causal inference is being able to describe a phenomena when you know what is the intervention and what is the outcome that you are changing. What is the intervention that you're giving and what is the outcome that we are changing? So studying about these concepts by making use of observational data or trial data and making use of maths for making all this counterfactual predictions. So that is the whole study of causal inference. So to answer your other question, what does this allow us to do? The common thing that you see in health science research is people don't ask proper research questions. So when you don't ask proper research question, then you don't know what is your inference. For example, if I say cigarette smoking, then you can see cigarette smokers. You can see three groups of cigarette smokers. One who had ever smoked. One who had never smoked and the other, the third one is currently smoking. So when I want to make an inference, I want to ask the question in a way that it suits all these three or any two or any one population. But I need to know for whom I am doing a research, for whom I am developing a policy. Okay, so is it for ever smokers or is it for current smokers or is it for never smokers? So if it is for never smokers, then who is my comparator group? I need to know all these aspects. So when I'm asking a question. And then doing an inference and estimating the risks of either smoking on diabetes or smoking on you know, cardiovascular disease. I need to know for whom is this risk being estimated? For example, am I estimating the risk of cardiovascular disease among currently smokers? Am I asking you know, what is the risk of cardiovascular disease among ever smokers? Or am I asking a question on cardiovascular disease in never smokers? I need to know these aspects. So what the causal inference has given us is a tool to ask these sorts of research questions. appropriately and then walk from the question to the model to the inference and backwards. So that is this whole new theory of causal inference that has gained a lot of popularity in the last 10 years. So it's

Amit Gupta: 19:24

a, it's a great explanation, Dr. Murthy in, especially when it comes to health and You know, even if you take that particular example that you just mentioned. So I come from a customer experience background where we perform surveys and ask people about their experiences, where again, this is a lot of statistics involved in terms of metrics such as net promoter score, and you look at the customer effort score and those kind of, you know, KPIs. Now, the one thing that troubles. An experienced designer is a response bias, right, where the respondents are, you know, just. Lying or, you know, they're not giving you the right accurate information. And how do you, address that in health? Because there it's the stakes are much higher compared to a marketing or an advertising survey here your policy and decisions could sway. So how do you eliminate or maybe adjust for those

Murthy: 20:30

response biases? Yeah, that's a really good question because like, this response bias is not only for health, like the social surveys, the marketing survey, they have huge repercussions because of the cost involved with the bias. So, like, when it comes to health, it is more because it's more matter of life and death in some instances, or some diseases. So, what we do is, again, like, of course, there are so many aspects, because, you know, This is again the current research that is mostly being conducted nowadays where the importance is given to these sorts of response recall bias, response bias. And measurement error. So this is a response bias is also called as measurement error because you know when you play or misclassification error, especially when you have binary data or multi dots like multivalued data, not the continuous data, but multivalued like, for example, very satisfied, not satisfied and neutral. That is a multivalued response. Okay. So when you have multivalued or binary like yes and no, then what you have is a misclassification error. So there are techniques for handling, but again, these are, once you do analysis using the observed data, what you can then do is some sort of sensitivity analysis. around the estimate that you're making. So those types of analysis, again, these are all counterfactual scenarios that you're creating, or you can even say counterfactual or hypothetical scenarios that you're creating. That means if the proportions in these groups are shifted by 10 percent up and above, like for example, yes, no, if yeses are 10 percent more than noes, How would my response or how would my prediction change? How much is the difference in my prediction? So you do this kind of measurement error, misclassification error analysis, and then you can account, you can't account, or you can't teaser, but like you can do a sensitivity, how far your estimate would be stretched because of the uncertainties that you have. in the response. So this is all the language of uncertainty. And so, you know, because there's a whole theory on uncertainty analysis. That is where people are giving more and more focus nowadays because you know, especially in health or even in marketing. When you see Amazon or when you see Google, when they do their predictions, they're accounting for all of these these sorts of misclassification and things so that they can create and they can create the whole not giving you a point estimate, but giving you a whole interval estimate where the. You know where the freedom for movement is much wider than just giving you a point estimate where you can move a little bit in your uncertainty of the estimation. Got it. Thank you. Yeah, I think

Amit Gupta: 23:43

that's that's definitely helpful to know that you're also factoring for that because that creeps in and I think there's also a challenge of your sample size, right? So you don't get to study. Too many people are not a lot of people are willing to try new drugs. So you would definitely want to factor in all these different scenarios and the permutations and combinations that can start influencing the

Murthy: 24:11

final result. Yes, of course, because you know, there's a lot of uncertainty due to unmeasured variables. You can't measure everything under the sun, so you can only measure something. So you're making an educated guess, if you want to say. It's the game of educated guess. So how good is your educated guess is based on how good is the information that you collect. Okay. And how good is the response or the information that is inside there, as you highlighted the response bias, how good or how truthful is your respondent in giving that information to you, based on that, like, if it is not, if every aspect is not a measured thing, even when you measure the machines can go wrong, like, for example, your weighing machine. Your blood pressure machine, the blood pressure machines, they don't show you one constant blood pressure measurement. Every time you measure, they give you three or four different measurements. Why? You know, your weighing machine, it lies to you in the first instance. Yeah, my weighing machine lies to me all the time. So you got to take multiple readings. So these are all like, you know, this is the knowledge that you have to bring. When you are analyzing data, the data should not be blindly analyzed using one observation or one one survey, but you've got to repeatedly do this. And find the similar effect in repeated studies and find similar component or similar cause in repeated studies to actually make it as a cause of the disease or cause of a of an

Rahul Mehra: 26:02

aspect. Right. Dr. Moorthy you know, you have done considerable amount of research on oral health, pain medicine, also diabetic research. Which of these topics did you find most challenging? And also, you know, I would say satisfying to undertake.

Murthy: 26:20

The satisfying is oral research, I should say. Even like even for the sake pain medicine, because the challenges in both these are almost same. Like for example, we don't take the mouth very seriously, you know, of course, we as kids growing up, our parents said, brush your teeth. We didn't know why are they asking us to brush teeth? All they said is, you know, you have teeth. You have you have you know, little you know, bacteria in your gums. So just, you know, wash your teeth or brush your teeth properly. But we ate Colgate paste and then we just spat it out. That's all we did. We didn't wash and brush our teeth properly. But the first introduction of any aspect inside your body is through your mouth. When I realized that like, Oh wow, this is truly insightful because the food that you put in your mouth is what makes you. The food what we eat, but the first entrance for the food to move, you know, move till your gut is through your mouth. And the mouth is such a wonderful complex system in itself. because you have teeth and all teeth are not same. Some are gum, some are molar teeth, some are, you know, the complex structure of them and the root canal and there's the bacteria and there's a fungus and there's a whole microbiome structure that is different to your other parts of your body. And these microbiome interact with your upper body and also the lower body. And so that was like, Oh, wow. So there's this complex colonies of microbiome that sit in, you know, inside your jaw and inside your mouth, and they interact and they dictate how much of the gut microbiome is interacted, and then they interact, the gut microbiome then interacts with your brain microbiome. Which is entirely different, which is called the gut brain axis. If you google gut brain axis, you can see that. So when I started working with oral health people When I was looking at the structure of the teeth, that showed me how complex the teeth is and how complex the information that we collect on dental or oral health is. It is very, like it is truly complex. And to address that, you need proper statistical methods. Of course, there are methods that are available, but you still need you know, some new techniques. And the sampling design for mouth studies is completely different to that of a regular randomized controlled trial. So this structural differences within the general disease analysis and oral health, that has interested me tremendously because of the complexity of the mouth and also the insights that I had. After working with these people is really helped me actually grow as the person and take care of my own oral health. That was

Rahul Mehra: 29:50

very informative. And actually, you know, speaking to you now, my mind is also going in the direction of statistics. And what I mean, I realize this, we don't think more about mouth is because we think too much about the face. So, yeah, we are in front of the mirror in the morning when we are brushing and we look at ourselves. What we see is what we understand more and what is inside is what, you know, so that takes a lot of, you know, we think about it a lot more than what we think about is going on inside. That we cannot see. So, sir, can you share a story where, you know, your work in biostatistics directly influenced patient care or health outcomes? I mean, some kind of real world impact

Murthy: 30:36

of your research? Real world impact of research is very, because like, as you know very well. Research happens, but the impact comes at a very later stage because you need so much of evidence and so much of trial. But like certain aspects that through my research, what I am trying, like what I am trying or asking my collaborators, my students is the change is coming in terms of asking proper research questions. Doing proper research analysis. And also promoting you know, the health literacy side from the health literacy side. What can we promote? Like, how can we promote? How can we engage more people? That is where the change is coming. But like, you know, where did it come from? I don't know yet, but like, yes, certainly sometimes when we get invited speaker you know, invitations and when we get people asking us to do workshops, then for sure, you know, that my research has reached certain audience and they want to learn about these techniques because for a biased registration, you don't see the you don't see the entire process and end of your research. It's only through these aspects like being invited to workshops, being invited to workshops you can see that people are interested in your work, people are collaborating with you, and people want to use the methods that you have developed or you have been using. So this is the kind of change that I see, and you can see that I have collaborators all over the world so people are interested in my work, people are You know, referring citing my work in their work, but like I can tell you one one work that I did with my wife on pain research and she is a pain researcher because she is a medical doctor, she is a pain researcher and we have worked on several manuscripts, both of us. And a couple of articles that she got invited was to give you know, because like in Australia, you have arthritis, juvenile arthritis associations, and these people have invited her to come and give presentations. To the parents to, you know, how to cope with children who are having juvenile arthritis and pain that they suffer. You know, how can you educate the parents? How can you educate kids? To better manage their pain or to for the carer to better manage their children's or the caregiving person's pain. So those are the kind of things that I have seen these instances that people are interested in our work. They want to use this. They want to use for the research translation. So this is how I I see a value for what I do. That's

Amit Gupta: 33:43

great. In fact we are also scheduled to meet up with Dr. Mani and do a session with her. So it is it is gonna be another interesting conversation when we meet with her and talk to her about Object. So, we'll make a point and ask her to dwell more a little bit on, on this

Murthy: 34:01

topic. Sure, yeah, certainly. Because recently she has been invited. By the New South Wales Arthritis Association, where she has given training and she's also working on translating the guidelines for you know, for doctors, nurse and other care providers, physiotherapists. With respect to her pain research. So really, that is that is where you see a direct translation.

Amit Gupta: 34:31

Absolutely. And her work is celebrated everywhere, even in India. Rahul met her actually at the. The award ceremony, she was recently awarded the award and in person in

Murthy: 34:43

New Delhi. Yeah. If you want to know who we are, like, the biased statisticians are there. You know, behind the scenes people, what makes things happen. So,

Amit Gupta: 34:58

Dr. Murthy let's talk about something that happens behind the scenes, which is the most buzzword in the industry today is AI and machine learning. Almost every sector or every aspect of life has been touched by AI in one way or the other. And I'm sure we can say the same thing about the health and wellness sector. Both machine learning and AI is being fused into new devices and platforms every day. There's a startup cropping up every hour, which is you know, AI enabled or AI powered, and there's machine learning in the background. You have spent a lot of time researching and number crunching. What has been your experience and most important findings in this area? What do you think is the potential and are we being responsible? With, aI today, and especially from a biostatistics

Murthy: 35:59

angle. Yeah, that's a very a heavy question, I would say, because there's so many questions, sub questions in it. But let me first start with the usefulness of AI. AI is tremendously useful in the digital age. Because the amount of information that we are getting, like if you take any statistics book or from 1950s, the biggest sample size that they'll talk is 30 samples. Because at that time, they were doing on hand calculators. So doing anything with 30 samples is a huge sample size doing on a hand calculator. But as you got like a, because I don't know if you saw engineering calculators that came in 1980s and 1990s. So those engineering calculators, you could do certain like till 100. You know, maybe 100 100 is also still big sample, but like when you started to get the computers and when you started to get more and more information and the computers allowed you to store more and more information not only on your. You know, growth, the physiological information, but you started to store information on your images, like your you know, like, like your MRI, your x ray, your, you know, see, you know, CT scans, or your electrocardiographic Information all of this has been digitalized and you get this information and this is not for one time period. You get multiple time periods. So the power of the computer or the power of the programming then lied in. How do I collate this information? How can I use this information? How can I analyze this information? So that's where your computer programs have helped you analyze millions of data points. But now when you get images and you want to extract image information and these images are hourly images like you, do you have a, do you have a smartwatch on your hand? If you have a smartwatch like an Apple watch that records every second information of your body. Right. These are called as wearable devices. So where you are extracting your heart beat, where you are extracting your precipitation levels, where you are extract, extracting your, you know, whatever the software you have or you put or app that you put for your blood pressure, BMI heart rate, or whatever that is. These variables extract information by second, by minute, by day. So think about the vastness of the information. So if we have to analyze, if we want to you know, analyze all of this information, you need to appoint lots of statisticians who will be looking not at the entirety, but looking at aspects of that information. Whereas with the machine learning, what you could do is now collate all this information and with AI, by training AI, you can analyze these images. Very appropriately, you know, or by writing because like the photograph is the defined space with pixel and pixel size and pixel length. If you write a program that will give you the size of the photograph, like if it is an 8 by 5, 8 by 12 photograph, there is a defined pixel. Size within that photograph. So you can fix all these parameters and then write programs that will allow you to analyze photographs very quickly. So that is what the AI is doing in terms of image analysis that collates information that works on digital or images, and then gives you the information. And now the machine learning part, when we have written programs, when the statisticians have written programs, because machine learning is a language that the machine uses. to analyze this data. So it uses all the statistical techniques that we have developed so far, or we know so far, and then applies those statistical techniques for analyzing the information that is coming out of these variables. That is coming out of these images and also it aligns with your electronic medical records like age and your blood pressure readings or your diabetes or your medication, pharmaceutical information, you bring all of these together. So, from that point of view, AI is very useful. But your question was, are we responsible with when you say, are we responsible? I think you're intending to say, are we responsible in analyzing? Are we properly analyzing this data? Are we creating more biases? Yes, of course, there are biases. Like if you look at, you know, big top journals like Lancet or New England Journal of Medicine. Yeah. They are saying like there are biases. The machines are creating bias. Yes, the machines are creating bias. Why are they creating bias is because inherently the medical systems. are biased. When they, when you say biased, they are biased because the medicine was not given to poor people. If you look at the historical development, it's not given to poor people because they can't afford it or they don't use they don't have access to medical services. And because of that, some certain sectors of communities have inequalities and inequities in health, access to health and gaining of health. So these inequalities and inequities, what they create is a bias in the information that is going into the machine learning models. And what that does is, of course, when bias information is feeded into the model, the result is a bias estimate. So we got to be very careful. We got to collect more information, we got to gather more information. And then when you train these algorithms. to be not biased then they won't pick up certain segments of population. So, to answer that question are we responsible? Yes, of course we are very, we must be very responsible. We need ethics around it. We need guidelines around it. How to use this, when to use this how to appropriately convey information out of this. For all of those things, we need a lot of ethics, we need guidance from medical sector, we need guidance from computer sector, we need guidance from manufacturers. We need ethics from all of these people. All these people should be responsible for the information because it is our, and at the end of the day, the patient should ask questions because it is their information. How is my information being used? And of course, all these business people make a lot of money out of it. At the end of the day, you can see recently in Australia the Optus, Optus, is a provider for your wifi and your telephone services. Okay? So because they didn't have proper tools for securing that information, what was created is a cyber crime. And because of the cyber crime, what happened with all information was leaked. All information of the patients is leaked. So when that, those things should not happen too frequently. We've got to, we've got to have proper regulations and. What happened to Optus? They fired the CEO. Is that end? There should be some repercussions for that CEO. Firing the people must not be, you know, people lost their identity. People lost their passport, you know, passport identity. So there could be identity thefts. All of these kind of things can be generated out of this open information or unsecured information. So you've got to be careful and people should be held responsible for these kind of events when they happen, the big organization. And then also the important thing is we see that it is not only this cyber information or, In a soft information, if an electricity grid is on computer and if that is stopped, all of a sudden, you can see a lot of mishap happening in the operation theaters because of the machines won't work because they're on a grid there. They need power. And if you don't have a backup as soon as the power is off, then people may die on the operation table. So who is held responsible for that? We need to think through all these things. Certainly it is useful, but there are also side effects of ai. Like any other scientific science-based innovation, you have side effects. Thank you, sir.

Amit Gupta: 45:08

So, I noticed you're currently looking for students or you're always looking for students who are interested in pursuing a PhD or postdoctoral work or even collaborative research. What do you look for in a student? Before blessing their application. So you must be receiving a lot of different applications. You're looking for their academics, but beyond academics, what are those traits that you would look for?

Rahul Mehra: 45:36

So for the students that we would like to know top three things. Yeah,

Murthy: 45:44

the top three things that you see, academics are secondary for me you know, when I say academics are secondary in the sense like you don't have to be in the top one percentile for me to have you as my student, you know, if students are listening to this, you don't need to be in the top one tier person. That is not my criteria. But what I look is a honest willingness to learn things that is very important. If you don't have an inclination to learn things, I can't teach you anything. If you say I know it all, then why do I, why do you need me? to be a supervisor. So I like a person who is very honest in terms of willingness to learn things and who has a deep curiosity. It's not like, I know only math and I'm very good at it. That is not sufficient because nowadays, Nothing is uni disciplinary. It's interdisciplinary. So I need to learn biology, I need to learn statistics, I need to learn programming, so I need to learn all this kind of things. So I need a person who is willing to put in that effort to pursue different things like philosophy or, you know, you don't know where your next idea comes from. Is it from a business? discussion or is it from an economic discussion or is it from a philosophical discussion? I don't know where the next idea comes from, but we need to borrow knowledge from different, you know, fields. Either it could be engineering, either it could be mechanical engineering, electrical engineering, or it can be philosophy, it can be economics, or even it can be arts. You know, because data visualization is completely an art, so if I want to think differently about my data, I need to have some appreciation for some art and, you know, paintings or looking things differently. So, one is, you know, a couple of things are there and then my, you know, you should have as a student, you should have a deep passion to be different and wanting to be different and wanting to give back to the community. That is very crucial, because if you say, I want community to give me everything, but I don't want to give back anything to the community. No, that doesn't work. You have to give something back to the community. Research is about giving back to the community, because you're being given so much of money to do research. And this is all taxpayers money. So the taxpayer must be benefiting from your research. That is what the research is about. You know, so those are the qualities. Willingness, persuasion and persistence and eagerness to learn things and be different. That is excellent, sir. I

Amit Gupta: 48:51

think the most important message that I took back from this was your willingness to contribute back to the community, right? So community is a two way street. And who could tell better on this you know, from us and Rahul. So we, we are kind of living the Indian community day in and day out. And and on behalf of the entire community again, once again, I would like to thank you for all this time and the insights that you've shared with us. I know we we are running out of time on this session, but I think we want to come back to you and talk more about some of these subjects, especially on AI and and machine learning. And I know from one of our previous conversations that you have a lot of insights and opinions on where AI will be and how we as consumers as humans will need to behave around it. I definitely want to park that conversation for another episode. And again, on behalf of the Indian community, we want to thank you and congratulate you for all the work that you do.

Murthy: 49:59

Thank you so much. Thank you. I'm very happy that we have this conversation. I hope students listen to this and benefit a little bit out of this. That is the main intent that you know, because as first generation people, when we come we never had this sort of information. We, our parents said go out and explore. And so we went out and explored, but going out and exploring things is not easy. You know, it is like finding a needle in a haystack. It's very hard. So. You know, for people who are coming abroad nowadays, they have so much of information because of people like you, who are encouraging these sorts of conversations and providing or putting this knowledge freely out there. So I really appreciate and commend both of you for doing this great job and giving back to the community and also giving me the opportunity to be part of your community. So thank you very much. The pleasure is ours.

Amit Gupta: 51:05

Thank you so much, sir. Thank you.