286: [Sumedha Rai] The Future of AI and Your Wallet

>> Anthony Weaver: This episode is sponsored by podvantage AI. >> Sumedha Rai: Uh, to a certain extent, it's a supervised learning algorithm, which means that you tell the machine, you take this type of content, you put it into this category, and it's going to do it really fast. So those type of jobs I do feel are going toa get replaced. But what can you do? You can see how to upskill yourself in that,

um, domain. Can you do something, um, similar to the domain, similar to the expertise, but can you do it using tech? Because if you're a person who is now using thei solution to enable your job, you're already an expert in that field and people need that. People need more workers who understand AI so that they can integrate it. That's in demand. >> Anthony Weaver: Welcome back everybody, back to another exciting show of day.

About that Wallet podcast. I'm, um, your host, Anthony Weaver, where we are focusing on the Saich generation, building strong financial habits so that they can spend money, talk about money and enjoy their money with confidence. Today I have somebody who has been in and out, out of the financial realm, but also with the AI spin and I cannot wait to get this started. So how you doing today, Samita?

>> Sumedha Rai: Uh, I'm doing very well. I'm very happy to be on the show and um, I'm really looking forward to these conversations because I've enjoyed your podcast in the past. So it's an honor to be one of the guests. >> Anthony Weaver: O thank you so much, so much. Um, and I'm really dying to really dive into your story because you've been, um, I would say internationally exposed to so many different things in the world.

And what brings you here today is what brings this question up is, um, so many people are pretty much scared of how powerful AI is getting and what is your take on the impacts on the human psyche? >> Sumedha Rai: Um, so I think with any new change that comes into our world, there is a lot of excitement, be a lot of panic, see a lot of misinformation. And I see this with AI as well.

People are really excited to try out the new tech. Every business wants it as a, ah, solution so that they can go out there and tell everyone, tell their competitors or the consumers're powering our solutions through AI and generative AI. Um, the other hand, there are also people, uh, really panicking about job security and they're saying it's going to take away all the jobs and um, replace the human jobs that we have with like, I don't know, some robotic jobs or A.I.

jobs. Um, I think the first step to any New change or any new tech in this case is understanding. So since everybody is talking about it, oftentimes what happens is that we get really affected by other people's opinions. And we see this one article that talks about how, you know, the next 10 years are gonna be really grim for us, and another article that talks about how it's gonna completely take over all the human jobs and

we start feeling the sense of panic. So at this point what I'd like to do is I'd like to tell the person, um, is this your opinion, uh, which has been formed because of some research that you have done, or is this just something that you've been hearing a lot? And I totally understand that if you hear something repetitively, um, it will make an impact on you. So the first thing that you need to do is understand what the technology is.

Understanding is the absolute key. And in this case specifically, knowledge will empower you because um, when you're introduced to new topic and you don't know anything about it, you go online and you read five articles, um, you feel really confident and you're ready to have a conversation about it. >> Anthony Weaver: There'five articles. >> Sumedha Rai: Y. I mean I've seen people who go like,

yeah, I'm not too sure about it. And then like one week later they'll come back to me and say I did my research and uh, they want to talk about it now. So there's a sense of confidence that you get when you understand something. And I'm not saying that you're going to understand all the technical aspects of it, exactly how the model's working, but just having an understanding of what this solution looks like. What can AI actually do,

what kind of jobs can it replace? Because at the end of the day there are certain jobs that are taking say like 8, 8 or 10 human hours and they're repetitive in nature and they can be automated. And if those jobs can be automated, um, people and businesses will want to do it like that. It's going to be done at a fraction of a cost with less error because there is no human fatigue or um, there is less um, um, human error due to monotony. So they will become more efficient.

But we also need to understand that there is like this whole new uh, spectrum of jobs that is not going to get affected just because we're not there yet. AI is just not there yet. So um, if you're panicking because you've been hearing about this a, ah lot and you don't understand it completely, go out there, read about it and um, if you're scared about your jobs, understand what is the nature of

your job and how is ainna affect it. So I was talking to someone recently and um, she mentioned that as a part of her team, um, there was one person who was categorizing a lot of data into uh, 150 categories. And she used to take 10 hours, 15 hours to go through one sheet because like just the sheer amount of data that was on the sheet needed that much amount of time. And then she said, we started talking to this vendor and they said they can categorize it, which makes complete sense to me to a

certain extent. It's a supervised learning algorithm, which means that you tell the machine, you take this type of content, you put it into this category and it's going toa do it really fast. So those type of jobs I do feel arenna get replaced. But what can you do? You can see how to upskill yourself in that, um, domain. Can you do something, um, similar to the domain, similar to the expertise, but can you do

it using tech? Because if you're a person who is now using thei solution to enable your job, you're already an expert in that field and people need that. People need more workers who understand AI so that they can integrate it. That's in demand. So I think, um, TLDR as we call it. >> Anthony Weaver: Yes. >> Sumedha Rai: Um, don't be scared immediately when you read something.

Please go and do your research. That's gonna give you a little more confidence, that's gonna give you some more knowledge and see if you can take some steps, um, to go in a different direction. Or maybe you're just gonna realize, okay, um, I'm a nurse, this is not going toa get replaced anytime soon. So you'renn do better. >> Anthony Weaver: So what are your thoughts on working harder, not smarter first and then work smarter? >> Sumedha Rai: Uh, okay, I think

it's a cool thing, uh, to ask. I think even with the tech that we have, the tech has already worked very hard. There are humans behind it that have worked hard to enable some humans to work smart. So, um, personally I feel any solution, any tech, has both these already embedded in it, whether you see it or not. So you can say that with the solutions you're going to work smarter, but also making the solution.

Somebody worked hard to create a solution and now we're working hard to improve the solution because you might have seen it. Um, a lot of companies are coming up with new solutions every day. Not all of them are vetted. Some of them might be trained on biased data. Some of them might have problems related to knowledge gaps. Some of them might have been trained till a certain date so they don't have a lot of recent data. Some of them are not showing you the sources.

We're still putting in a lot of hard work to make sure that um, it is a solution that is promoting uh, a better society, so to say. So we're still working hard and smart, but once we get to a point where we need to work a little less hard and a little more smart, that's when I would say, yeah, concentrate towards working smarter. Don't be one of those people who says, I'm not going to touch it. I'm very

comfortable. Because at some point uh, you're going to be pushed in that direction, whether it's by your company, by your peers, whether by the sheer fact that your uh, job might get um, obsolete. You will get. >> Anthony Weaver: Yeah, because what I'm thinking about is when it comes to job replacement and promote promotability, one of the things that I always promote or say to myself or anybody who's looking to get into their field for the first time is to

understand the business as a whole. Write the processes and procedures in place and streamline it so that you can promote or uh, replace yourself in that sense. Uh, so say if somebody is looking at their job and they're like, hey, this is really monotonous. Is it a way that, like what tool or program should they look into learning if they're looking

to replace some of those repetitive things? So like, the company might not be on go about this AI thing but say if they can just write a small Python script or small. I mean, I know c, I'm aging myself here, But a small JavaScript to kind of do what they do day to day. Uh, what are your thoughts on that? >> Sumedha Rai: Um, I think if you're a person who can actually take the initiative to do it, I would support it

completely. Um, um. A lot of solutions nowadays are built on Python and there are so many different environments that um, will enable you to do it free of cost. Google, um, Colab, for instance, is going to give you GPUs for free. So you can even try out computer vision models or you can try out language models which are very heavy in nature and sometimes cannot be done on someone's machine.

So if you're that kind of a person, I, um, would say the first thing that you need to do is uh, to understand what your problem statement is. So um, you break it down into parts and you see which part needs to Be automated and then you do a little more research about it. If it's ah, a general question, um, like I don't know, classifying it into certain categories as I was talking about before, the chances are the likelihood is that there is already an open source library that exists to do that

work for you. If you're someone who already knows a little bit of Python, that's actually a language that's used a lot in um, AI machine learning, some flavor of Python. It could be Piparark, which is used for uh, making sure that you can handle a lot of data. It could be Pytorch, which is used for more deep learning algorithms. It could be something else. But people have put in a lot of hard work already to create open source libraries for you.

And if your solution, um, the problem that you're looking to automate um, is, is something that is, that has been done before, there is a chance that somebody's already created a library for it. So then your job becomes even easier. You import that library, you use it, you test it out and you see if things are getting better for you from then on. If this is something uh, that you fancy, uh,

you're going to organically get pulled into it. Trust me, if you're a person who enjoys seeing the automation work, there is no chance that you're not going to want to do more. >> Anthony Weaver: Yes, I love it, uh, because you have a talk coming up, um, in New York City say in June, uh, talking about the navigating the bias in AI, just kind of promoting the ethical decisions making

in fintech. Can you just kind of give us a quick synopsis for those people who have no clue about the bias or the non bias in AI? >> Sumedha Rai: Right. Um, I think that talk was last year, uh, but it was a very interesting talk. I tried to capture some of the ideas. No, you're good. I'm so glad that you actually looked it up. I'm thrilled. Ah, there's another one coming up in April that's about fraud detection. But let me just talk about the bias in AI and why I think

it's a topic that we should be talking about. So um, most of the training data, um, as we call it, and I'm just going to explain what training data is real quick. And AI model needs to learn from something, um, it needs to predict something for you and for that it needs to understand what it's trying to

predict. So as a simple example, if you want the AI model to understand numbers, handwritten numbers, you need to show it a lot of Pictures of handwritten digits and then it's going to understand writing styles, it's going to understand a two is with a curve, but a one is a straight line, stuff like this. So it needs a lot of training data to understand what it's supposed to predict.

Now a lot of the training data that we had in the past, um, which has already been generated was based on decisions that could have been biased because you know, um, it could be because of human cognitive errors, it could be because of, you know, the society that existed at that time. But we're using any and all of data that we can find right now to train models that they understand a lot. So ah, the AI model becomes better if it sees a lot more

examples. It's like a child saying, if you show me an orange five times I um, will understand that it's an orange. But if you show me an orange a hundred times, I will definitely understand that it's an orange. So uh, we try to give the model as much data as possible. Unfortunately, uh, we might be feeding it biased data. And as part of our systems we're not always very careful that the results that we're getting are not

biased. Um, if we use such models to make automated decision making, it might give out biased results and it might affect actual human lives. So for me the question of understanding the results of a model from an ethical and um, unbiased perspective is very important. But a lot of companies that are integrating AI, ah, solutions are not making it as an important, like an absolute necessary

part of their pipelines. Right now they, you fine tune the model or train the model, they see how it performs on a uh, on a set of data, testing data, um, and they deploy it. But there needs to be this other chunk that needs to go before the results are deployed which somehow overlays the demographic data on top of your results to understand is it somehow producing biase results? If it's producing biase results, what can we do? We recalibrate it for a while and then refeed

it. Do we not use the solution as an automated solution? We could have a human in the loop. And is this something that is going to really critically affect the lives of people, for instance in credit decisions or in some medical decisions? So um, for certain things this is almost an irrepraceable part of your pipeline and we need to have more conversations about it.

>> Anthony Weaver: Yeah, because I'm thinking about the models back in like reading some of the books of how credit was actually dispersed to uh, different neighborhoods or people of color during that timef frme was like, they call it redlining and based on, you know, the credit score or they have the valability of their business actually being successful or the banks are willing to

take that type of risk. So is this kind of where we getting back into, hey, they don't know enough information about us, uh, meaning us as people of color because we were always left, it seems like we always left behind and it was always trying to play this catch up game. How can we now be part of that um, that conversation or be part of that data set so that we aren't left out of it? >> Sumedha Rai: Right. I think this is a very, very important

uh, topic that you're raising right now. Very important question to address. And you're right. There were certain subsections of the population that were not even given the credit. So there is no data about them or you know, the decisions were all adverse, which means there is

negative data about them. So uh, one of the things that we start with is to give it more representative data, which means if you pick out a certain subsection of the population, you give it both positive and negative cases and it needs to be more

balanced. So for instance, in your data set, if you have say five different um, geographies, you give it enough positive and negative data in each of those geographies, uh, four for um, the model to understand that this should not be the primary factor that I can make a decision on. And this could be something to do with age, it could be something to do with um, um, racial bias, um,

that we've been seeing. It could be something to do with geography, it could be something to do with um, the credit score. So unfortunately, credit scores have been linked with um, geographies and races for some time. And this is the link that we're trying to del link now. Now, assuming you still see the results, I understand in certain cases you actually do not have

data, so you need to train it. Assuming you see some results, once you see the results, um, AI models are inherently predictive in nature, which means they're giving you probab scores. Can you do something to the probability scores to bring these results on the same um, level so you can recalibrate the scores to say this negative is equal to, to the same negative in terms of like the absolute number. In terms of the probability number that I just gave you, these two negatives look equal to me.

Um, and then you use this data again to train your model so then it understands that I'm looking for patterns right now. I'm not going to just concentrate on these, like probability measures. Another thing that's really important is to quantify the bias. So, um, let's imagine that, uh, you're training a model and there's a person who's going toa evaluate it. The person needs to evaluate

it on a certain metric. Now what I'm saying is that when you overlay the demographic data on top of your results, the false positive rate, like a person being denied credit, for instance, but wrongly, should be equal in both the segments. Like if the model is denying something incorrectly, it should deny it in both the segments equally. And now if it's not, let's work towards getting this number correct. So there are things that you can do and hopefully, you know, we're going towards a society

that's more equal. So the data that's going to be generated now, now it's going toa be more equitable. So when we use this to retrain the models, things are going toa look better. And again, for this thing, we can t have the solution be automated completely. We need not just one, but several humans in the loop. And I say several humans because we as people are also biased. We have human cognitive biases. So you cannot just

have, um, one person's opinion. Sometimes for crucial things, you need to have five people's opinions. It's like taking the average of a model. So you get the results from the model, you run it through like five people's decision making and then you generate some results which are hopefully more equitable, more accurate. >> Anthony Weaver: Yeah, okay. Um, because I'm just thinking of, I'm glad that there are people like you who are actually stepping up and saying something in this world

of tech. And I'm sure you had your set of boundaries to kind of push through, uh, what were the hardest challenges of actually getting into this industry? >> Sumedha Rai: Um, I think one of the things that is hard for people historically is to not come from a tech background that can sometimes be a blocker. Um, I actually started, um, in economics and then I did a master's in finance and I spent some time in investment banking and private equity. It had nothing to do with

tech. And, uh, when I entered the field of tech, there were already people who were working, uh, in tech for the last like five years or something. So for them, programming was something that was second nature. And for me it was something new that I had to spend some time on. I taken, I'd taken lessons in Java and C

during high school. Yeah, and you know, like, it builds, uh, your logical thinking, it builds Your reasoning, thinking, you know how if statements work but then um, you're now programming in a new language uh, which has a different syntax and it takes some time to get used to it. Whereas the people who have already been working in this. But then every person can have their strengths when they're entering a

new field. So for instance for me uh, domain knowledge was a point of strength and statistics was a point of strength. So um, I was entering the field of data science and AI where the decision making also relies on uh, the statistical assumptions that we make. Some math driven uh, things that we have to take into account before we reach the programming stage.

So um, I found my strengths over there and then I built up on this other side of like the tech knowledge and infrastructure knowledge and programming knowledge. So if you're a person who wants to get into tech but is not exactly from a tech background, it is not exactly a blocker. In today's world at least I see so many people entering it, taking some time to upsill themselves and we have reached a point where

we're trying to democratize AI as much as possible. There are so many open source options available. When I started my journey, um, I was very clear on the fact that I have to do a formal master's program and I did it and it was very useful for me. But then I also saw people just going to boot camps. You know it's harder like compress that amount of knowledge in just like 16

weeks or something. But people were using this option and then there is uh, like this whole segment of people who just go online and they pick up courses and over time they are doing very well as well. Like they, some people choose to specialize in a certain subsectionion of tech or AI and they get very, very good at it. Like um, visualizations. It's a very very

important part of conveying your point. And I still think there is a lot of work that's needed from my side to find the perfect graph or the perfect um, um PowerPoint slide. That makes it, it so much better for a person who's from a non tech background to understand my point and people have mastered the skill and they are absolutely required. Like I'd love to get some tips from

them. But uh, I think again my point is if you're interested in the field of AI but you don't come from a tech background, don't let that be a blocker. >> Anthony Weaver: Got you. So giveniving your experience growing up um, and studying in India, the UK and also the US like how does could these diverse cultural and educational settings like really shape your approach to problem solving?

>> Sumedha Rai: Uh, I think collaboration is one of the biggest strengths that I have gained uh because I worked in teams that were culturally different, that had a different set of opinions and how they put it in front of you and they were also you know um, on different skill levels. So um, what I saw in one country was not necessarily how things were uh managed in another country both in terms of like communication, in terms of their um

approach to problems. But what really helped me understand is that you can gain something out of every opinion, every perspective. Um um one of the things that we do in statistics sometimes or you know in modeling sometimes is pooling averages. And um, if the model is not performing great we get a lot of smaller models together and then get the average out of that.

And to me uh, my cross country experience, my cross cultural experience was somewhat of a pooling average because I try to listen to everyone's opinion with the mindset that I'm going to pick out this one thing that's going to really resonate with me or I'm going to have this h. I didn't think of it like this moment and then I'm going to work on it and I'm going to understand and uh, how to go about it from a different perspective. And I actually I do gain a lot from conversations.

It could be a person sitting in the UK and South Africa and India or my co worker in the US and talking about the same topic like how do you approach fraud detection in AI gives me so many different perspectives that my resulting model is actually very strong.

>> Anthony Weaver: Nice. Uh because you also mentioned like um I'm thinking about from other things that you've talked about before but considering your work with like natural language processing or NLP's that you've talked about before and understanding like the user sentiments, um are they like the emotional challenges that the Sadwich generation have which is um the design of AI doesn't offer really practical

assistance for them. So like how there are ways that they can actually utilize it in their day to day. So I would say they got kids and then they got parents to deal with. Is there any AI tools or suggestions that you have for them?

>> Sumedha Rai: Um, I think since we're thinking of um a person who wants to us AI in their day to day lives I could come up with so many suggestions because um, we've really handed these uh, machine learning solutions to people's uh hands literally like fed them saying here's a $20 subscription much like

net go interact with chat GB. But uh, since I'm on a finance podcast, let, um, me probably talk about some solutions that people should absolutely start using if they're not using it already. Um, I think money is an important topic in every family, for every individual. Um, so sometimes we do not know how to get started with money conversations.

We don't know what to do with our money. And one of the worst mistakes that you could do with your money is to just let it sit in a bank account without doing anything with it. Um, so you absolutely need to start your journey to do something with it. Um, it could be that you want to save a lot more and then invested later. But you're, you know, to begin with, you're not saving at all and you're finding trouble understanding what am I doing with my money. Start with a

budgeting app. Um, I've seen people who create these beautiful Excel sheets. Everything is color coded, everything is linked. But as you mentioned, a person with two kids and a full time job may not have the time to do it. So there are these apps that make budgeting really easy for you. Um, they look back at the transactions that you've made in the last, I don't know, two years, five years.

And um, given a set of goals that you have, the future, they're gonna start nudging you in the right direction, saying, you know, okay, in the last 24 months it looks like, um, you did not spend so much on travel, but suddenly I see a lot of Ubers right, happening. Are you just doing this because you're, you're not, you really needed it, or is this because of something that you

could have avoided and you save instead? Now once you've started saving, you, um, have to start investing in something, uh, in somewhere because your money needs to make more money, so to say. So, um, what are you doing in your investment journey? Um, you, are you using, are you comfortable enough to say I already know where I want to put my stocks? In which case maybe a robo advisor could

help. Like, you know, it a fraction of the cost and you could just do with some extra financial advice that a financial expert could give you. But maybe you couldn't afford the financial expt, so you could look at a robo advisor if you don't want to do that. There are investments that are more passive in nature. So, um, you could, uh, you pay a subscription cost or you could pay um, a dollar cost for every transaction that's made. But you start uh, tying up your investments to an index

and you're passively investing it. And um, there are certain AI solutions that also like, tailor to your needs. Like, um, um, this is the ROI that I want to achieve in the next two years. These are my long term goals. I'm saving to buy a house. This is how much the house can cost me. Um, these are my risk preferences. So you give it a whole bunch of parameters and then it starts rebalancing your portfolio, given the market conditions,

and starts putting the money in there. So those are also good solutions to have. Um, I've seen people who are very good at investing. They're like, look at the market every single day. I'm going through my own articles and I'm doing my own research for such people. Um, um, getting a summary of the financial news every single day could be helpful as

well. Uh, I think the Bloomberg terminals have it where when you open the terminal, the first thing that you see on the screen, or like it's a part of the terminal somewhere, is five short bullet points about how the financial markets are looking. And for me, that's going to be really useful information because I might say, okay, this is something that I don't even invest in, so maybe it's not

useful. But then I might find something to do with the Magnificent Seven and I will click into it and I'll start reading about it. So there are solutions, uh, for a person to look into their money more closely. And, um, I think people should absolutely start doing it today. Like, start getting close to where your money is going. How is it going? What can you do to make it go in a better place today if you're not already doing it?

>> Anthony Weaver: And that brings up a point of now want to learn more about you, if you don't mind. >> Sumedha Rai: U. >> Anthony Weaver: Uh, it's like, did your parents, um, bring you into the financial realm or is this something that you just kind of like? You know what? I want to learn more about finances because we didn't talk about it in the house. >> Sumedha Rai: Uh, right now I think

this is a really cool question. I would say yes and no. Okay, so, uh, my parents are actually doctors, so n, um, topics about investing were not exactly table conversations. Uh, not, um, we were not discussing this over dinner. I could tell you so much about, uh, Tylenol and Advilce and the thoughts that make up these medicines. Um, but, uh, we were not exactly discussing where to put the money or what to do with the

next paycheck. But I do remember that my dad was very particular about doing his own taxes with his accountant So I remember he used to maintain this, uh, book which had all his transactions written down, so that when he's giving this information to the accountant, he's not a person who doesn't know what, where his money is going and how it's being treated or, like, how his taxes are getting filed. He knows exactly what he's giving to the accountant. Um, he knows his

transactions. There are times when he would just like, call his accountant and say, oh, I see this over here, but I think I did this. Can you go to, like, page 14 and check this out? I have it in my note. And, um, that image stuck with me. And, um, I didn't start investing right away. Truth be told, I entered the investing game a little later. But now that I'm already in it, I do understand that if I'm managing my own money, I feel a lot more comfortable with it. Even if I'm

asking someone else to do it with me. I would feel a lot more confident if I also had knowledge about it. And, um, this whole, um, uh, making sure that you know exactly what you spending habits is, is the key to it. Um, my parents used to discuss these things with each other. So, like, sometimes I would hear a conversation. So I know the conversations were happening

in the house, but they were not exactly happening with us. So maybe this is what I would advise people, that there are certain conversations that you can also start having with your kids. Um, what does it mean to save? Um, what is actually the value of money? There's a really cool, uh, thing that I remember, uh, when I was talking to my nephew this one time, I asked him what did he want for his birthday? And he said, you could either get me a fidget spinner or you could get me an iPhone.

And I remember thinking, oh, my God. I don't think you, that he has any idea of what, like, the cost of these two things are. >> Anthony Weaver: Right? >> Sumedha Rai: And this is just a funny example to say that, um, start teaching your kids the value of money. Like, what can work for them? What is $10? What can it buy? What is, you know, a thousand dollars? What's a good income to live off of, say, in New York City?

And I'm not saying you need to have these conversations when the kids 3 years old, but when they're in their teens or when they're in high school. Um, these are conversations that should be almost a part of our, um, daily talk. Uh, another thing that I would strongly recommend everyone to do is get close to your taxes. Um, in the last five years, I've just started doing taxes myself. It just makes me, uh, it forces me to look into what I did with my money retrospectively. Helps me

plan a little bit. Uh, which is not to say that a lot of people have much more complex taxes. They have business entities that they need to take care of. But you could hand it to an accountant, but also exactly know how the tax was filed. U. You are going to learn so much more when you do your own taxes. And you might even get ideas about, like, what to do better next year.

>> Anthony Weaver: So what tools are you using to file your taxes if you don't m mind, like, caus I was thinking, like, the big box ones, and I have a tax account that comes on, and she was like, well, you can go with them, but if you really want the money, you got to go to somebody now. >> Sumedha Rai: That's true. I'm not an expert on, like, tax structures and the best way to save money, so I use very simple

tools. Right now, I'm. I so like, pick up TurboTax or Sprint Tax, but then it asks me so many questions to fill out a return that I feel like when it's asking me so many questions, I have to go back and look at the answers to those questions. And that informs me u. Uh, a lot more than. Than just like, giving everything, giving my W2 to the accountant and, um, and just asking the

person to file the tax or e file it. Um, my point is, like, going to an accountant and getting the right help and getting the right ways to, uh, save money or filing the right returns and getting the right refunds, um, it's a very good thing to do. But also, do not make this a thing where you're completely uninformed. Because I've seen people just say, I'm not really sure my accountant does it. I'm not a big fan of that answer. Um, honestly.

>> Anthony Weaver: Well, it shows up in your work ethic. And actually not. It goes back to the data set that you were providing, your AI model. It's like, is it doing what you wanted to do? And if it's not, you know why? Because you understand what the data set that you've given it, in a way, it should operate, and it's not. So it actually shows up in everything that you do, which I have picked up on just in our short conversation. Um, I'm sure you're like, you know, why is my food

tasting this way? Do I need to add an extra spice? >> Sumedha Rai: Oh, my God, I'm such a big fan of research. Like, if I don't Understand it. I sometimes go crazy and I've been told this in the past that we don't exactly need 100% accuracy. 20% will do it. And sometimes I deliver. But I see myself going back and saying, okay, yeah, but what happened to the 10%? >> Anthony Weaver: Yes, yes.

>> Sumedha Rai: Um. >> Anthony Weaver: Cause it gets you to think about like these models that go out, um, far as like, oh yeah, you know, 29% of the people are doing this. So I'm like, well, what are the 80 mean or the 71% of the people doing that? The 29% are doing? >> Sumedha Rai: Like, uh, you know, this is actually a recurring theme in uh, AI as well. The, the choice between explainable versus explainable AI versus black box models.

And um, there are certain businesses that actually choose explainable AI over black box models, even though black box models are going to give them better metrics just because the explainability part is so critical to the business. Like, um, if, if a doctor was using an AI model and really wanted to understand the result, um, he or she needs to go back and see exactly how the result was generated. Cann I just go like, I don't know what happened.

>> Anthony Weaver: This I just handed to the AI team. >> Sumedha Rai: Yeah. And um, believe you me, people are actually doing that in the development world where they say, they wash their hands off and say, I'm not really sure how this result, um, uh, you know, was generated because I fed this to the model and this is what it gave me. Now

for certain things it might be okay. Like if I'm generating a summary of something and I get high level bullet points, I'm probably okay with a black box model as long as the result actually looks okay for something like, uh, you know, mark this as spam or not. I'm not too worried about, um, just marking something else's spam and saying the next time, just make sure this is also included. And I'm not really fussy about why did you

mark something as spam or not spam? I mean maybe not, but for certain things, like as I was mentioning, credit decisioning, if a person is denied credit, um, I need to know exactly what happened, um, at what part in the model pipeline was this decision made to go from, yes, this is what I looked at and like this is how I came to the decision. I need to know these things. So yeah, explainability is very important to me.

>> Anthony Weaver: That is awesome and I appreciate that you taking the level of detail to go through, look at all the fun stuff. That is not really many air quotes here with the fun stuff. Uh, because I'm thinking about like now as we get older, we move into like the third segment of the show, which is the features. Um, and I'm thinking about like when your parents are getting older, are you planning to have them live with you doing, um, what they call it, the assisted

living. Have you thought about or had that conversation with your parents yet? >> Sumedha Rai: Um, I think yes. Um, we've almost started talking about it, I guess, but it's really a decision that's very, very personal to them. So I mean, I can have the conversations with them, um, but really, at the end of the day, it's exactly what they want that's gonna happen. I know and I know this decision is a very important one. And for different people based on different situations, it' the

result might look very different. But at least in my case, um, it's really up to my parents. Um, they uh, were good with their finances, I think, and that'that's great. So they are able to make that choice independently. And, um, I will probably have very little say in this. >> Anthony Weaver: Okay. I actually think what it be possible that people can actually just put in their parent information into an AI model and be like, hey, what is the best solution in this

situation? Because we can't figure it out. Just dump it all in and just say pick one. >> Sumedha Rai: Uh, wow, uh, wow. Uh, very, very sensitive question.

Uh, to answer that question, I think yes, there could be a solution, but also coming back to the same conversation and um, first of all, you understand why that prediction was made and secondly, I mean, make sure that someone has a choice to say yes or no. And I'm sure like people are gonn to do that, but were something like this to get created, um, the end result should not be. I think the a said so, so this is what we'renna do. Uh, it should, um, also be like,

oh, uh, this looks like a cool result. Maybe we can talk about it, I guess. >> Anthony Weaver: Yeah, I think that would be cool. Um, at least to get the conversation starteduse a non bias. Yeah, not saye person, but a non biased entity, uh, is providing a solution based on the data that is s provided. I think that'be pretty cool. >> Sumedha Rai: Yeah. Again, like, was it really non biased though? So understand what the

data was trained on. You need to go to the very start of the solution and say, are you sure it's not biased? And yeah, that's when you rely on the non biased result. >> Anthony Weaver: So to say, uh, yes, yes, yes, full circle. I love it that you. So, um, right now do you want to focus on You. So what areas um, are you focusing on improving in your life or even your career?

>> Sumedha Rai: Right. I think having the right conversations for me is very important to me at this point of time in my life. Um, I feel like I get 24 hours in a day and um, a lot of that is spent on working um, I don't know, like preparing for a speech at a conference or, or writing an article. I love doing these things. I inna put more information out there. Good, um, information out there

if I can. Um, and so there isn't a lot of time that's honestly spent talking with people, uh, and having a uh, very long conversation. So when I do get the chance, like if I go at a conference or if I am m talking with a friend, I have started making this conscious effort to also bring up topics that are important for me to grow as a person. Um, and I really encourage everyone to do it. I mean um, that is not to say that you should not have conversations about uh, things that are not related

to this list of important topics. It's um, I can totally understand gets very stressful. Sometimes you just need to vend. Sometimes you need to have conversations about, I don't know, um, shoes and uh, um. Those are also important conversations for your brain. But also start making a conscious effort to start having think the right conversations now. Um, the word right I wantn stress on it because the word right might um, be different for different people based on where they are in their lives.

But if you have a list of these topics that you think are important, um, start talking to people about it. If they're in the industry or if they're in that domain, you're going to get a lot of good perspectives and it's going to help you build on top of it. If you're a person who likes to research about topics, just getting those conversations started, it is going to be very good for you.

So for instance, when I go for a conference, um, and if I'm speaking at a conference, um, I'm very alert right before the speaking topic. I'm very very alert. So I would actually uh, try to get a slot that is later in the day which forces me to listen to all the conversations that are happening beforehand. And I would consciously make a decision to just go very early, listen to everything and really absorb. And as I mentioned before, there are times and I when I tell myself

h. I did not think of it from this perspective. I already had an opinion. But this is also an extra opinion that could really, you know, um, um. I could have a different take on, on things if I looked at it from this point of view. And those things are important for me. So at this, at this time in my career, I'm looking to have good, useful, uh, conversations with a, uh, lot of people in my, um, in my, um, industry. In industry, it could be in my life, it could be a part of

my friend circle. Um, I think when I hesitated with the word like industry, it's because sometimes the conversations are just about AI and how awesome it is. And then I met someone recently in journalism and she said, you know, I absolutely do not understand it at all. And my simple ask, um, for her was, okay, like, we don't have time right now, but let's set up some time, have a cup of coffee and just discuss what you don't understand about AI and I don't

understand about journalism. And then let's inform each other about how it could be maybe a collaborative, collaborative effort. Uh, uh, some time back, I was speaking to a person who works in economics, and he said, oh my God, there are so many conversations around AI. Sometimes you don't need it. That is true causality. I need causal inference. I need to understand what thing caused the other thing. I don't need predictions for my problem. And I was like, yeah, that's, that's great.

I understand. I'm working in AI, but not every cause has to be championed by AI. There are certain things that might not require it. So I'm really trying to have these conversations with so many different people, whether it's at a conference, whether it's at a meetup, whether it's with people in my own company. But I'm consciously making an effort to carve out some time to have these important conversations. And another thing that I'm trying to do is to put my

efforts in the right direction. Um, that's because in the industry that I'm working in, there is so much noise. Like the word AI is an umbrella term for so many different things about it. Yeah, I really need to do my research to understand in the next three years, what am I going to concentrate on? Am I going to start learning a new language? Am I going to start

learning a new concept? Am I going to start looking at it from a project point of view, a research point of view, a business point of view, what am I going to do in the next three years? And I'm trying to chart a course for myself because this, um, tech market is changing overnight sometimes. So I need to make sure that I'm, you know, I'm up to date with the topics. I'm in a way ahead of the curve a little bit. So I'm putting in a lot of time and effort to research my path moving forward.

Um, conversations, uh, are a part of. >> Anthony Weaver: It, um, because you got me thinking about conversations just in general. Um, and these are the things that I talk about, um, with other podcasters. It's like the future of podcasting. Um, what tools are people using, how they streamlining their services and so forth. Like, we can talk hours. Like, I could

talk your head off about this stuff. Um, and I'm sure you can talk my head off about AI and like, how you guys are looking into pretty much the same topics but just at a different take of, of like, how are you streamlining your process? How are you streamlining your code? How are you looking in power consumption, uh, with these models? Because you hear about, I think it says it's almost like an ocean's worth of cooling that you need just to run some programs.

>> Sumedha Rai: Energy corprint is very high right now. Yeah, yeah. >> Anthony Weaver: Um, and so it's just like how we all have the same general topics. It is just how do we approach it and the mindset about how of it all, like, then again goes back to do you even really need it? Because sometimes the simplest solution, like how I think it was what NASA when they went up in space and it was like, we gota find a pen that can do anti gravity. >> Sumedha Rai: My God, use a pen'a.

>> Anthony Weaver: Pist. >> Sumedha Rai: Right, yeah, I've heard about that. Uh, yeah, there were issues with that as well. The lead can break and stuff. But I understand, like, coming back and saying simple can help, doesn't need to be overco complicated. Don't overkill. >> Anthony Weaver: Yeah, I think that might be the simplest, uh, way to sum up this episode. >> Sumedha Rai: Yeah, of course, yeah. Um, I say this to, uh, startups

and businesses looking to integrate AI. Think about three things for sure. Need time and money. Do you need it? Do you have the time to, um, create it? Do you have the money to afford it? And I think the need part is very important. Like, answer that question first. Can it be done in a simpler way? Uh, do you need to throw this into a chat bot or a generative AI solution and pay for each of those API calls? Or can you do it in a simpler way? Yes. >> Anthony Weaver: Love it.

U. Um, before we get to the final four, is, ah, there anything that you want to leave the audience with before we dive into the final four? >> Sumedha Rai: I think this became like a recurring theme in this podcast for me, but my mantra is research it and prepare for it. So specifically for this technology that has, uh, so much going on for it. There's so much talk about

this technology. Research the technology, research how to use it, research how to, um, how other people are using it, research how it's being used in your industry, and then sort of like start preparing for what comes next. Um, if you feel that you need to upckill yourself, start taking those steps right now. >> Anthony Weaver: Perfect. All right, you ready for the final four? >> Sumedha Rai: Yes, of course. Looking forward to it. >> Anthony Weaver: Awesome.

Number one, what does wealth mean to you? >> Sumedha Rai: Um, uh, a, ah, really, really cool question. And I actually thought about it and I think, um, I would say that wealth for me is having the right skills and the competence to be able to generate the tradable currency in the future if I need to. And I say this because your skill set is very important. Something could happen tomorrow. You might not have your job, you might have

to use up your savings. Um, a lot of things could happen to this, this fiat money that you keep in your bank accounts or to this commodity money or to like, uh, a land you on or something. But do you actually have the skills to generate more tradable currency that we need, um, to survive, to like, buy goods and services for us? Do you have the skills? Do you have the competence

to do it? Do you feel confident, couldn't enough to do it? And if you feel like, okay, I could lose my job and be unemployed for the next six months, but I'm pretty sure with my current skill set, I would be able to get, um, on the right track to start generating this wealth again. Um, your skills, in my opinion, are then your real wealth. And um, the second more philosophical take on it is that the definition of enough can be very different for different people.

So do you feel like you have enough and does that give you a sense of contentment? So that sense of contentment for me is tied to the idea of wealth. A person with two kids might have a different sense of what wealth means than a person who is maybe, you know, um, does not have kids or has like dependent parents or is working in a different industry that, that you know, does not generate enough numbers to get to a certain, um, salary

bracket. But at the same time, given your current situations, given your current geography, given how you've been working, do you feel content with where you are? Do you feel content with the number, um, that you have in your bank account? If yes, yes, you do have wealth, and then you can plan for how to sustain this sense of contentment. But if you feel like I m. Don't think this is enough, then ask yourself, what is your definition of enough? And work towards it.

>> Anthony Weaver: Right. This is a sad question. I'm just thinking of, like, when you say enough, because I know you like to, you know, like gu. Um, um, to do your research. When is enough research for you to say, yes, I'm moving forward and I believe this to be the case. >> Sumedha Rai: Know at some point I just get tired of the screen. Uh, no, let me, uh. Yeah, that's

a good question. That's a good question. Uh, I once told someone, you did not go to the seventh page of Google Page. I think enough for me is to start feeling confident that I can, um, start the Endeav. So if I'm creating a model for fraud detection and I know nothing about it, enough for me is to start looking into what space I'm working on. What do the features mean to me? What exactly are the rules that govern

the model that I'm looking at? And then at a point where I feel, okay, let me just start coding this out, I feel like at that point the research is. And then I obviously, like, there are certain questions that come up, um, when I'm coding, then I come back to it again. But the first part of research is always longer than the successive parts after that. It's like a marginal addition on top of your knowledge. So, yeah, I think typically enough for me means I'm ready to get started.

>> Anthony Weaver: Okay. >> Sumedha Rai: After that, it's going to be, um, an incremental value to every data point that you give me. >> Anthony Weaver: I like that. All right, number two, what was your worst money mistake? >> Sumedha Rai: Uh, I think I said before I let my money sit in the bank account, a, uh, savings bank account, generating

no, uh, interest for some time. And I, I could have started sooner, and I did start, but if there is anyone out there who's still not doing it, get started today. You need to make your money work for you as well. >> Anthony Weaver: Love it. Number three, is there a book that inspire your journey or change your perspective? >> Sumedha Rai: My. Ah, God, this is a very tricky one, honestly. Uh, there are so many different books that give you so many different takes on

things. So, um, sometimes it really depends upon what I was looking for at that point in my life. Um, and the book might change, but then there's this book, um, called the Alchemist. A lot of people have read it, of course, and I think the reason why I'm mentioning it is that it brings out a very, um, very famous Kind of, um, way to approach things in life. Start learning from the journey. Like, start learning from

your experiences. Of course, like, having an end goal is very important, but while you're chasing that end goal, uh, don't forget that every small thing, every small thing that goes wrong is actually like, not really going wrong. You're learning something from, um, it. Because depending upon our goals, it could take days, months, or years to get there. And if we don't start learning from our experiences, it's gonna feel like it's a very long journey to get where we're trying to get.

But at the same time, if we have this mindset saying, let me learn from every small thing. Journal it if you want to, because when you look back at it, it'll feel like a positive reinforcement and it's going to make you, um, work in a more positive way towards your new goal. But start learning from every little thing. Um, and I think that book really talked about it in a very nice philosophical way. So I actually ended up reading it twice at different times in my life.

Um, another thing that I'm reading right now is Atomic Habits. >> Anthony Weaver: Love that book. >> Sumedha Rai: Yes, it's a great book. Um, we are all, um, a slave to our habits. Sometimes we want to break the old ones, the bad ones, and we want to start with the new ones. And again, it has, it has a very similar theme saying, you know, learn from the little ones as well. But, um, if you guys haven't checked out that book, you, you should think about reading it.

>> Anthony Weaver: Definitely. U, uh, yeah, I'll leave that one. We'll talk offline. >> Sumedha Rai: Yeah. >> Anthony Weaver: U, uh, number four, what is your favorite dish to make? >> Sumedha Rai: Wow. Uh, I like making a vegetarian lasagna. So, um, I'm vegetarian, I love Italian food. And, um, I started experimenting with it, I think two or three years ago. I burnt a lot of them over B. But I've come to a point where I know how to layer them properly and, uh,

to use different kinds of cheese. I know which one works and I can customize it based on dietary requirements at this point. And honestly, this is sometimes my catchphrase for someone who I'm inviting over, um, as a guest. I'd be like, yeah, let's play some board game. And you know what? I do a mean lasagna. And, and just leave it at that. >> Anthony Weaver: Check them. Are you using, um, are using like tofu? Like, what are you using as your, your meat

based, quote unquote meat bas. Uh, because I heard some people use like lentils, um, or like some Type of toefu base. >> Sumedha Rai: I like to mix vegetables, uh, and create like a medley. So I chop them up real small. And so there's squash, zucchini, um, um, spinach, carrot, carro, um, mushrooms. And I like to chop them up real, real nice. So then it almost feels like, um, it's. It's one base, but it's actually a medley of these bases. And it. The flavor is really nice in your mouth.

>> Anthony Weaver: Yes. Oh, man, this is. I need to see a picture. I. To follow your social. To see. >> Sumedha Rai: Yeah, I'SEND one to you or, or the next time that I'm inviting you to a party again, I'm going to use this. I'm gonna say, hey, uh, let's play some board games. Let's talk finance. But also, I make a mean lasagna. >> Anthony Weaver: Okay. I'll be sure to be hungry. >> Sumedha Rai: All. >> Anthony Weaver: Ah, right.

The very last question of the show, which is where could people find out more about you? >> Sumedha Rai: Um, I am on, um, LinkedIn. I have a personal website. Um, the website has a link to shoot me a personal message as well, if you want to. I love having conversations about AI Whether you're a person working in it, it. Whether you're a person looking to find out more about it, just shoot me a message now. We'll make sure that I get back to you. I want to enable as many conversations as

possible. Sometimes one on one is not always possible, so I do a group conversation. Um, but yeah, I'll make sure I get back to you. Um, so hit me up, please. >> Anthony Weaver: Awesome. Well, thank you so much, Samitha. I greatly appreciate all of the information that you provided to us. Um, especially about the AI stuff. Like, I'm thinking I'm doing AI, but then people like, well, as long, large, uh, language models, and I'm like, okay, well, whatever is AI in

industry. So, uh, you know, I greatly appreciate everything you provided us. And one of the things, everybody, if you're listening, if you made it to the end, I just want to let you know that you have what it takes to go to that next level. You know, you listen to Samita'story um, you saw where her parents came from. You saw that she took a whole different path than what they did. You have the ability to

change your life, your trajectory. All you have to do is just put in that effort and take time out of your day to invest in yourself. I wish you all the best. We out. Peace.