Courtesy of NHS Digital's Rajeev Chakraborty, below is a transcript of his speaking session on 'Are we using AI-related technologies to automate enough?' to Build a Thriving Enterprise that took place at Digital Process Automation Live.
Are we using AI-related technologies to automate enough?
Is the head of portfolio delivery of test and trace digital for NHS digital? So Rajiv, I'm going to ask you to, please turn on your camera and join us here.
There is a regime. Ladies and gentlemen, Rajiv. Thank you so much for joining us. Rajiv is, is a leader, again, in the NHS digital, where he leads technology, AI, related automation, and many other aspects of technology innovation and innovation for value creation as, a broader as, a broader topic. So we're thrilled to have you with us, Rajiv. We're very much looking forward to your insights, to the principles. And fundamentals that there that you apply when you look at digital process automation. Thank you again for taking the time to share your expertise of our global audience today.
Thank you, Julie, appreciate it.
... guilty as charged, I work in the NHS. And I've been working in several different organizations for quite a long time.
And as part of that, you tend to get used to the idea of, in many ways, how people see technology, how people understand technology, but more, more specifically, how people adjust to technology.
And for me, that's quite an important bit the last bit.
how we adjust to technology.
So if you think back, if you went to somebody and said, I would like to be in your house, and I would like to know everything that you're doing all the time.
And I know a I suppose you understand where I'm going with this, is I'm talking about things like privacy.
Things like everything being watched or even otherwise, things like, for example, if I told you that, if you have to open up your door, you have to do something different. Where something as simple as this. All of us have a door lock.
And if I told you that the way the door lock is fitted, you'd need to have a screwdriver you put the screws in every time you go in and you pull the crews out every time we do that.
You'd look at me and go, Your ..., it doesn't make any sense.
And it does, Yet, when we talk of technology.
Well, ah, we've got this wonderful device in front of us where you have to put a username and password.
So, these are things where humans have evolved with technology and where we've started adjusting to technology as opposed to the other way around. And that's essentially what I do for a living.
I look at digital transformation, start the basic set up, what do you really need?
And then take it all through the journey and putting it into a place where the technology is stopping you.
And when you do something like that, I think the most important question is, is actually doing something that I wanted to do?
And that's what I'm going to discuss in, in the world of AI.
So, I'll talk about, what is AI, or does it actually mean, how ubiquitous it is, how? Where do you find this around you in your everyday life?
And then also, what other things that you can do with the technology that's available, and I'll keep it very generic.
I don't wanna get into that technical aspect of it, because I'm sure if you go outside, you'll get a dime 12 outside to tell you exactly, exactly how the data lake is configuring all the rest of it.
And then after that, I'll show you some things which we are building as an organization. So I own a company called a London L Hub Technologies.
Our website is WWW dot A R Y a LTD dot com.
That's REO LTD dot com or limited dot com and where we have built some technologies and mainly to show you that we don't dog food. If we say we like privacy, we can show you how it's done.
And believe you me ever, super, super difficult. So, I'll, I'll go through that little bit of a journey at the tail end of this whole presentation.
That's essentially what I'm thinking doing.
Is there anything you want to ask? You have any questions before I proceed, JJ, is there anything else you want me to elaborate?
No, that's great. Whenever you are ready to share your presentation, you can click the button and have it broadcast.
OK, so Um.
Just to confirm, you can see my screen now.
Yes, we can, but we're seeing the one that has the preview on that. So, if you go under the Sharing tab and if you can select the one that has only the presentation mode, right now, we see the presentation with a Preview of Light.
That's the one perfect.
I have, this is what I mean by technology. OK.
so, um, This is the strange thing having worked in technology for so long, yet having to know the nuances of it and so I'll get to that.
So this is us. Arjo Associates, we have a product called ... Hub. If you go into app, so Apple App Store, specifically iOS, we have an app as well. We provided for free. This is for benefit of people.
And we're going through some new and fun changes.
So that's where we are.
We mostly tend to work within the UK, but the have started working at outside India is outside the UK as well. We work to India. Look. But the US, in fact, work for a fairly big company in Atlanta. So we have presence outside, but generally, we're very focused. one of the ethos we have is we work with who you like.
So, starting with, what is it when you say technology and, and then you see, there's a whole bunch of things that come in. I mean, you, you receive and capture all kinds of stuff. You store them. And then start categorizing in the technology world.
You analyze them, you predict them and then you start providing meaningful information But that's no different from what we do as humans as an observed things. And then we decide how are we going to remember them all?
Take pictures or make notes all, get them home, and keep them in a certain way?
And then based on that, we start using them to use them. We start looking at, how did we actually use them? What's the best way? If you think of the typical phone, I'd use this in a certain way. Or if you think of, I don't know, a new charger that you use in a different way, or, it could be your chair where I've decided to use this in my office content instead. Say you do that.
And then you start thinking about how comfortable as it, is it great, if you like it, you end up using it more, if you don't, you don't use it, use that at all.
So it's essentially trying to categorize the technology world in a very human way, how we react, So what is AI?
So in the simplest of business speak, so to say, and I know this is not running business, generally, it categorizes into three areas is, you take a lot of data, not sits in the world of data ops.
And then you begin understanding the data.
Meaning, if you see a date, and you see that against, let's say, a set of people entering a building, that it's the relevant just time-stamp, whereas, if you see the date, that happens to be about starting a computer using them, That's the time-stamp use in different way. So, there are different context in which you start, and then you how you, if you categorize it.
And then, how do you operationalize it. So, meaning, what is going to happen to that data?
Is that so and in that worldviews machine language, for the operations?
And there you talk about, how do you integrate this, this information set, or dataset, or the application and itself, talking to each other? How do you do that?
What's the point in having that? How do you build yourself found that basis? How do you keep that in a secure area?
How do you actually make that, that whole solution quite stable, so people can use that, and how do you put this into real-world where, if you were using it at home, you now use it outside. So it's the same thing as you prepare food at home. And if you decided to go to the market and sell that, that'll be a different approach, And that same thing happened in the world of technology.
It's how you take that, and how you operationalize the whole thing.
So, when you understand these three categories, the biggest thing, the biggest challenge that comes across, is think of ...
as the, the people of the world.
We sit in different cities, city, in different countries, think of other creatures as well. They say they live in the sea, we live in the air.
But when they haven't, they have to interact with each other. Some things are common, some things are the way to move, the way they react. Those things are common. And there are others where it may not be common way. It could be the language. So if I said something to someone in my native language, I grew up in India, and I speak in Hindi. they wouldn't understand unjustly on the streets of London.
There might be some people who also might not.
That's where the challenge comes in the same thing in the world of technology. So when you have some dataset that comes out of pull, and there's another one which comes out of Android.
And another one comes out of, I don't know, from Windows.
Yeah, of course, everybody loves Windows, Uh, sorry, I had to take a swipe policies.
And when you take all this data and you let it talk to each other, it may not necessarily talk to each other. And that's a fact.
Because some something as simple as, in English, if I say gender, gender can mean man or woman for some people. For the others that have more categories, and fairly. So, everybody has a right to choose that, so they have many choices. Same thing in the world data. You might write the data, but as I don't know, 1, 10, 96, and another person might write it as 0 1, 1096 were chosen British. Or you could write it as the month, date, and year. And you might use two notations to fall.
These are small little things that complicate that, but far more complex things where one system will only give you data, which only they can understand, no one else can understand.
And then imagine another one, by they'll say, OK, we'll give you the data, but it'll be encrypted and you can't read that.
That's where the engineering aspect comes in.
How do we take that data? How do we structure that?
How do we engineer it?
So that, one, it's the least hassle, which in the world of technology, Auggie, quote, loves complicating things.
So, idea would be, How do they take that, and how do we not make it complicated to run it. And that's the ideal way to do this.
How do we keep the data in such a way that, like humans, you can still stay in your own space, and still be able to talk to each other, so a translator that sits there.
Then, how do we move dataset from one place to another? So, when you travel from one country to another, you're going through the, the, the airport and the Visa anabolic controls and so on.
So there are controls there, which identifies you, and then tells you you join this. Q So if you come to the shores of Britain, it'll say, Well, if you are from Europe joined this, Q and if you are from UK, join this, Q, If you're from the rest of the world. You're lucky.
That's exactly what you do data as well, end up categorizing it.
Then you have, how do you store the data? That could, there could be something chabal where all the data comes in and you just put them altogether.
But then, there will be data lakes where you don't know. So this could be, example, would be, where there's a march on a street.
And you have all kinds of people coming in. You'll say, OK. All of you go through this. And eventually, there'll be segregated.
So that's what data, make sure that they will take everything, put everything in one place, and then start segregating eventually as systems start understanding it.
Then you bring it into the world of business, understanding and say, OK, I have this data. What does it really mean to me?
So you could have simple way of thinking of it is fatal data.
And you've got all kinds of payroll data coming in.
The question would be, what is the benefit of me analyzing it? What do I, what am I trying to achieve?
And that's a very, very, very important point.
Does one way you think of tech, where you'd say, I'll get some graphs and don't give me the answer.
Hmm, hmm. I don't know that that's not really why they use technology. Technology is there to augment human thinking, not the other way around.
So, then you'd ask yourself, What am I trying to achieve? I'm trying to understand.
What's the gender pay gap in my company? How long have I had it?
Having done some things in the recent past, have I made a difference? How I actually made anything, Which makes it easier for my, my colleagues, and my, my employees to feel better about themselves?
Or, um, if you haven't done anything, how do I know?
I need some data to drive my decisions? Not just, I think, so one example is imagine a company what you love, and this is a fact, I wouldn't want to name names.
You have lots of people, well, different offices.
You sit there and the gentleman said today and he says, he's telling me that? I'm sure when you do the analytics, you'll find out that they are very, very good in terms of D&I, diversity, and inclusion. So I asked him, What do you mean? Like, for example, cintas of just hire three women in our organization, when we were looking for five, so that was a majority higher.
And that's fine. I'll take you four, Take your word for it. You're paying Meeks, Oh, yeah. I'll believe you. I'm not going to say no to you.
And then we went in, We started shot, looking at the data.
And he goes through the whole thing. And you let machine drive analytics and understand what that means.
We find out that those are the only three people they hired in the last whole year.
And that assumption came from him because he thought he was driving it.
So where he was having that visibility, he made it back, See other places he didn't really.
And he didn't like that.
Now what that tells me is that that's not data driven decision.
That's a warm fuzzy feeling you get, which is governed by emotions. Emotions don't always give you the right answer. That's why they are emotions.
And where you bring logic, and that's different logic is based on evidence on facts.
That's fair. Business understanding comes in.
You're saying, I want to know whether we're really D&I compliant, or the good for diversity and inclusion, at that point, you let your data tell you whether you really either.
And when you look at that, then you see the overlap in the way you've captured the data, and you understanding the data.
That's when you start looking at the data and say, hmm, it does tell me these things, but I'm not necessity, capturing other variables. Like, for example, did this person come in from a Challenge background?
Did he actually have university education? Did his parents have to see the UK voluntary, And when forces you to do that?
And it's completely anonymized, where people, people asked, I've been asked whether you, your parents, had any university education.
And at that point, people can choose to answer, But they don't. That's fine.
But at least, you've got a route to put data enrichment in place.
And depending on how you don't talk to them, meaning, what do you put it on the screen, the user experience of it, it will then decide, the user then decides whether they want to give you more information.
That's one way of doing it.
There are other ways of data enrichment, social media world.
Again, I can't name names for fear of being sued but there are other ways where they will look at your digital fingerprint and I'll come to that shortly, where they'll they'll buy information for outside.
Bear that inside the dataset, and put an identifier against that.
So I could be subject number 105779.
And then they'll bring source one Social Media source, do social media, source tree company data source for previous employee data, and they will form a picture that is what you do, in terms of creating that data science to business understanding.
And then what you do with that is the operationalization.
So, you say, OK, now I've got the data.
I need to get this ready.
That's when you start looking at the application aspect of it. But this is basically the starting off.
What is AI? This is what AI generally does for you.
So, there is a second part to that, which is way, how it does it not. I'll come to that shortly.
What other key components, generally, you will have, and this is something I've used from AWS where it's come and use that, You will have some, something to define the data movement.
You will have something that defines the governance. So how will you store your data? What kind of backup do you have? Do you even have any backup? Do you need to store that data?
When do you get rid of it, or do you really need to keep that this day and age? Everybody thinks should keep everything. I, I have an exact opposite thing in our company is We don't keep anything. We only use it for a need, and I'll explain that a bit later.
And then there's metadata.
Metadata is data created from data, and then use access. So, metadata.
To put it very simply, you take a set of datasets and make some information from that.
And that's metadata, so it's It's basically taking it and making some inferences on that. And that is owned.
That is an extremely important thing to understand, and that's where, in the wider world, we are the product.
And our metadata is actually the stuff that pays for it.
And then, of course, there's use access. How do you access data? Who's got the right to access it?
You hear all the time. Not as being data breaches of all sorts. What does that actually mean? It means that someone was not actually putting proper governance and user access in place, which is why my username password was compromised someplace. So these core things you need to have, whether and, to be honest, whether it's AI or otherwise, you normally have that.
And how do you store that? Is it central?
And then, on top of that, then, you start putting, you start with analytics, meaning, you're expecting results, and you know what you're expecting.
And then you train meaning, You get that system to start predicting data the way you've actually seen it, and then you start building that algorithm further. So, you start seeing trends.
You never thought you could even believed up as possible.
That's basically the components, then, let's go to what a typical layout looks like. Typical layout generally, would look like this.
He would normally taken, um, data that's it's just a data lake is why it's a huge amount of data coming in from different directions, structured unstructured.
And you could have an on premise.
You could have a real-time, you can have it on, on the Cloud. It doesn't matter what you keep it. You can secure the Cloud.
And then, on top of that, you would have Analytics running.
And machine learning is doing its magic.
That's generally what you would want to do, where it is very open and comprehensive, because it's constantly churning that data. It has to be secure. It has to be scalable.
If you're working for a large organization, even otherwise, the whole idea of churning data is wanted scale of it you wanted.
When you say durable, meaning, you want persistent data. You want the data come in and make sense of that city. And of course, everything comes down to cost.
And this is where I say a techie world gets very counter debate very, very hard way, And we make it very complicated for our own liking for our own code. And this is where, in the new world where we start doing things, which are, what we call loosely coupled, where you can take bits of it and put something else in without destroying the entire system. Microservices based. Meaning you start creating small little functions rather than creating one giant thing.
And then, apart from this, you have to make sure these are modular, so you can use one bit across different systems, and that makes it very cost effective.
That's generally the long shot.
In terms of how you do this, there's all kinds of things. You can do this. So, you have, this is just a random example I got from a finance company you're working for, So you do the sentiment analysis. You look at customer lifetime value. You can look at debt prevention. You can do all kinds of things.
Think of it as where an intelligent person would say, Look at the data and make some inferences. That's exactly what the system can do for you.
This is my biggest bug bear.
It's AI and my data.
And, let me first explain.
Wow. Oh, what? It doesn't explain the why, and, and some of the how.
We all have data permissions given to businesses, it could be for sales, and marketing, but before operation, products, stuff, and customer insights, finance all of the wonderful, fancy business terms. What does that actually mean?
It means that someone's taking the data, in all such examples, and if they've made you sign terms and conditions.
And if you, if you think logically, next time you walk into a store to buy a pack of cookies, or you're buying a television, You never have to sign over there saying that.
I shall not kill you with this laptop or the television that I buy.
Don't do that.
It's the most idiotic thing, anybody would ask you to sign.
I will not take this television and display my neighbors' pictures on this one.
Don't have asks you that when you go and buy television.
Um, I can, can can actually go on with these examples yet, when you go and go and use an app and not just once, repeatedly, we asked to sign terms and conditions.
These terms and conditions are made by lawyers and some very, very senior lawyers who've gone and understood how they can be taken to court and justify the actions that we have.
Now, there are good reasons where we need to have that kind of information available.
And we need to understand why we have that data but in this case, I think in current day and age it's mostly the tail that wags the dog. Most places, very rarely find otherwise, and this is one of the first things we did when we started building a hub called shop. And then comes to this whole thing about digital fingerprint.
I'll give you a good example: All of us use Wi-Fi.
Every device in relation to which Wi-Fi is it unique, and I'll explain to you how.
So imagine you yourself, along with your, the person. You love it. It could be, Well, of course, if you don't have an individual using technology that you different story, certainly not your bet.
In some cases, yes, you're using technology to monitor them.
There are good tools for that.
But, if I'm talking about people, Lecce, that's me, my, my, my partner, and I've got two kids.
All of them have a device and each one of us has a Wi-Fi station on.
So, when we log into Wi-Fi ..., and every time you go somewhere, you've gone and put the password in, it remembers the Wi-Fi station, that my First Nation is unique.
If you look at your device and look at the Wi-Fi stations' you've got on your device, check the one on your laptop. Check the one on your mobile phone. Go and check the same thing with your partner.
Joel are your colleague who works for you, or your best friend, or even your twin for that matters.
It will be different.
That's a digital fingerprint. That's one point.
Then you think of the same thing.
Every time you go and browse something, they give you this wonderful screen with Chase.
Except earlier, before we had the EU regs for Data Regulations you had genetic stuff and genetic means OK, and gone. Now they give you even more fancier ones. You love to choose 1, 2, 3 all.
If there are even others who are more exotic, it'll take you to another page altogether.
You have to select those, and it'll take forever for you to come back. So, once you've done this twice, you've done this, I knew this Despite knowing this.
I get fed up, and I'll say, OK, just take it belonged.
And you'd say, OK, that's when you're giving them information, like, No browser goes out.
It has a unique ID on your device.
There's a unique ID for your IP address, IE, from where you're connecting, there's a unique ID for the application you're using.
There's a unique ID for the hardware components in different ways, the configuration of it, and not to mention the cookies, so, meaning that, although at that point, if someone says that.
Hmm, hmm, I've taken your data, and I've not actually taking your personal identifiable information, B, I If you can take anything away from this session, take this away.
P I, it's a thriving market.
That's where, all your free products drive stock and that, then tells you what they need to know about you.
So in this case, they will take the PI, and they will then use that to determine certain things about you.
And that's why privacy such, if somebody told you that I'll be sitting in your house every day and watching everything you do, or if you went to a coffee shop, and they want to see when I come in.
When I scratch my beard, where was I looking, what color shadows are wearing?
And by the way, is when the person gives you coffee.
So where did you come from, and where are you going? Can you please tell us, if you don't, as, we just know anyway recorded here. You'll be bonkers. That's exactly what it does to you.
Every time you go to a website, every time you accept everything and these are not great things.
The reason it's important to understand that is because when do you do this, when you accept these things, Yo.
You're giving this to a system back to the point.
If you think of it.
Sorry, I'm going back to the screen.
Now, You're giving them data over here.
Now you've given them all the data, and now it goes eventually into this machine machine learning operations.
What do you think the machine is going to do?
So I'll give you another example, just too.
Visualize that thinking.
That is, imagine you have a child growing in your house.
You teach the child that every time that child looks at Uncle John.
He will want to walk, He's learning to walk from Uncle John.
But what you want the child to do is learn to walk, because he's growing up with Uncle John.
But Uncle John has a limp.
But you don't want to do this him to know that.
I can limp like uncle John.
It can do things like uncle John.
I can mimic every one of his behavior.
And imagine no army of bots that you don't know anything about, can start understanding you know, behavior.
And, in fact, know, things about you better than you can start predicting your behavior.
That's why it's important to know what goes in, and how do you analyze that data.
So we created something called a learner's hub.
four, and now I'm part of that. So I went in into the organization.
We were trying to build something called as SV Hub. It's a social value experts down to very big on social value.
And we're trying to build a platform there.
They said, They want to do this.
So, they want to do just capture social data.
So, I explain to them how it works, and I'll come back to this in a short while, just to give you the context.
What we did was, we said, let's take a look at the user journey on the Hub.
So people come in, they take the device, tell, define what the browsers are, then based on the role, I will go into the admin learners profile.
They'll have big companies who will work for small companies, tell them to start recording their article value data, and it feeds back in, in that way, video.
Putting anything out there in terms of corporate social value, They say, Oh, yeah, we can show you how we measured and how those dashboards will tell them how they actually did, where they couldn't understand us.
how many points could people's personal information.
So, if you're doing something ethical surely, data privacy is one of it and respecting people's privacy and, and information like that is important.
So from, from that perspective, if you think of it, the moment this user comes in, then device data can be captured.
This data can be captured when they go in and login the personal data can be captured.
When they go in and start learning, you can create metadata to understand what would they learning.
Why are they learning, at the individual level, or even look at how was which punishment was involved in pushing which one, and so on.
That's too personal, That's not why we are building the system.
We were building this system to make it more, more ethical to collect information, share with them.
So then we decided, OK, what do we do. So he said, we need a warehouse to collect that data.
We need a learning module. You need analytics. We need integration insecurity.
So the question was, how do we do this in a secure manner?
And then, after speaking to several different service providers, be realized, no way where you go, can you own the metadata?
So we'll have to hold a date ourselves.
If we hold the data ourselves, then we have to have the burden of securing it, but then also then we have to make sure it's integrated with other systems, so then we built this ... Hub Platform. That hub platform to be eight years to build.
And the reason why it took me eight years when I started off, I couldn't find anything where, if I loaded something onto a system, whether it was the unknown software of the world, I didn't know how much of it would stay secure.
So we had to build our own secure walls to make sure anything that comes in can stay that way.
Then from there, we decided, OK, you will have insights you are consulting. But that means it's truly obfuscated.
And what I'm trying to say, Oh, we heard about this one is when I say truly obfuscated, it means that this particular thing will always report back on group data.
Um, specifically, data that cannot be identified against individuals, because the algorithm will stop stop doing anything that starts correlating information to create digital fingerprinting.
That's very important to understand.
Because most applications that you do on site, only reason why they do this is for user experience.
But then, honestly, if you ask yourself, and you say, user experience, is that really user experience that we're talking about?
Are we talking about maximizing our sales?
And then it goes back the same logic that if I go to A a coffee shop, then I'm going to buy coffee.
And if I'm selling shirts, I'm selling loans and I'm selling on watch, then I will not focus on selling the best coffee in the world.
That's a very important point to understand, because then, if I'm not selling anything else, I'm only selling coffee.
And my platform only lets you sell coffee that I will want to sell the best coffee in the world, automatically my entire value chain changes.
And that's the ethos we tried to build. And the integration platform, especially Hub, where, that's what we do, will give you the information that's relevant to you.
There are places where you can identify, but those are very, very specific ones, where you want it to be done for specific use cases, and there'll be a different process for the different user journey for them.
So, what can we do with AI?
You've done, I will give you some examples of what we've done, things, where you take just basic things, and you take management information systems, your data data, you create, or a wonderful graphs, Excel does, amazing.
Power BI Desktop, Bartley machine, So you will have that basic stock, and you have described the data. This is how it looks, and what it is, and so on.
Then, you can go further down and say, mm, not just this.
We can probably start looking at a bit more in terms of where we say, unsupervised machine learning, so you understand the data you explore. What is the art of the possible? And you do the different analytics within that.
And when you start doing that, you understand certain trends, but alongside that, also looking at supervised and unsupervised learning, which is where you start doing predictive stuff where, if you say that if I take Rajiv ...
picture, and he has this expression on his face, it means he's smiling.
That's predictive, because I've told the system.
Then I will take 14 pictures of rajeev, smiling in different ways.
That's supervised learning.
Then I will give it to the data and say, no, try giving me something else.
And that's when it starts giving me that unsupervised learning way.
It'll find another one which will probably be smiling.
And I'll say, yeah, this is fine.
Now, the tick machine learns that is a good algorithm, the weight identified, or will come back another one where I'm not smiling, convincing, I'll say, cross, That means, Amping lighting system, the system understands there's a difference here, it'll try and understand the difference.
And then it starts learning. And that's going goes in unsupervised learning.
That is when your machine starts doing a little more than just your average analytics. That's when it's not AI. It's machine language that kicks in.
And this is why I always say AI is very, very abused.
Artificial intelligence is only when it's really intelligent. It works on its own after a little bit of teaching or training.
So then, why not Excel? Yes, you can. So there are people where we've had, they've taken data from Excel. Put that into SQL Server Engine, database tooling.
And then from there, you start taking data, and then start, what we call the scoring, where, based on certain things, we started giving it scores.
It knows this is the right thing, that's the bad thing, So that taken across that I was talking about, and then it goes into something called VR. We use Power BI Analytical Tool, and you can, this is a pre configured tool, as provided by Microsoft, and you can go and configure that, start making and do things.
So there is already tool in the market. You can use that that comes through what we call the structured data.
You have to get it into a shape where it can start doing that.
And then you can do all kinds of fancy reports, where you start talking about things that didn't make any sense. Only this is a company in India where we worked with them and provide them a solution.
You can do all kinds of exploration, exploratory, data analysis, so we looked at credit card transactions, what features are there. So what was the number which that the transaction happen, which account number, date, and time, or you could have the transaction features a number of transactions from there in the last 24 hours, or average number of days. Average number of that.
Then, of course, you're going to have the cop dive and exploration date, and all the rest of it.
Then comes the next part where you take the data, and you say, I need to clean this thing, I need to look at what's missing in that data. Sometimes the data will come in, and, for whatever reason, it may not be the right data. Maybe ****.
That needs to be corrected somewhere. So that's treating the data. Before you treat data, first, you identify what's missing.
Then, you look at what it could be, because it might follow a trend. And based on that, you think, OK.
Although, Chase, 10, it might mean 10.04, because it's showing the trend would change that.
And then, as you start going through that, and you start analyzing, then you look at the, the overall picture and say, which one is not really making sense? This discrepancy.
Take that out before you start providing any kind of, making sense of that information.
And that's when you start preparing the data, so you categorize them into new medical sets.
In this case, because it was numerical data, You look at the data type, in terms of the format that is there.
You sample it, put it over a period of time, and then you start doing the analysis. So, these are different types of analysis. You can run the regression, uni variate, multivariate, and so on.
There are so many different things in the world of statistics, now, keeping me, as going super excited, And we'll go into Seattle Shi'a, not just getting bored.
However, what's important is the attributes that defines a lot of things.
You need to know information that is defining that particular dataset.
So in this case, it could be lhasa Chase the date and time, the amount of transaction, because we're talking about a point of sale. That's device where you goin top Ocado. You pay somebody at the diligence on.
So that's the kind of information you're looking for.
That's when you start having the actual idea of the information about the data, which is, as we call exploratory data analysis.
And based on that, we did things like, you could get a client client background and say, Hmm, this is the kind of information we have, and what was the problem. They're trying to look at. trade promotions, they were trying to leverage the promotions. So, they can offer better incentives now.
Same thing, if you are really keen, you can do this in a very, very data, privacy, respectful way, and one of them would be, you actually look at this customer experience, rather than understanding the customer's personal details.
That's very important in a business world, and you're thinking of AI, And because, then, you're actually looking at the core issues in a business.
You're not looking at why John or, oh, I don't know Js or josey are or someone else.
We're looking at certain things to buy.
There, you're looking at, this individual in his late forties was looking for a specific USB key, which is quite common in that space, because 19 year old never, probably has never even heard of these stupid USB keys, because they are used to clicking something on a cloud, getting a free cloud space, and uploading things that they don't care.
However, rajeev in his late forties is very concerned because the data's valuable to him.
That's the kind of information you're looking for.
That's when you really understand what promotion will drive Jeeves behavior, at that point, then you create the data audit for a group of .... People in similar brackets for you decide that's how I want it.
And when the system brings that information back, it doesn't bring any identifiers.
It's bringing outcomes that you're interested in, which means it doesn't store any of the past.
Which means after a file, it should be able to get rid of the original data.
And there's no reason to fear that because, if you've done your job properly and if you're looking at business value, then that's what you're interested in. And you're constantly then focusing on the business outcomes you're not interested in whether it was Rajiv about it. Oh, Josie, who bought it. And Rajiv Stingy, Fellow who just bought a very, very cheap.
USB stick and Josie comes in and he buys a very nice fancy one because he's really cool about these things.
No, I'm just knowing that.
Can interest, knowing what's a profile, it works, what, why doesn't work out as well.
So you don't collect that information, but the machine then start analyzing that.
your algorithm will do what you tell it to do, which means you have to decide at the onset what you wanted to do.
And of course, then, you have various kinds of analysis and all kinds of stuff that you could do inside eco, literally overboard.
You could do high, actually in this already. I'm not gonna go through that, but this is all the kind of stuff you could do.
And then comes the important thing.
This is very, very, very important, especially in our day and age.
So when you watch a movie and there are so many online platforms and you go and say that I liked this movie. I disliked the smoke. I liked this movie. I dislike this movie.
What that does is somebody sitting behind is an algorithm or solver, which records that from you.
This is where it can make it very personal to you.
Then, dare it's important that when it's making it personal to you, it should, ideally, be encrypted in such a way that no one can actually see the data.
That data, the encryption decryption happens only for the machine.
And that's quite important to understand, because then the outcome is going to be.
Rajiv tends to like the sci-fi movies, and he also likes a few other movies, which are comedies.
Specifically, he likes this one, particular comedian, or this particular director.
Then, all it does is starts looking for movies, which have a similar attributes.
So when you go to the likes of music and Spotify, the algorithms are quite superior, because they don't just look at what the category of that music is. They also understand how that music sits in the binary world. What trends do you see inside that music?
And based on what I like, it'll start understanding the pattern of that music.
And they'll start bringing the nearest one closest to it, so, especially in the world music.
one thing you may may not know is they used to things that you're familiar with.
So familiarity breeds a lot in your tastes, especially.
So they, they understand that. That's applying those things in the probability.
They'll start bringing more things which you tend to like, because they've understood that.
And the algorithms, and they'll try and put that back.
And constantly when they see something slightly different, they'll push that back to you, Because they've noticed that one, which is very nice that you like.
And constantly one could be that they've noticed that people at a certain age, and two like that, or people of a certain attribute, that that could be age, Or it could be gender. Or it could be something to do with your location. It could be to do with your nationality. It could be any of those. They tend to, like, they'll still push that. And if you didn't like that, they'll create something more closer to that, So they'll create different ones.
..., we have one minute left for the end of the session.
Oh, sorry, my apologies, So, that's essentially it.
Um, to put it, simply, we created tests.
What it does, is this: you could create situations where you have an alarm when notifies people It doesn't hold anything.
You could do that for children going to school, and you can have a geofence around it if they tend to go outside, notifies the school or the parent.
Or, you could have something called Cloud management where you could literally click on a screen as an organizer, and you can notify them, basically, it's still doing the same thing, machine language without without ever identifying people.
And within that fi trying to put augmented reality, meaning, you literally physically reach a point, and then you could be given a ticket in the virtual world. So, until then, you don't get anything. That's what happens in the real world.
You pick up your ticket, so you could pick up, pick that up, virtually, and then you could go in there, and then inside the stadium. You could literally have geo locations on the basis of which you could be notified, and so on.
Essentially, that's what we do as fisher. Feel free to go there if it ever, interest you.
Come back to me. Go back to the company's website, a R Y, L, T T dot com.
And I'll be very, very happy to talk to anybody if you ever think, you want to go get some idea about what you should do about AI, any project ..., very happy to do that. It's all on the website.
Rajiv, thank you so much for such a comprehensive presentation. You got into the details of AI, the promise, and the reality of AI applications today. So we really appreciate that comprehensive approach. We had a number of questions that came up. So while we have, because we don't have time for the Q&A here, we're going to ask. The folks to do is that they are going on LinkedIn, and they have posted. The question is on LinkedIn. So later on, I'm going to tag you on the LinkedIn post. And if you can provide some perspectives on some of the questions, that would be wonderful.
Thank you. As you can see, I'm quite passionate about this topic, and I completely lost track of time, but, Because you are quite a bit of depth on the, on the, on the thought processes, and the reality is of AI. A lot of the one of the comments that you made that really resonated with me is that AI is so misused in the other day. I had someone who was the multivariate regression analysis and they thought they were doing AI or like Kyle this is like techniques for like a century ago if it doesn't have any level of cognitive you know capability. But in any case, you really cover the fundamentals farewell. And then the applications. And there are some interesting, lots of interesting questions around privacy laws and how organizations adapting to this very fluid system.
Especially here in the United States where every state has their own privacy laws that there are enacting right now. It can become quite complex. So Rajiv, on behalf of our global community, I want to thank you very much for taking the time to share, for sharing your expertise with all of us today.
Thank you very much, much, please.
Ladies and gentlemen. That was Rajiv Chakrabarti with us sharing his expertise, knowledge about AI, and the implementation, the fundamentals, and the practical applications and the implementation of artificial intelligence in the way that we work, and the way that we operate in society. We're going to shift gears in with our next guest at the top of the hour. We're going to be taking a break now. And when we come back, we're gonna have Angela Perry. Angela is a leader is a vice-president of the Strategic Alignment for Transformation and Consumer Innovation at UPMC Health Plan. And Angela is going to be talking about aligning population health with digital transformation. It's a fascinating topic and coverage by Angela. I hope to see all of you back at the top of the hour for her presentation. Thank you.
Head of Portfolio Delivery of Test and Trace Digital - ATE, IBT and AI Reader,
September 07-09, 2021
October 12-14, 2021
Watch On-Demand Recording - Access all sessions from progressive thought leaders free of charge from our industry leading virtual conferences.Watch On-Demand Recordings For Free
Courtesy of DC Government's Ernest Chrappah, below is a transcript of his speaking session on 'Going Digital To Enhance The Customer Experience' to ...
Courtesy of PMO Global Alliance's Harris Apostolopoulos, below is a transcript of his speaking session on 'Digital Transformation: The Game Changer ...
Courtesy of 's Anu Senan, below is a transcript of his speaking session on '' to Build a Thriving Enterprise that took place at Enterprise ...