The conversation this week is with Avinash Malladhi. Avinash is a PSPO certified professional with over 15 years of experience and implementing finance, artificial intelligence and related IT projects. He holds an MBA in accounting from Wilmington University, and has completed his expertise in finance with Fintech courses from Harvard University. Throughout his career, he has been able to network with other Fintech specialists from over 30 countries and stay informed about the latest developments in AI and Fintech. Currently, he is working at SAP as a consultant on the open text vendor invoice Management Implementation Team will see our solutions include machine learning capabilities to extract value using OCR
If you are interested in learning about how AI is being applied across multiple industries, be sure to join us at a future AppliedAI Monthly meetup and help support us so we can make future Emerging Technologies North non-profit events!
Resources and Topics Mentioned in this Episode
Avinash Malladhi 0:00
OCR, let's suppose even if the OCR the old base OCR, I'm talking about a decade old to see our systems where it was not calculating it right or if it was not recognizing it right. And it was getting passed through, it was just getting passed through. Now with ML and AI built into it, it has its own set of validations built so that it says okay know, the last time when such an order was placed, it had such and such capacity is built into it. Today, you're putting something which is way over any tolerance, give a alert to the user, the user can look at. If somebody reads to some order, and if they say it's not the right one, it can be routed back to OCR to retrain it manually.
AI Announcer 0:48
Welcome to the conversations on Applied AI podcast where Justin Grammens and the team at emerging technologies North talk with experts in the fields of artificial intelligence and deep learning. In each episode, we cut through the hype and dive into how these technologies are being applied to real world problems today. We hope that you find this episode educational and applicable to your industry and connect with us to learn more about our organization at applied ai.mn. Enjoy.
Justin Grammens 1:18
Welcome everyone to the conversations on applied AI Podcast. Today we're talking with Avinash Malladhi. Avinash is a PSPO certified professional with over 15 years of experience and implementing finance, artificial intelligence and related IT projects. He holds an MBA in accounting from Wilmington University, and has completed his expertise in finance with Fintech courses from Harvard University. Throughout his career, he has been able to network with other Fintech specialists from over 30 countries and stay informed about the latest developments in AI and fintech. Currently, he is working at SAP as a consultant on the open text vendor invoice Management Implementation Team will see our solutions include machine learning capabilities to extract value using OCR, which if you didn't know stands for optical character recognition, which has its roots in computer vision, which is a branch of AI. So thank you have an edge for being on the program today.
Avinash Malladhi 2:08
And thanks for having me. It's good, good.
Justin Grammens 2:12
Well, good. The point of the conversations here is to get it probably a little bit more detail with regards to maybe some of the more interesting things you've been doing in your career. And I gave a sort of brief background with regards experience and finance and artificial intelligence. But maybe you could talk a little bit more about, you know, kind of what got you into this career and what you've been doing.
Avinash Malladhi 2:31
Yeah, sure, sure. Accounting as such, people take it as very a boring subject to be honest, because it's lots of numbers. It's lots of rules and regulations around it. And also, as such, it is very rule based and very strict oriented subjects. So not many tried to get in there, and those who get in there don't really think more about how technology could be embedded into it. But software such as SAP, Oracle, etc, etc. Those software's have really built solutions, which can automate most of the processes. But still, there are such processes, which are mostly very manual. And people don't really love doing those activities. And those activities one of those is so when there is a paper invoice that comes into the company, there's lots of information on it. And as a human that has a good amount of chance that you can end up typing a six or a nine or a three for six Shuri audios the keyboard we all do. We all have made those type of mistakes. But sometimes that one decimal can change how it was goes through. Absolutely. To be very honest. That's where my journey started as it would sound very cliche, but I entered an invoice as an AP member with an extra zero. And I thought I hit ENTER. No, I hit a zero. So I was like, I just hit enter, and it went. And it went all the way to my manager to approve it. He calls me upon and he's like, Did you see the invoice and still it's just a zero so it didn't really catch my eye then that's where it started. I was like, Okay, fine. Now he rejects the invoice, it comes back now I have to manually enter it all over again. What happens? I spend around three hours, man hours to get everything again sorted out why? Where and everything because it has gone all the way up the ladder. So that's where my thought process started. And I look for opportunities where I could collaborate and like, give my input as an accounting graduate, and I can take it forward and that's where the journey started. As it went through I started working with companies who were already in had baby steps of doing such technology based solutions and accounting and finance started off with accounting where invoice parsing was the major focus, where the OCR reads through the invoice, grabs the value, and passes that on to the next step where it can be processed. Now, so that is that was a baby step one. Now as the time went through, we all want some more of the technology. Now we know that it can grab the values what next? So every time it comes in, it can mistake a zero for an eight. Or if it doesn't read it, right. We all know it's technology, it has stuff built into it, which if not trained, right doesn't work right. Now, the training piece comes into picture. Now what else can we do to improve it? That's where the machine learning aspects come into picture where you can reinforce training into it, and then pay tries to read through it. So now that's good for English invoices, there's an English language, because that's where we are operating. And that's where at least I have been operating. Then comes the other languages into picture. So this scope is unlimited. Now you read through English invoices, now we have trained it to do certain invoices, but what if somebody changes the invoice? Sure, sure. It's not just wondering your your process when you're processing with 1000s and 1000s of vendors as a company. So that's how the process started. Me and being a subject matter expert on how to process those invoices was a big, big contribution to the development team. And then I started learning development. So the journey keeps going on every day, and we every minute, from invoices to sales orders to purchase orders to what not document that can that we can train that we can develop because everything is more or less based on template based based learning. And how do you train that particular application to read through it? So that's have been contributing to this fascinating world of OCR and stuff. So
Justin Grammens 7:20
yeah, yeah, no, that's, that's amazing. It's interesting that you kind of have a finance background, and then you sort of pick up pick up the programming along the way.
Avinash Malladhi 7:28
Yeah, so a little bit, not much. But I have picked it up with the subject knowledge on my hand, which I can talk to business about it to the people who are really processing. Now, I also have to talk to the second side of the coin, which is the development team. So have to have both sets of knowledge on hand. So start learning a new thing. And that's how I picked up a little bit on the programming. And now I'm able to probably guide through and sometimes get my hands dirty on the programming
Justin Grammens 8:00
as well. That's cool. Yeah, I worked on an OCR project back in 2015, I think is probably when it was we were using some open source software called Tesseract. I don't know if that's still used much in the field by we're using that we're also using computer. Well, basically Google vision, right? So you can just send stuff to Google and it will go ahead and and transcribe it. And I mean, geez, I think about that was probably eight years ago, now it's gotten, I'm sure the software's gotten a lot better.
Avinash Malladhi 8:26
Just office, I've got a lot better. Now, it really makes it very much interesting for people like me who are spending there all day long on it. And when you see new features and new stuff that comes out of it, on how you can train, I mean, our boss does speak from a finance perspective. Like there are things that for example, when a Chinese invoice comes into picture, it's hard for me to read it because I don't know Chinese, but then the program has been developed in such a way that it can really identify everything from it. So sitting in any part of the world, any nook and corner of the world, as a company, I can set up a shade workstation where it can be processed. So that is kind of very interesting aspect that as I said, it has been evolving and evolving like any anything else in using artificial intelligence. As a matter of fact,
Justin Grammens 9:21
you know, one of the things that I recall, as we were working on on our invoices were you know, it's not only the the number you're talking about, you could have like a wrong zero, but also just all of the information like what was purchase. So this particular application was dealing with food was basically invoices that were being sent to restaurants. And what they were doing was they were capturing a lot of information around the pricing of like produce of any types of stuff. So one restaurant in New York is paying so much for eggs and a restaurant in LA is paying for so much for eggs, but you needed to make sure you had the item that was written correctly. And it wasn't only the amount but it also was obviously the item but then what was really fun. fascinating you I'm sure you can elaborate more on this was not all the templates were the same. So one food distributor had one template and it didn't it another one a different one. And of course it's at a different location that you need to sort of parse out how you guys had to deal with that, I'm sure. Oh, yeah, sure.
Avinash Malladhi 10:14
Yeah, I mean, that is always there. I mean, it's just not with the numbers as you are saying the text, but it matters a lot, because text is what gives the template actually a path forward. Because each template and each the words that they use, not every, as you were saying, not every vendor, not every product that's taken example of an apple, and Apple doesn't come with just the word apple, they could say, the brand of the apple, and then they could say the green apple, not something else that comes along with it. So now what if I haven't had the juice and a green apple both on the same invoice. And the template has to identify one as a beverage and another one as a fruit or food item. So that differentiation, everything is being actually fed into the system, it keeps learning. And there's a reinforcement learning that we built into the system. And also just not the system learning but the users contributing to the learning. within SAP when we, for example, our speak a little bit about SAP here and the open text product that we have in there is there is a solution that we use, just not for the invoices, but to process sales orders and to read through the purchase orders. When the customer's purchase order comes into picture, the customer could call my product something at their end. And but at my end product is different, because I am manufacturing the product. So now the customer says I need a box of so and so. But now we have built a mappings into those social solutions in such a way that the customer is sending certain product. We also want the product to be identified by the OCR the same way but the way we map it, because the solution we designed a has like an input and also an output screen right there on the screen so that the vendors product, everything is listed there. Good. The customer shows us what they need. But at the same time asset, we want a salesperson also to be able to say, Okay, now if they're ordering like they need a carton of certain item, we want to make sure we are able to give them the feasible availability date, a delivery date, everything at that point in time. Straightforward, because these days even people keying in as I'm saying, there's no king anymore. With this OCR technologists getting so rapidly out. The real customer is ready, sending the purchase order in and immediately calling. When is my availability? Yeah, sure. I cannot have my salespeople start typing in and searching. We need certain mappings in there so that it should go in, identify the product, get what is availability, check everything possible in the background systems, which are the root systems and give back and output all this probably in less than a minute. Yeah, sure. So that's how good the OCR systems are like enough.
Justin Grammens 13:17
Yeah. And what was so much well, I guess a couple thinking in my head, it's like, there's still a lot of paper based systems, I guess why we just haven't gotten to a point where we've digitized everything.
Avinash Malladhi 13:26
Oh, yeah. Yeah. I mean, yes, we have not disliked there are companies who still want to use the paper paper base, especially in the US, the European market as bit different. When we come to this aspect. I have implemented such solutions in probably more than 20 countries now. And I've seen, it's in the US that we use a little bit more paper based process yet still, compared to Europe or in other parts of the world. digitalization is picking up everywhere. And it's more than I would say, yes, people are concerned about nature. But that is actually the secondary aspect. According to me, the more primary aspect is being the speed of how it can go, and how it can get to the solution that they want or get the process up and running as soon as possible. Because we still use checks here in the US. I'm in US and Canada, we still do that. Europe, I have seen very less amount of checks being used. Most of it is electronic are there. So each or each offered has its own advantage. Companies still prefer checks for their own financial planning. Whereas electronics have their own advantages. So we're paper based and electronic based are really two different aspects that we can delve into. Yeah,
Justin Grammens 14:49
yeah, that's an interesting aspect, right? So it's not just invoices, you touched on checks, right? So anytime I write a check and I use this all the time and take a picture of the check and deposit it into my bank account that uses OCR. What about something as simple as mailing a letter? Right? It needs to figure out you know, there's not humans sitting there trying to read all this stuff, right? Yeah,
Avinash Malladhi 15:08
no, no. Yep. For sure. When when you when you receive something in paper, technically, who has that patience in this time and world? So you can read through the whole document? I mean, I personally don't. Piece of paper I would rather bro probably like, Okay, what is this? And then done? I don't know. I think most of them are like me. They want automated stuff, I'm sure. And as you're seeing the checks, the life has got so easy, since you can just deposit a check, right, sitting where you are. That is something that is truly a thing that we have developed over the past few years here in the US. And I think I have not seen this in in most any other country where they can do this. Because each aspect is different. And when it comes to the same thing, if you're not using it, you don't invest your time developing a solution around.
Justin Grammens 16:03
Sure. Yeah. Yeah. When most other countries are not using paper based checks. They're already automated. It's
Avinash Malladhi 16:10
almost, almost not. But I think that is a generational thing. The elderly generation still prefers checks all over the world is what I have seen that they don't completely rely on the electronic stuff there for multiple reasons.
Justin Grammens 16:26
Sure, sure. Yeah. Interesting. So do you do you see OCR, at some point, just completely going away? Once everything is digital? And we don't really like have this anymore? Do you see this going for?
Avinash Malladhi 16:39
I don't see this going away? To be honest, because I have seen this solution where I help them to develop it for converting it into Braille. So where there was this document that had to be converted to Braille? How would you do it? I mean, somebody had to read. So the way the easiest way was put that in the OCR connected to a brain, I mean, I'm just saying that very superficial layer of it. Sure. Put it in there, it's gonna read through it, OCR is going to recognize it, and there will be a validation that you can do on it. And then it gets generated and bred. When I was trying to help them develop the solution. I was actually very happy and proud of myself that it is something and a product that we are developing that that helps people who cannot really look into it. Like they, obviously they were unable to read this. But now that is not they have that ability to do it. I don't see how else can that be done? Somebody has to read through the whole document for somebody to do it. And if it just imagine somebody had read through the Odyssey and created into Braille. There are options. But yeah, but I think this is the easiest and this way, I think the OCR isn't that was going away for next decade.
Justin Grammens 18:01
Yeah. And I was just sort of thinking I mean, there's everything it with regards to when you buy a home, for example, right? There's a mortgage, there's just like hundreds and hundreds of pages that are still no one's been able to get that into a digital form. And a lot of ways,
Avinash Malladhi 18:13
at least right now, it's not this decentralized. So no government in the world is that heavily digitalized, where they can go and do everything in a digital format. Still, it is paper format, which I personally have digitalized them. So everyone. So I think the basic step, the baby step was to scan them. We have I think not all of us have a scanned copy of each and every document that we have the OCR is the extension of it, where I can, if I were to search for some legal term that I wanted, because my insurance claim didn't go through. I'm not gonna sit and read all the 150 pages document that interests me, right? I'm going to try to search where is that clause? Where is it? And if it's just a scanned document, how am I going to search for it? Until and unless it has been OCR?
Justin Grammens 19:01
Yeah, they're building it right into the camera app. You know, like, on my Google phone, I take a picture of something. And it's amazing. It's like transcribe, or you know, basically OCR, I'm like, Yes. And then it just it finds every word. I mean, it's amazing.
Avinash Malladhi 19:15
Yeah, same thing. A few. Probably it was introduced last year on Apple. I don't really remember the timeline. But that was really fascinating that now you could take a picture. Hard preset, now it read captures everything. Now you can copy paste it into anything. Right? And that makes life really easy. For most of them. It just for me it does because now I can just document and scanning it. copying the text sending it over saying, dude, this is where it is in a document. This is what it says I don't have to type it for 10 minutes. Yeah, yeah, exactly. The life gets easier. That's what I say. From all the aspects. Yeah.
Justin Grammens 19:55
So to bring it back to artificial intelligence, I guess you know, one of the things I was thinking about out, as you were talking is, you know, I think AI is really good at a number of different things. I mean, one is is things that humans aren't good at, which is really meticulous, you know, precision on a lot of things over and over again, I can see how you can make a mistake by accidentally typing a zero by, you know, by example, right? So, perfect solution for an AI to sort of take this task on,
Avinash Malladhi 20:22
it doesn't equal into a typo. And it doesn't have to be one single thing I'm in, I come from a hardcore business perspective. So I'll say, for example, I'm creating a product where I'm writing a weight limit for on it. What if I put an extra zero and the whole configuration of the product changes? Right? And if I don't realize it, then and if I just push it for manufacturing, what is the done effort out of it? What is the ripple effect of it? So as a matter of fact, as you said, AI, now the AI component in it as like, I can put in the value. Even if the OCR let's let's suppose even if the OCR the old base, OCR, I'm talking about the decade old to see our systems where it was not calculating it, right. Or if it was not recognizing it, right. And it was getting passed through it was, it was just getting passed through. Now with ML and AI built into it. It has its own set of validations built so that it says, okay, know, the last time when such an order was placed, it had such and such capacity is built into it. Of today, you're putting something which is way over a new tolerance, give a alert to the user, the alert, the user can look at it. So and if somebody reads to some order, and if they don't, if they say it's not the right one, it can be routed back to OCR to retrain it manually.
Justin Grammens 21:57
That's what I was gonna ask you. So you guys, I mean, this is what we were doing back in 2015. Avinash, we were we were actually having human in the loop. We were basically using Mechanical Turk. And we would basically have Google vision, give us an output, but then we were checking it as well, and sort of trying to improve upon it or you guys, is that still a concept that you guys are using to sort of improve these?
Avinash Malladhi 22:20
I would say it is still in use for sure. But not to the extent that it was in the past. There are so the levels are like, first the mission does it, then the AI checks that and then it is like we'd have trained the template based so much and so much that only the outliers, which I would say about two to 3%, you it cannot be 100% automated if I say if it's 100% automation, just it's not possible. Possible. Yeah. So when the outliers come in. And that's when the human aspect comes into picture so that the human can train it in that point in time, or for most of the solutions that I have worked with both companies that I've worked for, it has gradually increased from a bar, when I initially was full fledged into it, I saw about only 70% going through, but now I can proudly say that systems have developed so much. And I've been able to contribute so much to them that I hardly see five to 6% of the output being flawed. Which certainly is not just because of one aspect, there are multiple aspects and the reasons behind it, I would say it includes pride from the quality of the paper, the quality of the brand. So the learning aspect. So it's a combination of multiple things.
Justin Grammens 23:44
Yeah, that is one of the things that we saw a lot of is is you know, people not writing. Well, either poor poor penmanship, right, or you're right, it's a physical piece of paper, so it gets wet, you know, or you can shine it can shine through on the on the backside of it's not thick enough, you can start seeing content or images from the backside of the paper when it was scanned.
Avinash Malladhi 24:05
So now with OCR being improved in the way that it is being built, now even being able to identify that difference. So we call it a confidence factor. I think everybody calls it a confidence factor that is built into it to see if it is an image or is it something else with the quality of the paper? Or what is the issue? So that is how I am seeing how it's getting evolved? The more the confidence factor, the more the training, the more how should I say the more you process, the more volume of data, the data sets as the datasets increase. Even those things that we think are obvious are going away. I mean, as a human remain on we read it but the machine is able to do it. That's more funny than it is. I have customers who ask it looks like this to me. How did the the mission said so I'm like, Yeah, different It's additional aspect was as you have explained it to the business on how it works, because everyone these days want to know how things are working. And that is good educative for them also,
Justin Grammens 25:13
is an interesting concept. So this whole ability of Explainable AI, right? How can you actually explain what what's what's what's happening to people that are set a business, because he can't just be a black box who can just put something in and then Burton and take verbatim whatever it says that ever comes out says, well, that's gonna be what it is, you find yourself having to kind of talk a little bit about some of the reasons as to why it might be wrong or Yeah,
Avinash Malladhi 25:36
yeah, sure. I mean, you have to previously, there was a clear distinction between a beer business user and a technology user and a technical person who's tech savvy, and who's not at all tech savvy and stuff. But now everybody is tech savvy, hardly seen users who are not tech savvy and who, and who are not asking for an explanation. You have to Amman, I have helped Southwest Airlines in the past the users Oh, there are so tech savvy, they have asked me good enough number of questions that actually gave me also food for thought. One, however, yeah, I'm in. And those are the challenges that keep you actually driving your solutions. As a matter of fact, now I work for Calumet specialty products. It's an again, an oil and gas product. So an airline's and so we may think that oh everywhere, it's the same thing, not the accounting is not the same everywhere. Nor is the finance nor is the production not nothing is same when it comes to the documents, it's totally different. And each client requires a different kind of solution that they want to be based on their business processes. And that kind of keeps us on the edge to evolve even more with technology and provide them with a solution that is interesting, both for the client when for us, and I'm the kind of guy who would stretch it out even more and that's fun. It's just technology. So yeah, well, I
Justin Grammens 27:07
was gonna ask, one of the things I do ask is is like, most people in the program, what is the typical day in the life of, of your job?
Avinash Malladhi 27:15
Typical day is mostly working with the users to identify such aspects in there where they say, now I have this invoice from certain country, which I've never received before, or have received this document which the OCR is failing. So now we have a solution, building the and now they want us to include this. Now, a typical day we'll go and identifying if it worked, why did it work kind of things it tends because now we have not taught such a template before is the system evolving is the system reading it through is a system self training it to a new template. And obviously, if it fails, we know kind of reasons why it fails. Some of the days, it's more for me that if it's a completely new document, and if the system is able to successfully read through it and produce at least 80 85% of the result, I would like to go into the weeds of it and see how the system has learned. How did the code evolve to self learn it that is more fun to understand these days, then to identify what failed because kind of failures are expected. The success is something that Oh, wow. So it's an auto generated code now, okay, it is self learning. That kind of am mostly spending my time to research and analyze and understand these days. Gotcha.
Justin Grammens 28:40
Well, good stuff I played around with OCR, and just and actually more just character recognition is from like the M NIST database, right? I like 50,000 handwritten things, which surprisingly, is, is pretty good. I mean, even if you only trained a portion of that you can train a model pretty quickly. But it's mostly stuff you're dealing with, like all labeled data, do you think or or is there a lot of unlabeled unstructured data?
Avinash Malladhi 29:05
There's lots of unstructured data to be honest, because things don't come in as they are supposed to. Right? Yeah. We think the data set is going to be perfect. No, the data is as it's far from being perfect. To be very honest. It's far from being perfect. And lots of data is mostly unstructured these days. I mean, I'm working on.
Justin Grammens 29:28
Yeah. And so is the ultimate business goal, I guess, is to probably, you know, kind of eliminate humans from doing work that they're not good at, and then put them on other tasks and maybe improve efficiency, I guess is that the general thing most businesses are looking for?
Avinash Malladhi 29:43
Yes, that's for sure. Efficiency is the one I wouldn't. That is the word that everybody looks for. Because obviously if there is a process that has been either slowed down by human interference or a boring task, I would call it if it's a boring task. For it's a very manual step oriented tasks, that is where everybody is looking to include all such activities. Everything that I'm seeing these days AI oriented, somehow or somewhere AI is being used, and OCR is no far from it. It has exactly what it is today. I call it AI powered OCR. So AI powered OCR, I think somebody sent you a coining a term there. I'm like, I don't I just call it AI powered OCR, maybe others are also using it, but I've seen it, it is powering it. My concept is, it is a different power that is being provided to work with because of which I myself have seen a good career opportunities. As a matter of fact, and I have been talking to students at a few universities, where I talked to them saying that you are an accounting field, and you don't any typical carrier for you is either going into accounting or a CPA side. And guys on the technology side, you don't really think about this aspect, because this is no different. I mean, you're just looking at from a technology perspective. Now, when I do a seminar or a webinar with such students, I try to bring them together and see what if you could do this, although this is not part of STEM, think about it. It's a good discussion. And when you do it with the group of the grad students and stuff, the ideas they come up with, man, that's really interesting. I have had some good nice brainstorming sessions with those students and that few universities. And
Justin Grammens 31:37
yeah, as we're talking about before we started recording, one of the things that I do like to ask because people like, Hey, I'm a brand new graduate student coming out of school, where do you think I should go? Like, obviously, people are learning some of the stuff in school. But I mean, a lot of it is probably just more self taught, right? You just going out to conferences or books or anything like that? I mean, is there any places you might advise people to look for a lot of this information?
Avinash Malladhi 31:58
A straightforward answer that I may have for this because this is an evolving technology today. So there is no one shot one stop shop for everything. Take me for example, I start I when I started, I just did my MBA in accounting. And when once I did that, I still was so interested in into accounting, but at the same time technology was at its rice. So I combined it. And then I did through SAP books. I did my SAP certification. And then I did the OpenText vendor invoice management certification, where I learned more about how OCR is being used, how it can be used for invoices. And slowly then I was actually working on a different project to learn more about it. And the same time open text also, probably that was the time in the world when people were elaborating more on the technology with open text also came up with a solution, which was more for other business processes, and not just for invoices. So the timing matched for me. So I started doing that as well. It kind of helped me. So that's what I would say, the recent grads, that there's lots of niche technology or nice subjects out there, which they can really explore.
Justin Grammens 33:11
For sure. For sure. Are there any interesting applications that you've seen recently? Maybe not even on OCR, just maybe just in general, I guess he may have read about in the news. You're like, oh, that's an interesting application of of artificial intelligence
Avinash Malladhi 33:24
just from last two days. Sundar Pichai announcements with the healthcare with everything. Those are really Mammon. Hopefully that will revolutionize the way that we, we are going to live forward with our lives. But those things are really too much to even thought of. I remember I don't know if anyone remembers, but there was that cartoon way back while Jetsons? Oh, yeah, yeah, the Jetsons? Yeah, the Jetsons. I always think of em. And whenever there's an advancement, I just go back. And I think, man, we saw that when I was a kid, and we never thought this could be a reality. And now, it's truly a reality. I mean, you said you press a button, everything's delivered to you. They had probably 10 years back as on the sun and home improvement video. Because one, he actually designed his house, conceptually taking inspiration from the Jetsons. And today, you don't have to design houses. It's already made. So last, but not the least, my friend ChatGPT.
Justin Grammens 34:25
Yeah, we haven't we haven't touched on us. But the biggest game changer I think here the past six months. Do you? Do you see that? Again? I think it's there's lots of aspects where it can you could even throw an image at it and say, explain to me what you see in this image. Right. And so I was thinking about that.
Avinash Malladhi 34:41
Yeah, that has really changed at least men I would say it educational level it has, but I'm not just charging it as a matter of fact, the bang and barred and whatnot, everything. I mean, everybody's coming up with a new thing. These three are just the popular ones. You have to call it and chat GPT just stirred up the pot too much. And people had to come with their own stuff. Because just touching on that aspect, one of the other companies that I helped altered, so now they're using those API's to get those chatbots in there. So that's a company which never heard of anything AI. Now, that's an AI chatbot in the and that is fascinating. That is how much I think ChatGPT has impacted. To be very honest. I don't think anything else in the recent times I was reading this said, if you thought technology evolution was something that was going to impact your life. Just wait until AI gets to its full bloom, which is obviously not in the near future. Maybe I don't know. But those are the things Yeah,
Justin Grammens 35:49
it's certainly getting better and better for sure. For sure. Yeah, I
Avinash Malladhi 35:52
hope it's, it's getting better and better. Not for the doomsday. Sometimes.
Justin Grammens 35:57
Yeah. Are you are you concerned worried at all about what's what's coming down the line with AI?
Avinash Malladhi 36:02
dimes? I am actually because everything, as we all know, anything in excess is bad. But who will determine that excess? Something that is excess form isn't for you. But at times it is really, it is dangerous, to a certain extent is what I think I mean, I did a column for a newspaper in India, where I was actually mentioning about that I was like, it's a good to have it embedded into every every nook and corner of the world. It can improve the way of life and everything. But without a certain regulation or without a certain because I come from a finance and accounting background. For me regulations are something that I deal with day in and day out. Because if I'm developing a solution, if I don't develop an OCR solution with PCI compliance, for example, for banking, information, compliance, governance, my solution is not going to fly through with the business. In the same way, I was just looking at this, except for the European Act that was passed a week ago or two weeks ago. There's no strict AI laws anywhere. There needs to be something. I mean, people who are saying there needs to be regulations. I believe they know something more than we do. Probably there's some other products out there. I don't know. I don't want to be a conspiracy theorist here. But yeah,
Justin Grammens 37:25
sure, sure. But yeah, you never know. Yeah. Yeah, for sure. Where do you see this being in 510 15 years or so it seems like there's still always gonna be some sort of need, I think to translate something from the physical world, that's written down somewhere into a digital form
Avinash Malladhi 37:43
a hiatus, see this going at least for two more generations, if I have to call it because the small example you have old photographs of something or really from your grandparents time, and you want it in there, you have some documents that you unearthed after 20 years down the line. So I think this is going to stay this is not going away for sure. The usage may go down gradually, with more digitalization coming up. So I think from books, we have all graduated Amazon Kindles, that is one like a regular use example. But there's always this as an accountant, I always try to write things up. Yeah, I like to write it down and then think about it. Now, what will I do once I've written it down? How would I get it into the system? These days? Just take a picture, copy and use it what Apple has a lot going for itself? So yeah, I don't think it's going away. Probably it will, it will improve more, where if I write a number, it will compute it automatically for me and give me the result.
Justin Grammens 38:48
So there will be a lot more intelligence sort of built around it. That's interesting. Yeah. Once ChatGPT reads all the books and stuff, I mean, my wife works for a company that publishes legal books, right. So that's, that's what they do. And there is a huge plant out by her building. And you know, everything's online. But there are still many, many firms that just want to buy the physical books. It's a huge wall of books. So yeah.
Avinash Malladhi 39:12
Oh, yeah. I mean, yeah, that is also that it's also a personal preference, if I may call it, because I personally love to read a hardbound book than reading it from Amazon Kindle. So if I'm going to research, I rather have a year fall at the end of the page so that I can refer it back when I'm doing a certain other pitch. Sure. It's yeah, I think I get a copy. But it's personal satisfaction, I
Justin Grammens 39:37
guess, for sure,
Avinash Malladhi 39:39
for sure. Well, how
Justin Grammens 39:40
can people find you
Avinash Malladhi 39:41
though, I'm available. I'm active on LinkedIn, very active on LinkedIn as a matter of fact, so I can be reached through our LinkedIn. And that's the easiest way. As a matter of fact, every episode
Justin Grammens 39:52
has like liner notes. So it has a transcript of all this stuff, but then we have notes with regards to everything we sort of talked about here. We'll make sure to put all your cards Taken information from LinkedIn. And there, I have an issue. And yeah, I just thank you for, you know, sharing your experience. That's what this show is all about is just getting people that are doing very interesting stuff in all aspects of artificial intelligence. I mean, OCR, I don't think I'm ever going to run out of topics. You know, when regards when it comes to applications of AI on this on this podcast, everyone always has a new thing. And OCR is is like I say it's kind of old. But it's also new. I mean, it's gotten it's just there's no way our society and our commerce and our financial systems, our post offices, none of that stuff can function without actually be unable to do it. Because it's just too much information for a human to do so. Well. Yeah. I appreciate you sharing your you know, the impact that you're having and the the cool work that you're working on. Thank you. Thank you, Justin. Thanks
Avinash Malladhi 40:45
for having me here. It was nice setting it. Yeah. Yep.
Justin Grammens 40:49
All right. Well, we'll have you back on the future. And we'll we'll talk more we'll see where things are at. So thanks again. Thank you.
AI Announcer 40:56
You've listened to another episode of the conversations on applied AI podcast. We hope you are eager to learn more about applying artificial intelligence and deep learning within your organization. You can visit us at applied ai.mn To keep up to date on our events and connect with our amazing community. Please don't hesitate to reach out to Justin at applied ai.mn If you are interested in participating in a future episode. Thank you for listening