EP 6: How AI, Data Science & IoT Go Hand in Hand w/ Jags Kandasamy

ABOUT THIS EPISODE

To say that the internet has exploded is a vast understatement.

Today, we generate 22 exabytes in 4 hours — which was the size of the entire Internet in 2002.Yet we only use 5-10% of the data we collect.

In Episode 6 of #UnleashIT, Jags Kandasamy, Co-founder and CEO at Latent AI, explained the future of data science and the explosion of IoT.

We talked about:

  • Making all devices intelligent — even offline.
  • How to become a data hoarder
  • Whether your org needs an AI ops role
  • What constitutes true digital transformation

A couple of resources we mentioned during the podcast: "HBR on not waiting to adopt AI" and "The Economist bills data as the new oil".

To hear this interview and many more like it, subscribe to the Unleash IT Podcast on Apple Podcasts, Spotify, or our website.

We create these systems to enhance our ability to interact right. So Ai is augmented intelligence from my books. Welcome to unleash it, a podcast where we discuss the experiences and ideas behind what's working in enterprise architecture and digital transformation within the IT landscape. Unlock Your Business has digital capabilities. Transform your enterprise architecture. Unleash it. Let's get into the show. Hello and welcome to this edition of unleash it. Today I'm very happy to welcome our guest, Jazz Katasani. Jags is a CEO of Laton AI, and we're here to talk a little bit about artificial intelligence and how that's having the impact on it organizations. But to start, Jaggs, why don't you tell us a little bit about yourself and a little bit about Layton Aih tharting. Thank you for having me on this podcast. Very excited to be here. I said the leading day I is technology spin art from Sri International. SRI is a book plays of the Computer Maeus and Serie, among others. Late in day I we bring developer tools helping ai developers bring models down from the cloud to run on computer constrained devices. So you can run ai on sensors all the way up to the cloud, on every device possible in between, right. So that's what we do. Think about it right, like if you talk to silly today when you're offline, you will not get any answer unless you're connected to the Internet. So we are trying to change that. We're making every device intelligent, even offline. Okay. So, hence the latent ai exactly got it, got it. You know, there's some world wide spending numbers that are out there that there are pretty incredible when you think about it. I mean AI has been around for a long, long time, but right now, all of a sudden it's getting it's so joe, so to speak,...

...is that's a good term. Why do you think that is so? A couple of things, right. Like, AI, of all, a has always needed a lot of data. MMM, since we went online in the late S, or are early s to late s and the last twenty years we've been collecting so much data about a lot of things. That data is fueling the training algorithms to learn. Machine learning means machine needs to learn from the data, and you've got a lot of that data. And then, with Moore's law, our compute capacity has increased, doubling every two years. Right, so the compute capacity, along with the data, is what is fueling the AI growth that you're seeing in the recent years. Yeah, and there must be other technology. Is certainly things like Iot, Internet of things, you know, transactions and e commerce sites. All the data are collecting is having a tremendous impact on this. Do you see that? I think specifically around I ot in devices. Yeah, so here's my famous quote. Number quote that I have is that all of the Internet was only twenty two x Avite in two thousand and two. Today we generate that amount of data in four hours. Then you believe that, right? Like you, you get six internets in a day in terms of data. IOTI is what is fueling that. Right. If you look in we passed the analog to digital conversion. Basically, you have a thought, you you're speaking something and you're typing it using a computer mouse or using a microphone to record things. All of this required active interaction from you. But today, if you look around, there are cameras that you don't even need to activate. You just walk by if it's continuing to record everything. Right, microphones that are picking up like for example, your Alexa or or your iphone and stuff. It's continuing to hear everything that you're saying and the analog to digital conversion is happening in an uninterrupted, seamless way. That's why the data explosion is happening. IOTI...

...is one of the contributors. So how have we been handling this data explosion? I mean, is it forcing us to know, to more of a cloud architecture? You know, is that migration from on tram being accelerated because of all the STATA that's needed to start doing ai? Absolutely right. That's where we started off with. Right, we started doing things and it was the client server architecture and data was wherever it was, and then we said like all right, let's form this big data lake and push all the data into that and then people can go and do their analytics and do all sorts of data signs around it too, to drive intelligence or build ai models or other analytical models out of it. That worked for a while and then people started realizing that ninety to ninety five percent of the data that they collect nobody's using it. It's useless data. Only about eight to ten percent is is much more a signal time data. That's that's driving business intelligence and business outcomes. Right, so that is starting to slowly come about from as a realization. So see, IU is considering, why should I be collecting? Why should I be a data holder? Right, data is good. Data is the new oil, as economist protect. But why should I be a data holder? Well, where do you put out the data do you want? Do you need to collect all the data? First of all, yeah, there are privacy issues, there are security concerns. You collect all the data, you sitting on a pile of data. Your attackmaker just opened up wide. Right. How do we reduce that? That's where AI comes to play. Right. The whole idea of Ai is, hey, can I cut short the amount of data that I'm collecting? Data's getting created automatically. Can I only filter and choose the right data set that I need to store that? That actually makes business intelligence for me, for my business processes and my...

...outcomes. Hmm Right. So by AI doing that, by them only identifying that data which makes sense for them to work on to to reduce these positive business outcomes, are the number of models the AI professional data scientists is looking at. They expanding. You know, when you have. Early days I would excited we had like one or two miles that would work on a whole data set. But now is we're starting to segment the data in more interesting ways. We must be developing a huge amount of new models to really better understand and train that data on. Absolutely. I'll draw parallel to the human resources in an enterprise. Right, okay, if you look within a particular organization, let's say sales operations, right, you will have different teams doing very specific functions within that organization. Right. If you look from the larger part, okay, sales operations is helping you get your sales people enabled and train them and then help them close the sale and and the deal and all that stuff. If you go with an individual organizations, it is somebody that's managing the crm. That is somebody that's it's managing that the pipeline and stuff, somebody helping you close the contract and stuff. To each individuals are doing separate work within them. That's how the models are getting siloed and getting down to further and further granularity to run more efficiently and also to get better intelligence out of there. Ye, but how do you manage all those models, different models across an organization. If I'm a CIO, all of the sudden I need like Ai Apps, you know, where they're basically looking at my artificial intelligence. It's is as it's operating across, you know, a multitude of the applications and processes in my organization. Yea. So, you know, we were introduced to the term develops the recent past. Rate, AI OPS is becoming the next big thing. Yeah, what would those people do?...

Actually, I mean de APPs. We finally all understand, you know. Yeah, we're about, you know, releasing, getting out, you know, new software and services. But what is the AI APPs going to do? So Ai OPS, from my point of view, is going to consist of data science folks as well as operations folks. Those are those are the two. They'll come together to form the air ops. So you're going to be looking at, of course, the streams of data that are coming in. How is your model performing? Right, there is a business as usual process that's doing like all right, this is my threshold. This is what I expect these ults to be. Is My AI model operating within the threshold? Is it exceeding it? Is it is there a normally there showing up that we didn't catch with our normal business operation that the AI is catching or the AI is going completely Ri and and and going in a completely different tangent than way we intended. So those are like few simple examples. And then if new data types are coming in, you how do we route that into our training process? So we have a complete cyclical training, deployment and inference methodology developed and deployed with them. An organization is a cremasure. Are Using something like an f measure, which is used for n LP, you know, to see. You know what we call and frequency. Is there sensor of measurement like that in AI? There are different, not sophisticated tools yet that that's still under development. A lot of startups are looking into that, I including us. We have some work going around there. See, if you look at from a AI perspectively, you've got time series data, you've got visual data, which is like computer vision type pictures and stuff, and then you have an LP and then you have some textual data as well. Right, so each one will have a different measure and measuring tools that you will need. so I anticipate that the skill set of the person running these things, that the data scientists. As you're getting more and more closer to each of these different models. Is Different tools, you know, data...

...sets that are coming in the road. The data scientists actually, as you have, a much more specialized as well. Yes, the data sign does need to be much more vertically aligned. They need to understand which business that they are supporting. MMM, can't be a generic data scientist and be a Jack of all prints. You can, but you need to get specialized within each one of the business functions that you're supporting, because without knowing the business, you don't know what the data is telling you. So this is really an area to where it because I assume the AI porson will fall under the the it organization. This is getting a business managers involved as well. So I'm a marketer, I want to know as much as I can about my personas you know, what type of data, what triggers them to make a purchaincing decision? Are there trends or there's certain steps buyer takes that I want to be aware of so I can fine tune an optimize that system. So you might have a data scientist for each one of those lines of businesses. Exactly exactly you would need a let's say an AI model is supporting the operations functions of all of your marketing department. Right, that data scientists need to be tied to the hip the way you think, the way your organization operates, the way organization measures success. HMM, right, those are the input criterias for the data signed to go build the models and to work on data that they're looking at. Yes, so this is really a big part of an organization's overall digital transformation. I mean sometimes really think of digital transformation. We think just to that, you know, a business going from on front to the cloud or adding digital business and all this is this is this is even further than that, though. This is really, you know, the yeah, taking taken a further step on. Yeah, absolutely right. It's a simple thing, right, like look at on a actory floor. Yeah, right, process automation and our operators are there. How many factory floors...

...still operate on a clipboard basis? Right, they're taking a clipboard and then they're there their noding down the staff and whatever, readings and and and metrics and they go back at the end of the shift, they push that information onto a spreadsheet or onto an APP. HMM, at the end of the day, that is not digital transformation right that? Yes, are they digitizing information? Yes, they digitizing information, but a digital transformation there would be like, how do I capture that information in a timely manner and it is not subject to a user or an operator error? HMM. Right, there, with the amount of data that is being generated around one asset in a factory floor, right, there will be at least ten censors of operating over there. A human can possibly not record all the ten cent sensor data at every given or every minute or every five minutes, at ten minutes, right. That's where AI comes into play. You can collect all the data and then start to infer from them to the sense of fusion and staff and try to understand what the machine is doing and what are the the heartbeat of it and continue to record it. We have the human done exactly. That is the true digital transformation. Let the human do what he's capable of doing. We are a high intelligent being. Let us do that right, rather than being the data entry operators in front of a machine, because that is a big fear that you know, the machine will take over the human, where in reality it's freeing up the human to do more human, challenging activities. Exactly. For me, the AI definition itself as augmented intelligence, not artificial intelligence. Right, it's augumenting. Yeah, we don't create systems for anything. We create these systems to enhance our ability to interact. Right. So Ai is augmented intelligence from my books, so I read somewhere to that.

I mean, there's tremendous amount of investment going on in our defition. I think worldwide spending in two thousand and nineteen was up to thirty five point eight billion. It's targeted to do about seventy nine point two billion by two thousand and twenty two. Are we seeing the return? Or is that return with all these investments, going to take time? Like, are we coming better at what we do through ai or we just we are we still in the NACENCY that we're just trying to stand up these these things in the hope that they'll bring a large amount of I am an internal optimist as far as technologies concern. Others I wouldn't be in this field. I think the returns are slowly showing up now, but the returns, we have no idea how big the returns are going to be. In the near future. It is going to be from my from where I sit and see, it's going to be tremendous. One of the new terms probably you heard of, is edge computing. Yes, so what is edge computing? I like we've got these data centers and if you look at a ws or Azure, Google cloud or any of these datas and they're all in remote locations. They are far away from the city scape. Most of them are in Nevada because of the low power right. Or Microsoft, as I think, deployed some data centers under the sea so that they can control the the cost of cooling and stuff. All of this is away from where the users are. So every time we talk or we do something, it is to ping farther and farther away to get connected. Right. There is lots of physics that involve here. And and you know the the speed of light is only so much, right. You cannot improve that. So this is where edge computing is helping us with putting compute mini data centers and small computing powers around the city where the users are. HMM, right. So this is going to help further in reaping the benefits of AI and other investments are going into this right man, instead of you talking and sending your data all the way to Nevada, let's have from...

...from from the East Coast to Nevada to get a reply back. What have you just sent it to your city's telco hub and got a response back. Yeah, they make it much, much quicker. Exactly. What did some of the other hurdles that you know when you're you're certainly bring ai into your rezation that you can run into, whereas there's some of the you know, things you can like the start for me that you know ceio might run into. So for of all, I would I would say this. Howard did a Howard Business Review did a study. And you know, in in normal technology adoption there are earlier doctors and then late followers. Right, HB are article talked about. If you're not a earlier doctor, you are not a fast follower in the AI world. HMM, if you don't adopt now, your business is going to be irrelevant in the next five to ten years. Reason being, AI requires data. Right, we talked about this data is generated, but way we generated this data today is for us to read from it, not the machine to learn from it. Right. We wrote logs, we wrote report all of this for the human to interpret it. We never made it easy for the machine to, you know, learn from that. Right. So we have to transform the data into a machine learnable data. Right. So the earlier doctors are already in that process of changing that, right. They creating data, they making the data readable and then they training more algorithms, more models, and they can there in that circle, right, in a fast running circle already. If you're a late adoptor for a fast follower, how you want to call it, you are behind them already. They've already got models that are tuned and running and now you just entering the game trying to learn and see what you new, what data do you have? But what would your advice st t these late adopters, you know, to get ahead? I would say start yesterday. Yeah, figured...

...out what is the key important point of part of your business that that needs transformation. Keep that well, what is the lowest hanging food that you can get done in the next three thousand and sixty ninety days that you can see value? Get that done. You will get confidence, you will grow the confidence within the organization that this has impact. Aa has impact in our material impact on our business and then you can slowly grow from it. Thank that's great advice. So, Jags, thanks so much for joining us today. This was a great conversation and I hope it sparks in our audiences to consider what your organizations are doing with a I certainly check out latent ai. I think there are anothering coming start up out on the west coast. Correct, yes, we are. So please take a look at some of the lace technologies that they're working on to and we'll see you next time. Thanks. Now you've been listening to unleash. I T to ensure that you never miss an episode. Subscribe to the show in your favorite podcast player. If you'd like to learn more about enterprise architecture and tools to help unleash your businesses digital capabilities, visit lean ix dotnet. Thank you so much for listening. Until next time,.

In-Stream Audio Search

NEW

Search across all episodes within this podcast

Episodes (24)