Value Driven Data Science: Boost your impact. Earn what you’re worth. Rewrite your career algorithm.

00:00:00 Dr Genevieve Hayes
Hello and welcome to value driven data science brought to you by Genevieve Hayes Consulting. I'm your host doctor, Genevieve Hayes, and today I'm joined by Doctor Mudassar Iqbal to discuss the data science behind large language models, or LLNS. Mutasi is the founder and CEO.
00:00:20 Dr Genevieve Hayes
Of team solve, a company dedicated to leveraging AI for digital transformation with a sustainable focus. He has extensive experience in industrial AI, including multiple pay.
00:00:34 Dr Genevieve Hayes
And was recognised as an MIT Young innovator. He has also played a key role in the growth of his previous startup Visenti, and its subsequent acquisition by Xylem Inc, Madasa welcome to the show.
00:00:49 Dr Mudasser Iqbal
Well, thank you, Jenny, you.
00:00:51 Dr Mudasser Iqbal
Happy to be here. Thank you.
00:00:52 Dr Genevieve Hayes
Great to have you.
00:00:54 Dr Genevieve Hayes
I love chef.
00:00:55 Dr Genevieve Hayes
CPT's not so much because of what it can do, but because virtually overnight it made AI and data science mainstream. Suddenly AI and data science went from being seen as a nice to have by businesses to being a must have, and now we've reached the point where demand for generative AI tools.
00:01:17 Dr Genevieve Hayes
Is so great that AI companies are having trouble sourcing the GPU's needed to train these technologies. However, while most of our listeners will have experience with generative AI tools as an.
00:01:31 Dr Genevieve Hayes
User few will have had experience in actually building and marketing their own tool, which is something that you have done through your company team self. So to begin with Medusa, can you tell us a bit about what team solve does particularly in the generative AI space?
00:01:51 Dr Mudasser Iqbal
Yeah, absolutely. So we come from, it's a, it's almost like a 15 years long journey now in the sectors where we play primarily water, energy and commercial building sectors and through the previous start of vicenti, there was a I was a co-founder, we started in Singapore.
00:02:10 Dr Mudasser Iqbal
And had the opportunity and the privilege of solving some of the pressing water laws and water security challenges for for water utilities UP of Singapore and a lot of utilities here in Australia. So we did that very well and even in that solution we employed a lot of AI to be able to.
00:02:30 Dr Mudasser Iqbal
Captured the data coming from the IoT devices deployed on a water network. Crunch through the data. Try to understand the patterns of how the pipes are failing. So it was a lot of machine learning. Data science involved there.
00:02:42 Dr Mudasser Iqbal
And then obviously the solution was delivered to our clients through amazing dashboards where they can see what's happening 24/7, alarms and all of.
00:02:51 Dr Mudasser Iqbal
That so that was an amazing journey. But one thing we saw first hand, a lot of operations when they happen in these sectors, a lot of that knowledge sits in the heads of the people.
00:03:03 Dr Mudasser Iqbal
There are things that sensors cannot see. There are things that any algorithm cannot predict. If these are things which are happening on the.
00:03:11 Dr Mudasser Iqbal
Around none of that was being captured, and eventually, if there is a massive pipe burst in the middle of the road, right, there's a traffic chaos. You deploy your crew to go and solve a problem. They struggle to find the right information to find the right valve that they need to shut off to shut off the the the, the, the water geezer.
00:03:31 Dr Mudasser Iqbal
Going up in.
00:03:32 Dr Mudasser Iqbal
Here, then we saw first hand a lot of mistakes happen. Field teams, if they are even even get worse when they're inexperienced, when they go out, they make mistakes. Not because they don't know just because they don't have the right information available on their fingertips. A lot of information in his is hidden in the piles and piles of documents in the heads of the experts who have been there for the last 20-30 years. They know.
00:03:55 Dr Mudasser Iqbal
How to do things?
00:03:56 Dr Mudasser Iqbal
But that's in their head. So when we started Team Sol early last.
00:04:01 Dr Mudasser Iqbal
We had this thing in our mind that this is one thing that needs to be solved, that this knowledge which is human knowledge, expressed as a human language. When the field teams they go, they see things they are not able to capture that in any meaningful way. What are they actually seeing?
00:04:20 Dr Mudasser Iqbal
They have apps, they have different dashboards that they they that they go through, but none of those tools allow them to accurately reflect what they actually saw in the field have.
00:04:31 Dr Mudasser Iqbal
So there is a gap in what we call the total operational knowledge. There is a gap in that knowledge that this tacit piece of information, which is either coming from experts or coming from the field observations that is lost. But that is the information that powers a lot of successful operations. That is the key.
00:04:50 Dr Mudasser Iqbal
To unlocking efficiencies in the operation.
00:04:53 Dr Mudasser Iqbal
So when we started team solve this, this was our vision to primarily pull together total operational knowledge of the organisation, whether it is sitting in the piles of documents, whether it is in the databases or whether it is in the heads of the experts, capture all that, learn from it.
00:05:13 Dr Mudasser Iqbal
Discovered the insights, discovered the mistakes, discover best practises from all that knowledge and make that available to the field team.
00:05:22 Dr Mudasser Iqbal
Not through complicated dashboards, but through a simple human conversation like you and I having a conversation because the last thing when you are in the middle of a road, in the traffic chaos, that the water geezer is going up in the air, the last thing you want is a fancy dashboard. You want to know exactly where is that valve? I.
00:05:39 Dr Mudasser Iqbal
Need to go and shut. So that was the idea.
00:05:42 Dr Mudasser Iqbal
Pull together all the information in one place, learn from it continuously. Evolve that knowledge because operations keep on happening on a daily basis. Keep on adding new knowledge into the system and the system we call Lily and then.
00:05:56 Dr Mudasser Iqbal
Bring Lily as an assistant or a Co pilot into the hands of every engineer, every operator, every manager in the organisation that whether they're sitting in a cafe, they're sitting, they're right in the middle of the operation. They can just ask the question and get back the response. And not just that they can tell back Lily, how are they solving the problems?
00:06:19 Dr Mudasser Iqbal
So Lilly keeps learning and evolving. So that's in a nutshell where we landed.
00:06:24 Dr Genevieve Hayes
So Lily could be thought of as being a private domain specific equivalent of ChatGPT. Is that what you're saying?
00:06:33 Dr Mudasser Iqbal
Yeah, yeah, that's one way to put it. I would say it is more like an insights platform, which is basically delivered through conversational interface, but it's a lot more to do with how it processes all the tacit information, which is unstructured information.
00:06:53 Dr Mudasser Iqbal
Or all the structured information which is available in all the all, all the databases and all that.
00:06:58 Dr Mudasser Iqbal
Highly curated for the domain that you're working on. So exactly to your point, it's highly curated, highly targeted to the use cases that you're trying to solve because when you look at chat, GPD, for instance, you can go and ask a question, right? You get better back a response? Where did it find the response? Whether the response?
00:07:18 Dr Mudasser Iqbal
Is actually grounded in the domain or grounded in the truth or not? You don't know. You have no idea.
00:07:24 Dr Mudasser Iqbal
Even to the extent we all know about the hallucinations, to the extent that you asked for it, hey, where is the source of this knowledge? And it gives you the sources just for you to discover. None of that exists. It just made up. Even the sources are made up. Right. So the industries where we operate the trust on AI is key to adoption.
00:07:44 Dr Mudasser Iqbal
If those operators in the field cannot trust what Lily is telling them, they're never gonna use it.
00:07:50 Dr Mudasser Iqbal
So Lily comes with a curated knowledge from the organisation. It is use case driven so it understands the domain concepts. It is powered by the domain of where it is operating. For example, for some of our customers where we are, we are helping them in such as replacing pipes. How? What is the optimal way to replace pipes?
00:08:10 Dr Mudasser Iqbal
Right. Should we replace this pipe or not? This is a very domain specific question that requires understanding of how these pipes operate. How do they fail? What are the different modalities of?
00:08:20 Dr Mudasser Iqbal
Failures. The lady is trained on those very specific use cases. It is powered by the knowledge that everybody knows what knowledge is going into Lily, so it's not a black box, it's a it's.
00:08:32 Dr Mudasser Iqbal
A glass.
00:08:32 Dr Mudasser Iqbal
Box and then.
00:08:34 Dr Mudasser Iqbal
Finally, and very critically, new knowledge coming in to Lily is.
00:08:41 Dr Mudasser Iqbal
Provided as base feature. So like in ChatGPT you go on and you try to tell it something new. You have to go through a very long process of other fine tuning or prompt engineering and all of that. Whereas with Lily that piece is part of the solution where you tell Lily on a daily basis. How are you solving the.
00:09:00 Dr Mudasser Iqbal
Problems really keeps learning from it, so the entire knowledge life cycle is included. So check Expedia. I would say chat kind of interfaces like one part small part of it.
00:09:09 Dr Mudasser Iqbal
Whereas Lilly itself is a total domain specific platform to help you solve the specific use cases that are really your pain points.
00:09:17 Dr Genevieve Hayes
I know chat EPT makes use of LLM's or large language models. So does Lily also make use of those?
00:09:25 Dr Mudasser Iqbal
Yeah. Yeah, absolutely. I mean, large language models, as you know, have really transformed. I mean, researchers have been working on large language models for a lot for a long time. It's not that just after ChatGPT, we all came to know. So researchers have been working on largely models for a while, but nobody actually knew the public actually did not know the the impact.
00:09:46 Dr Mudasser Iqbal
Since activity coming up in November last year, now we know the impact right activity picking up 100 million users in a span of two months. And then by now it is almost a billion users, a couple of billion users.
00:09:59 Dr Mudasser Iqbal
Compared that with TikTok that took about nine months to get to about 100 million users. So that's the impact that now everybody knows that LLM is something, right? So at the base, what LLM really allows to the systems such as ChatGPT and then obviously with Lilly is.
00:10:19 Dr Mudasser Iqbal
It's a breakthrough technology that allows you to process human text, human language. That's what it allows you to process. And if you go back a little bit, what I mentioned in the beginning, a lot of operational knowledge in these industries is expressed as human language.
00:10:34 Dr Mudasser Iqbal
And the problem is that human language expressed information. It sits in the piles of documents. Nobody ever looks at them. Nobody ever learns from them. It's just humanly impossible. So now with the language models, we have an opportunity to apply the language models to go and learn from all that knowledge which is humanly expressed.
00:10:54 Dr Mudasser Iqbal
So a very quick primer, if you may say, on how these LLM's really operate is right. So basically they they basically work.
00:11:03 Dr Mudasser Iqbal
Smith, they are trained to predict the next word and that process is a very elaborate process. It starts with representing human words into some word vectors like numerical representations.
00:11:16 Dr Mudasser Iqbal
And why do you need a medical representation? Is because, for instance, if you look at a cat, a word cat and a dog, the two birds don't look similar at a.
00:11:25 Dr Mudasser Iqbal
Well, but they they're pets, let's say, right? So belong to one class. So if you can numerically represent them in such a way that the numbers that represent cat and dog are closer to each other than numerically it is possible to see. Oh, these two numbers are closer to each other, so they must be similar. So that's where the word vectors come in, they represent.
00:11:45 Dr Mudasser Iqbal
All those concepts, all those words into numerical vectors. But then you take from there, and then you start to build a neural network. It puts all these words in an imaginary field where similar words are clustered around each other.
00:12:02 Dr Mudasser Iqbal
And then you give a job to a neural network, which we call a transformer network, which basically basically layers after layers after layers of neurons that try to make sense of looking at this field, of all the words that try to make sense, if you give it.
00:12:17 Dr Mudasser Iqbal
A sentence it.
00:12:18 Dr Mudasser Iqbal
Tries to make sense what should be the next word given.
00:12:22 Dr Mudasser Iqbal
It's similarity along around its cluster.
00:12:24 Dr Mudasser Iqbal
So this transform is kind of layers and layers of neural network, with each layer making a lot better sense of the sentence, like for example ChatGPT has about 96 of these layers, so it all came from that. Google's 2017 paper on attention is all you need, so that's what the Transformers do. They they try to put attention on the specific words.
00:12:44 Dr Mudasser Iqbal
And learn from that to predict the next One South. As you can imagine, what you really need for this is a lot of words, right? You need to fill that imaginary space with a lot of words. So.
00:12:56 Dr Mudasser Iqbal
That so that the Transformers can really learn how in different ways same word can be used in different ways. So if I say for example Sydney Harbour Bridge was completed in 1932, that's one sentence. And then I say we need to bridge a gap in our understanding of LLM. So the word bridge in both sentences.
00:13:17 Dr Mudasser Iqbal
It's the same word, but it's totally different.
00:13:19 Dr Mudasser Iqbal
Meanings. So that's where comes the training. Pardon. So we have words, we have Transformers, but then you need lots and lots of training data to fill up this imaginary space as much as.
00:13:30 Dr Mudasser Iqbal
You can for instance ChatGPT again the GPT model. It is. I think it's about half a billion gigabytes of textual data that was used to construct its imaginary space from it, where it's learning.
00:13:43 Dr Mudasser Iqbal
So that's where.
00:13:44 Dr Mudasser Iqbal
Just going back to the point that you made about Lilly using LLM so in these vertical.
00:13:51 Dr Mudasser Iqbal
Industries where we operate, water, energy, the challenge is you do not have a lot of good data available.
00:14:02 Dr Mudasser Iqbal
To really construct a very good high performance language model that can answer a lot of your questions straight away, you just don't have enough data to train a language model. So Lily uses language models capabilities because it is able to understand.
00:14:22 Dr Mudasser Iqbal
Human language. So if an operator in the field says, Hey Lily, I fix this pump by changing such and such component. So if somebody says this we bring this to a language.
00:14:33 Dr Mudasser Iqbal
And language model dissects this entire sentence and helps us understand what are the concepts, what are the different things this user is talking about. So it.
00:14:41 Dr Mudasser Iqbal
Pulls that out but.
00:14:42 Dr Mudasser Iqbal
Then that information is taken out, not kept in the language model that is taken out into what we call a knowledge graph, which basically constructs it's a representation of all the assets and operations and the learnings.
00:14:56 Dr Mudasser Iqbal
Which are happening in the organisation. So this dissected sentence is taken and then it is put in the right place in the graph wherever it fits.
00:15:05 Dr Mudasser Iqbal
And that starts to build the entire knowledge picture of the organisation.
00:15:09 Dr Genevieve Hayes
That's very interesting.
00:15:11 Dr Mudasser Iqbal
Yeah. So the advantage of this is now we are not limited by that. We need a lot of these sentences coming from these operators that we do. We don't need like 1000 statements from these operators to train our language model. Like when somebody asks that question next time an operator goes to the field and they ask the question, hey lady, this pump is not working. What should I do?
00:15:34 Dr Mudasser Iqbal
We are not limited that, oh, we only have one such example from the from before. The language model is limited, but our knowledge graph is not like.
00:15:43 Dr Mudasser Iqbal
So the system hits the knowledge graph. Machine learning happens on the graph, it learns. It picks up. So this is the best. This is the right way to do it. And then it is fed back into the language model to generate a human like answer. So it's the generative capabilities of the language model which are employed.
00:16:03 Dr Mudasser Iqbal
To have an effective conversation with the.
00:16:06 Dr Mudasser Iqbal
User, but then a lot of learnings, a lot of insights discovery, machine learning happens outside the language model. So it's so that's why it's slightly different from how the.
00:16:15 Dr Mudasser Iqbal
Chat GDP operates.
00:16:16 Dr Mudasser Iqbal
But it is it pulls together. It's a hybrid approach I would say.
00:16:20 Dr Genevieve Hayes
This is really interesting because earlier in the year I had an episode where I interviewed Alessandro ***** about knowledge graphs and one of the things we were talking about then was ChatGPT, and also how knowledge graphs can be used for similar purposes and what you're giving now is.
00:16:40 Dr Genevieve Hayes
A actual real life example of what he was talking about in his episode.
00:16:46 Dr Mudasser Iqbal
Right. Yeah, yeah, yeah. It's just the limitation.
00:16:49 Dr Mudasser Iqbal
Look at the very base. A language model is transformer neural network, which is very well trained to pick up patterns. It is very well trained to do predictions. That's the power of language model that we are using in Lily, not so much it's text generation.
00:17:09 Dr Mudasser Iqbal
Capabilities and not so much as capabilities to write an essay on all of that. So that in itself is a very powerful feature. But then because we know the limitations in the industries where we operate in the industrial sector, we just don't have a lot of good.
00:17:23 Dr Mudasser Iqbal
The data I mean if you look at how open AI and and all of the the companies which are building the foundational models, if you look at how much investment is going in, in building a language model from scratch, mean 1000, NVIDIA GPUs employed by open AI to train a language model, construct the GPT, it's just impossible for a water utility.
00:17:44 Dr Mudasser Iqbal
For example, to do the same right? It's just not never going to happen one. You just don't have enough data. And two, you just don't have enough money and resources to deploy. So that's where our approach is a hybrid approach that we can still this industry can still benefit from the power of a language model. So you take away.
00:18:03 Dr Mudasser Iqbal
All the pains of #1 all that tacit information expert knowledge is now buried. We don't know what to do about this. No. Now we can do a lot about it now.
00:18:13 Dr Mudasser Iqbal
We can really pick up all the insights from it because language model can understand all that as long as expressed in a language that the language model can work with. So one.
00:18:23 Dr Mudasser Iqbal
We can start to learn from all of that and two, the insights can now be delivered through a human conversation. You don't have to worry about those complicated dashboards. And I mean, I saw a picture from one of the water utilities.
00:18:36 Dr Mudasser Iqbal
They had this field operation going on and there was in the picture there was a Wan and the the vector of the van was open and there was a guy sitting in the on the backside of the van and had about, I think 7 or 8 screens in the van be beside him. And I can imagine what this poor guy is doing. He's basically trying to keep an eye on every little graph.
00:18:56 Dr Mudasser Iqbal
Every little GIS map.
00:19:00 Dr Mudasser Iqbal
Which is related to that operation which is going.
00:19:03 Dr Mudasser Iqbal
On and I was thinking.
00:19:05 Dr Mudasser Iqbal
How many more screens are we going to fit into this van? That's just not the way this is just so not productive. So LLM, have really come into revolutionise how the operations, how the information is captured and delivered in an operational setting.
00:19:21 Dr Genevieve Hayes
Going back to what you're saying before I agree with you that it takes massive quantities of data and resources to train an LM with the LM's that you're using for Lilly. I'm assuming you're not training them from scratch, is that correct?
00:19:39 Dr Genevieve Hayes
So is it some sort of transfer learning type approach?
00:19:42 Dr Mudasser Iqbal
Yeah. Yeah. So I mean, a few different MLMS are employed in our system because different language models are trained on very specific tasks, right? Some some are trained for good text generation, some are trained for good intent detection. What the user is really.
00:19:59 Dr Mudasser Iqbal
Trying to trying to.
00:20:00 Dr Mudasser Iqbal
Ask some are trained for for very good classification so and also there are cost implications as well. So some language models are available as a service, so you every time you use it you have to pay for it and some language models which are not large language models. It's like a small language models which are not trained on a very massive corpus.
00:20:21 Dr Mudasser Iqbal
But they have enough power to be able to be fine tuned for your specific task.
00:20:27 Dr Mudasser Iqbal
So what we in our architecture is we employ an ensemble of those language models. So for some cases where the data is not sensitive, some of these service language models which are available as service might be might be used of course with.
00:20:41 Dr Mudasser Iqbal
Agreements with our.
00:20:42 Dr Mudasser Iqbal
Customers, but for a lot of other.
00:20:44 Dr Mudasser Iqbal
Tasks we will.
00:20:46 Dr Mudasser Iqbal
Be fine tuning existing small language models.
00:20:49 Dr Mudasser Iqbal
Like you have burp, you have slant 5. Some of those smaller models are fine tuned 4 doing very specific tasks that really requires through the process of knowledge capture, knowledge, understanding, knowledge classification and learning insights.
00:21:05 Dr Genevieve Hayes
And given that they're smaller models, I'm assuming they take a lot less data to fine tune than something like GPT 4 would take.
00:21:14 Dr Mudasser Iqbal
Exactly, exactly. I mean, I mean you can, you can imagine it's a, it's a if a model is smaller if a small model is not pre trained on a lot of data, it means every new piece of information you put in will have a bigger impact as opposed to chat. GPD for example 300 billion trillion calculations were done to train its entire model.
00:21:36 Dr Mudasser Iqbal
175 billion weights all pre trained. So imagine if we have a small use case where one of our customers requires the power of a language model to help them maintain their treatment plants and they got sixty treatment plants. The data from these 660 treatment plants will be like peanuts.
00:21:55 Dr Mudasser Iqbal
As compared to what's already in the model, right?
00:21:59 Dr Mudasser Iqbal
So you will have to generate examples after examples, examples and variations after relations after variations to have some impact. And then obviously there are limitations on is very expensive to fine tune such massive models it takes.
00:22:14 Dr Mudasser Iqbal
A lot of.
00:22:14 Dr Mudasser Iqbal
Time. So if there are day-to-day operations happening, there's no way that that data can be fed into the language model model.
00:22:22 Dr Mudasser Iqbal
Fine tuned. And then tomorrow you will have is also available for the next operation. There is no such guarantee because it just takes massive amount of time and money and it just can't be real.
00:22:32 Dr Mudasser Iqbal
Time. So that's where these smaller models come very handy. They're not. They're scope their the parameter set is small. You can fine tune them and every fine tune you do it is impactful.
00:22:42 Dr Genevieve Hayes
Did you start?
00:22:43 Dr Genevieve Hayes
Building Lily before or after the release of ChatGPT.
00:22:47 Dr Mudasser Iqbal
Yeah, we actually. So we we were having quite a joke about this. So we started January 2022, that's when we started Team Saul. And then obviously we started brainstorming and some of the initial versions of clearly were already put in place.
00:23:02 Dr Mudasser Iqbal
It's with some earlier cognitive systems and libraries which are available pre jet GPT. Obviously we started as a startup, we had to go and find customers, so we started talking about that mode of interaction and what what was possible to some of the earlier potential customers. And they're just scratching their head and we know what you're talking about.
00:23:23 Dr Mudasser Iqbal
And then November comes chat. GPD comes out and suddenly everybody knows what Lily is all about. Suddenly everybody knows so.
00:23:33 Dr Mudasser Iqbal
Obviously some of the arc we have after ChatGPT, we have have some tremendous architectural enhancements because of the frame availability of a lot of new frameworks. So that has fast track, a lot of development for us. But development of Lily started and this concept of using cognitive capabilities to process human.
00:23:53 Dr Mudasser Iqbal
Knowledge and learn from it that started way back, so we were just kind of saying, OK, thanks technically you really helped to to spread the message and helping our customer base to really understand the what's possible.
00:24:06 Dr Genevieve Hayes
Well, even the things that aren't as sophisticated as something like Lily.
00:24:11 Dr Genevieve Hayes
Before Chats CPT, you had people saying we don't even want to do anything that's more sophisticated than, you know, basic Excel type calculations. You know we will never get to AI because that's beyond us. And then suddenly, you've got people who are in industries that you would not imagine.
00:24:31 Dr Genevieve Hayes
Using AI, going over to it, it's incredible.
00:24:34 Dr Mudasser Iqbal
Exactly, exactly. And we are so, like very positively surprised when we started talking to these customers and especially after tax TPT. It was like to be.
00:24:45 Dr Mudasser Iqbal
Nest the biggest trend we are seeing is at the management level at the top levels there is an appreciation of what is possible. There is an understanding at the top level of the benefits and the risks.
00:25:02 Dr Mudasser Iqbal
So a lot of.
00:25:04 Dr Mudasser Iqbal
Managers a lot of C-Suite, the leadership in our target customer base, they are talking about.
00:25:10 Dr Mudasser Iqbal
That yes, they need it, but with the proper fine tuning with the proper understanding of the risks involved, the actual user base where whether it's the planning teams who are going to use such systems to plan for, they want the system to process a lot of information they already have, and then they want to ask.
00:25:31 Dr Mudasser Iqbal
Or to ask questions. Discover all the insight.
00:25:34 Dr Mudasser Iqbal
Or it is the field teams who are actually going out on a day.
00:25:37 Dr Mudasser Iqbal
To day basis.
00:25:40 Dr Mudasser Iqbal
Normally are difficult to adopt any digital tool and you bring an AI to them and they will say, Oh my God, what? What is it? Right. So that group of users we are seeing is taking a little time is is taking their time to get to gel.
00:25:59 Dr Mudasser Iqbal
With this concept.
00:26:01 Dr Mudasser Iqbal
That instead of going through dashboards and through complicated menu systems, now they can just ask a copilot in their pocket. They just ask a question, they get back their response, and should they trust that response and do what? What Lily is telling them.
00:26:21 Dr Mudasser Iqbal
And that's where the explain ability and the trust factor comes in, which we have baked in our system from day one. Explain ability when Lee tells you something and if you're not sure, you can ask back, hey, Lily, why do you say so? And Lily goes back and then explains. So that helps with the potential that helps with the adoption.
00:26:40 Dr Mudasser Iqbal
But look, what we see is a lot of technology adoption happens top down in a lot of these organisations. So we have seen through COVID in one country, they put cameras in the in the vans in the fleet vans, so that when these people are going out doing their job.
00:27:00 Dr Mudasser Iqbal
The cameras will monitor. Are they wearing masks?
00:27:02 Dr Mudasser Iqbal
Or not. So that stopped down, right. And then in another place we see another company they implanted implanting chips in the in the thumbs of their employees to track and basically apparently saying that now you can just you don't have to swipe any card or no retina strands. You just come in and.
00:27:23 Dr Mudasser Iqbal
Your chip in your in your hand will be will be automatically scanned and you can come in and out. You don't need to log in and log out anymore, but that's that's tracking right? So.
00:27:33 Dr Mudasser Iqbal
So we see that a lot of technology adoption is top down, but over the last 15 ish years with the previous start up, we have really worked hands in hands with the field teams with the middle management who are actually the drivers who actually go and use these technologies.
00:27:53 Dr Mudasser Iqbal
So we have learned that this group of people, you really need to work with them, hold hands, get them to adopt, get them to understand and appreciate how this technology in the first instance is beneficial for themselves and then for the organisation. And that's where.
00:28:09 Dr Mudasser Iqbal
While the organisation helping with the top down, we are working bottom up to to help with adoption. So interesting trends and interesting behaviours we're seeing AI.
00:28:20 Dr Genevieve Hayes
Out of interest with that example you gave of the people with the chips put in their hands was that in Australia or in other country?
00:28:28 Dr Mudasser Iqbal
In the US.
00:28:30 Dr Genevieve Hayes
In the US, really.
00:28:31 Dr Mudasser Iqbal
Anyways, and the way it was done was like I just can't get my head around it. They organised some party and they they invited all the employees there and they say it's a party and what we're going to do is a volunteer basis come forward and we'll implant this chip in your so yeah.
00:28:51 Dr Mudasser Iqbal
Not sure about about the way the these top down approaches are.
00:28:56 Dr Genevieve Hayes
That's incredible because I remember, you know, when with the COVID vaccines, one of the conspiracy theories that was going around was that it was a secret conspiracy to in bed tracking chips in people. And I think Bill Gates was behind it or something. And and that's what I'm thinking of there. And yet.
00:29:12 Dr Mudasser Iqbal
Yeah. Yeah, yeah, yeah, yeah.
00:29:16 Dr Genevieve Hayes
It sounds like in your example, people were voluntarily signing it.
00:29:19 Dr Genevieve Hayes
Tracking chips.
00:29:21 Dr Mudasser Iqbal
Right, right. Yeah, so.
00:29:24 Dr Mudasser Iqbal
So why on one side the privacy that the systems such as ChatGPT and all of these systems, they're going to monitor us on a day to day basis, whatever we talk to them like when we when we listen to open AI and all the narratives that they did not go back in the lab and just kept working and came out.
00:29:45 Dr Mudasser Iqbal
GPD 7 in one go, they came with GPT 2, then GPT 3 and GPD 4, and the primary reason is they want a lot of public feedback. They want to see that adoption going through, so I think that public engagement is really, really.
00:30:00 Dr Mudasser Iqbal
Critical and that's going to take time, but it's just inevitable.
00:30:04 Dr Genevieve Hayes
Going back to the privacy issue, I'm guessing that's that's something that's of concern to a lot of your clients, and I'm sure pretty much everyone would have raised that when you're training Lilly, do you have a separate partitioned version of Lilly for each of your clients?
00:30:22 Dr Mudasser Iqbal
Yeah, yeah, definitely. So the data from one customer stays with that customer. They own the data and when we train our models, the knowledge graphs and and the and the language models, that all stays for that customer. It is not shared across.
00:30:40 Dr Mudasser Iqbal
So so that privacy is is top concern. Obviously when we work with the customers, they tell us they give us the data that they are comfortable with in bringing into lately. So they identify the use cases they.
00:30:51 Dr Mudasser Iqbal
The data and really strain on that so that that's taken care of. Obviously the, the, the other concerns on this, the data how good that data is is that of any use or not, right. So that's something that is that is tackled at the onset of when we train.
00:31:10 Dr Mudasser Iqbal
So whether such learnings when a language model in a certain country has learned some best practises that awarded utility A has such and such best practises, whether those learnings can be brought to some of the smaller water utilities who cannot.
00:31:30 Dr Mudasser Iqbal
Afford to deploy 10s of thousands of sensors or engage very expensive consultants? Can we really help those smaller councils and smaller water utility?
00:31:43 Dr Mudasser Iqbal
Ways to leapfrog in the way they operate their water networks by using the learnings from the larger utilities in the region. That is something really on our mind that we are trying to find a model whereby the learnings which are not customers specific learnings.
00:32:03 Dr Mudasser Iqbal
But the learnings how to fix a certain pump or when a pipe breaks, what are the best practises how to solve it? Not which specific pipe right when a certain when a flood happens? What are the best ways to in a in a response scenario right and we take those learnings?
00:32:23 Dr Mudasser Iqbal
And construct a language model which is fine tuned with these learnings and brought to the masses which is small water utilities or small SMEs who just cannot afford so affordability is is a key thing on our mind at team solve and that's something we're working with some of our larger customers.
00:32:43 Dr Mudasser Iqbal
To figure out how can we share share that knowledge because now.
00:32:48 Dr Mudasser Iqbal
With the language models, that is really, really possible. We can share all that knowledge. What are the modalities that something's still being figured out?
00:32:56 Dr Genevieve Hayes
OK. Yeah, that's good. Because, yeah, I was thinking about that and I was thinking there's so much knowledge that you're gathering there and if everything is purely partitioned, you're going to disadvantage everyone.
00:33:08 Dr Mudasser Iqbal
Yeah, I think it would be opportunity lost, I would say. And the good news is we do see amalgamations happening across the industry. So we do see, for example, in New Zealand in the water industry specifically, there is a reform where there there is an amalgamation happening with the 65 water utilities or so.
00:33:24 Dr Mudasser Iqbal
So what would that mean? That knowledge will be shared across. There will be a pathway from a regulatory perspective or the way the water sector is organised in that country, there will definitely be a pathway that knowledge can be shared across water utilities and we are seeing similar trends here in Australia.
00:33:44 Dr Mudasser Iqbal
Similar trends in other Southeast Asian countries and beyond. So we are very hopeful that some of these regulatory pressures will push the industry towards knowledge sharing and that's where technologies such as leading will be will be ready with with all that with all those learnings.
00:34:02 Dr Genevieve Hayes
I would imagine it would also help that most water utilities are government organisations and you tend to only have one water utility in a particular region, so it's not like with commercial enterprises where you've got multiple people in the same space directly competing.
00:34:19 Dr Mudasser Iqbal
You mean uh?
00:34:21 Dr Mudasser Iqbal
If you if you just rephrase the question.
00:34:23 Dr Genevieve Hayes
So for example.
00:34:24 Dr Genevieve Hayes
Well, in Australia in Melbourne the whole of Melbourne is divided up into areas and you've only got a single water utility servicing each particular area. So yeah, so north, South, East, West, etcetera. And so because you don't have say, 2 competing.
00:34:37 Dr Mudasser Iqbal
Yeah, that's right.
00:34:45 Dr Genevieve Hayes
Utilities in a particular area, there'd be less fear of 1 organisation sharing information with another because it's not like they can poach customers from.
00:34:56 Dr Genevieve Hayes
Each other.
00:34:57 Dr Mudasser Iqbal
Yeah. Yeah, yeah, yeah, absolutely. Absolutely. Yeah. And and that really enables that kind of demarcation really enables them to be able to share higher level insights and with each other because one thing we have observed in these industrial sectors is digital adoption is everybody wants to learn from each other. And that's why we see forums.
00:35:17 Dr Mudasser Iqbal
The swan and some of the other other conferences happening across the world where these utilities want to come together.
00:35:24 Dr Mudasser Iqbal
These organisations they do want to come together, they do want to learn from each other's findings. OK, you have deployed, for example, set and such 100,000 of smart metres. What were your learnings? But you're fine, right? So there is an already demand for knowledge sharing across the industry. It's just that the.
00:35:44 Dr Mudasser Iqbal
Enabling regulatory pressure and the structural organisation of how these organisations are structured within a country or within a region, it's just moving in a very positive direction, where sharing knowledge across the board will will become far more easier as opposed to, for example, which was few years ago.
00:36:02 Dr Genevieve Hayes
While back you were talking about how many of your customers already have dashboards in place and you have to convince the people working in the field that they should shift away from those dashboards to tools like Lily. How do you convince them?
00:36:19 Dr Mudasser Iqbal
So we are we are fine. We are having some very interesting experiences. So one of our customers, one of the field staff, he went, he went away on a vacation and during that time we have been training Lily the person had seen one or two versions of Lily and then he went away. So he was still still training.
00:36:38 Dr Mudasser Iqbal
And the person went away, I think 3 or 4 weeks and came back and obviously.
00:36:41 Dr Mudasser Iqbal
In that time.
00:36:42 Dr Mudasser Iqbal
We had added a lot more features into Lily and interaction and then the person was sent out to to do an operation in the field to fix to fix his side and the person at the side and had.
00:36:53 Dr Mudasser Iqbal
Lily on the.
00:36:53 Dr Mudasser Iqbal
Mobile phone and and actually thought for a second. What do I do?
00:36:57 Dr Mudasser Iqbal
Well, how do we talk to Lily? So it was 0 training the person. We never told this person how to use Lily. So the person just just went on like a normal conversation. Hey, Lilly, I'm here. What do I need to know? There you go. Conversation started and then the person was helped along.
00:37:12 Dr Mudasser Iqbal
Way. So these are some very positive feedbacks coming from those field teams that with zero training that they are able to come up to speed and they are able to do their job. They don't have to really remember. Now wait to click where was that drop down menu. But that's obviously one class there's another class.
00:37:33 Dr Mudasser Iqbal
Which are like slightly older, older age people who are not already not tech savvy, so they already hate it. All the complicated dashboards, all the graphs and.
00:37:43 Dr Mudasser Iqbal
And and chatting away in our lives has just become a norm, right day in and day out, we chat with our parents and kids and grand grandchildren and everybody. So even these older technicians or older people, when they have really as a chat on their on their phone.
00:38:03 Dr Mudasser Iqbal
It's no difference than them having a WhatsApp conversation with some other person, right? So even for them they will maybe hesitate for a little bit. Should I say something would is, is that gonna offend Lily or what? What should I say? And then they just they just say it and just it just go picks up from there.
00:38:20 Dr Mudasser Iqbal
So we're not seeing a lot of pushback or a lot of difficulty from those people who are already frustrated with all those complicated dashboards. So we are seeing a lot better traction, a lot better acceptance at that level. The challenge that we are seeing is with.
00:38:40 Dr Mudasser Iqbal
Kind of like the back office people who are used to looking at fancy charts and fancy graphs, and they're used to be looking at big screens and they're used to that way of working and for them to change that now look.
00:38:54 Dr Mudasser Iqbal
You used to look at a graph of. Let's say you you want to see OK you want to see the performance. Let's say the the the water loss is in a certain area. Yeah, right. So you you were used looking at a map and that map has all the different regions colour coded and the one which is red is the one which is which has the most losses. So you would use for this way.
00:39:13 Dr Mudasser Iqbal
Of looking at the full map.
00:39:14 Dr Mudasser Iqbal
And spotting the red one. And you say, ah, this was the this is the one now change the.
00:39:19 Dr Mudasser Iqbal
Way now you don't have to look at the graph. Look at the map. You say. Hey, Lily, tell me which which region is most problematic from what losses perspective and boom really picks up that one region and now they're saying is it correct? How does it compare with the rest? It's just a different modality, but this thing we also understand the need.
00:39:39 Dr Mudasser Iqbal
Because and I think that's another aspect of working with kind of ChatGPT and and all the other things. And I have experienced this first hand that when you ask a question when you're looking for news articles, right, what's happening in the world.
00:39:55 Dr Mudasser Iqbal
And then let's say there is, there is an event happening and let's say it is U.S. Open tennis championship going on and you want to see what's happening. You, us, Microsoft Bing, which is now connected to ChatGPT and it is updated. Latest information ask a question. It gives you a summary. Now how that summary is constructed may have a bias in it.
00:40:17 Dr Mudasser Iqbal
Because it is trained by people, AI is not going on its own journey, right? Some people are crafting the designing the neural networks behind it, so that's somebody comes.
00:40:27 Dr Mudasser Iqbal
Out, whereas in the olden days of searching on Google, I'll go into Google and I'll type the US Open 2023, everything comes out and I can just glance through look.
00:40:37 Dr Mudasser Iqbal
At all the.
00:40:37 Dr Mudasser Iqbal
Scores. I will be much. I will have a much richer experience of the information I will get as opposed to asking Bing a question and Bing just giving me just one little snippet.
00:40:48 Dr Mudasser Iqbal
So that's the thing that we are facing on the back office people that when they get a simple answer from systems such as Lily, they kind of feel that they're missing the full context that what is everything else happening in the network view. I want to be, we want to be able to see.
00:41:04 Dr Mudasser Iqbal
That so the way that thing is being cracked is.
00:41:08 Dr Mudasser Iqbal
Lily integrating with existing legacy systems, tightly integrating. So for example, if you use power BI for your dashboards where you want to see the fancy dashboards, the difference now is your power BI will connect to Lilly instead of you connecting to Lilly. So your power BI connects to Lilly, Lilly gives the right insights to power BI.
00:41:29 Dr Mudasser Iqbal
And Power BI projects that on your traditional dashboards.
00:41:33 Dr Mudasser Iqbal
So it just changes that the AI is now in the within the ecosystem. It is in the mix, it is powering your existing data analytics tools, your existing dashboarding tools. So you can still look at your dashboards, but those now those dashboards are powered by systems such as such as lilies, AI and the language models.
00:41:53 Dr Genevieve Hayes
And that makes it easier for people to accept and adopt A tool such as Lily.
00:41:59 Dr Mudasser Iqbal
Yeah, that that eases the the transition because I can imagine if a technician or if a if a command centre person who had who used to be looking at those complicated dashboards and still loves those dashboards goes out for a lunch and is sitting in a on a lunch and gets a message from his boss. Hey John.
00:42:20 Dr Mudasser Iqbal
Hey Mudassar, which region has has most of the losses? I need this answer in.
00:42:25 Dr Mudasser Iqbal
The next two minutes.
00:42:27 Dr Mudasser Iqbal
This person now will be just able to pick up the phone. Hey, Lily, which region has the losses? Lily gives the answer the person does the job. Boss is happy. I'm happy. I can go back and look at my dashboards. But I think this will ease. That that transition slowly.
00:42:41 Dr Genevieve Hayes
How did you come up with the name Lily? Is it an acronym or is it?
00:42:45 Dr Genevieve Hayes
Just you like.
00:42:47 Dr Mudasser Iqbal
Yeah, there is a. There is a story behind it. Yeah, I think a lot of these artificial these systems, the AI systems, they they typically you have Siri, you have Alexa. So we we started this in Singapore. So in that community is it's it's quite a common name. Yeah. So just just speak of a name just this sounds different sounds good sounds.
00:43:07 Dr Mudasser Iqbal
That sounds nice. Yeah, but yeah, we did. We did brainstorm on quite a lot of lot of ideas, a lot of names.
00:43:14 Dr Mudasser Iqbal
But we are talking to some of the other customers and some customers want to want to have their own names. So like in Malaysia for example, we were having a joke with one of the customers. They want to name it like city cities like a common name there. Yeah. So we're not really tied to.
00:43:28 Dr Mudasser Iqbal
The name could change. Yeah, where it goes.
00:43:31 Dr Genevieve Hayes
Do you have to say, hey, Lily, do this like you would with something like Alexa or Siri? Or can you just say?
00:43:40 Dr Mudasser Iqbal
Yeah. So Lilly is added.
00:43:41 Dr Mudasser Iqbal
If you're using.
00:43:42 Dr Mudasser Iqbal
Lilly on let's say your messaging platform like WhatsApp or Microsoft Teams, that's where you can access Lily. So Lilly is added as a contact just like you can add me as a contact. So anytime you want to talk to me, you just open my, my, my contact where you have all the conversations and you just just start talking to me.
00:44:00 Dr Mudasser Iqbal
There are some future some features that we are working on where Lilly can sit in a group conversation. Imagine you are in a command centre and Lily is sitting in the command centre as your virtual operator.
00:44:11
OK.
00:44:14 Dr Mudasser Iqbal
Winter and you are having there is an emergency happening. You're doing an emergency response. There are six or seven people in the command centre, Lilly being one of those. And now if you want to talk to Lilly and you want some answer, you don't want Lily to be listening, to be answering everybody's everybody because I'm talking to you. You're talking to me. Yeah. So in that case, you will say, hey, Lily, what do you think about this? Yeah, those are some of the future things we are working on.
00:44:36 Dr Genevieve Hayes
And yeah, I I was just thinking back to how some people wanted to rename Lily. I could imagine if you had a company where a senior executive had the name Lily, you could have some problems there.
00:44:47 Dr Mudasser Iqbal
Yeah, yeah, yeah. I think that's where the the transformer models come really handy because the, the, the base of the transform transform model is the attention which helps it to respond based on the context. So so that context.
00:45:01 Dr Mudasser Iqbal
Hopefully will help.
00:45:03 Dr Mudasser Iqbal
In Lily not jumping in when you're talking about another.
00:45:08 Dr Genevieve Hayes
And where do you see all this? Generative AI and large language model and language technology heading in the near future?
00:45:16 Dr Mudasser Iqbal
I think the key milestone that we see the organisations will will take in the in the next, I would say few years.
00:45:25 Dr Mudasser Iqbal
Is a lot in the in the in the direction. They will gain a lot of trust in AI powered specifically by the language models to the extent where I see that where currently they might be using.
00:45:37 Dr Mudasser Iqbal
It in a way.
00:45:38 Dr Mudasser Iqbal
It is more of a recommendation system, so you ask a question. You get back the answer right?
00:45:43 Dr Mudasser Iqbal
Or you just tell it what you're doing.
00:45:44 Dr Mudasser Iqbal
And it it captures it.
00:45:46 Dr Mudasser Iqbal
But then I see this moving more towards organisations giving it a little bit more control of making taking decisions. So one simple example is let's say you're in a command centre and there are tonnes and floods of alarms coming on your screen and you're not sure what are all these alarms. So now in today's world you will say, hey Lily, I have this alarm.
00:46:07 Dr Mudasser Iqbal
What do you think? What do I do? Should I dispatch the crew or not? So really we'll go and assess and tells you. No, you don't have to dispatch the crew because of XYZ features. Ohh really will say yes. Dispatch the crew immediately because it's a it's a massive issue. So that's the today's.
00:46:20 Dr Mudasser Iqbal
World in the next few years I see this where you where these customers allowing Lilly and allowing these AI systems to actually take the decision and dispatch the crew because they would have gained a lot more confidence and the and trust in the capabilities that the recommendations coming from such systems.
00:46:40 Dr Mudasser Iqbal
Or or on par with what a human expert would have already been been taking.
00:46:47 Dr Mudasser Iqbal
I think AI in general and specifically generative AI is already going. I think leaps and bounds in some of the other like healthcare areas. I was just yesterday. I was looking at like Google Deep Mind just released their research note where now this system is able to look at the sequences of the letters in the human genome to detect which letters.
00:47:07 Dr Mudasser Iqbal
Are out of sequence which will lead to the proteins being forming in the non right.
00:47:12 Dr Mudasser Iqbal
Wait, so again it is going to the the sequence of characters in the human human genome and spotting the disease sections. This is going to just continue. And it's, I think with the regulations coming in and with with a lot of trust building up it's it's it's just will transform the society.
00:47:32 Dr Mudasser Iqbal
Leaps and bounds.
00:47:33 Dr Genevieve Hayes
So it sounds like you're very positive about this. Do you have any concerns about where all this is heading?
00:47:39 Dr Mudasser Iqbal
Yeah, I I'm definitely not in the camp of the Skynet kind of things where, yeah, where you have somebody sitting out there and they're trying to trying to manipulate, manipulate everything I think.
00:47:49 Dr Mudasser Iqbal
A lot of us tend to think that this AI is running its own course, and it has gone uncontrolled, and that's not true.
00:47:57 Dr Mudasser Iqbal
There are only a bunch of people in the world who are dumping billions of dollars on the development of AI, so AI development is very guided and controlled by humans.
00:48:09 Dr Mudasser Iqbal
So I think the the only risk I see is who are those humans? What are their intentions? So as long as the intentions are right of the group of people who are making investments, I think you're OK. And the other flip side of this, I would say the biases that the data which is being.
00:48:29 Dr Mudasser Iqbal
Compared to the systems making those data sets more neutral will remove the biases. I was just looking at one example somebody asked a question to AI. I think it was GPT as well. You take a doctor, subtract man plus woman, right?
00:48:45 Dr Mudasser Iqbal
What do you care?
00:48:46 Dr Mudasser Iqbal
If the answer is nurse, that's a bias, right? Because why right? So, so those data sets making those datasets neutral, that will help. Obviously a lot of concerns around the job losses like chat, GPD passing 90% on the, on the on the bar exam.
00:49:07 Dr Mudasser Iqbal
And doing perfect on the SAT math tests and all of that. That's OK. I think it's going to definitely displace jobs if it displaces in a matter of single digit years. That's a concern. But if it displaces on the matter of double digit years, I think it's OK that has happened through the history of the mankind. Anyway, new technologies come in.
00:49:25 Dr Mudasser Iqbal
New jobs come in, so yeah, so I'm quite positive in the in the direction as long as we control it well.
00:49:32 Dr Genevieve Hayes
I would agree with what you say about the job displacement because.
00:49:36 Dr Genevieve Hayes
As long as it doesn't happen too fast, if the new jobs being created increase at a rate faster than the job losses, then there's no problem.
00:49:44 Dr Genevieve Hayes
Not at all.
00:49:45 Dr Mudasser Iqbal
Right, right. Yeah, exactly. Yeah.
00:49:47 Dr Genevieve Hayes
And and you would have seen that with the invention of the automobile, it would have displaced a lot of jobs with regard to horses, but it created a lot of new jobs, like cab drivers, for example.
00:49:59 Dr Mudasser Iqbal
Right. Yeah. A few days ago, nobody would know. The software engineer is a job and nobody would even cannot, can't, cannot even think about it. And I think a few, probably a decade couple of decades later some people say, oh, software engineer used to be a job. So these displacements are happening but change is that has kept us all going.
00:50:19 Dr Mudasser Iqbal
And yeah, I'm not worry.
00:50:21 Dr Mudasser Iqbal
About that.
00:50:22 Dr Genevieve Hayes
So what final advice would you give to data scientists looking to create business value from data?
00:50:29 Dr Mudasser Iqbal
I think any business value it stems from the realisation of either you understand the pinpoint that what you're solving or there is a possibility that you have discovered and that that's what you are able to to spell out very well. So as long as the data science community.
00:50:49 Dr Mudasser Iqbal
Grounds that is surged in the pin points has a has a business value to to someone to, to a utility or to an organisation as long as the pin points or are identify.
00:51:00 Dr Mudasser Iqbal
Right. Or you have such an amazing thing that you're building and you know the new possibilities. I think iPhone is a classic example of these new possibilities. Before iPhone, we never thought that's a pain point that we need an iPhone. Nobody had that pain point. But now can you live without an iPhone kind of device? I'm not sure about that. Right, so.
00:51:20 Dr Mudasser Iqbal
Either you have discovered a pain point, or you have discovered a possibility. In both cases, if you are.
00:51:25 Dr Mudasser Iqbal
To do one or two, the business value will be created, but then I also want to highlight that thinking outside the box will also come very handy. I mean, I come from a PhD background and we have a certain way of doing things. I think sometimes thinking outside the box and leveraging on different models which are available as a service.
00:51:47 Dr Mudasser Iqbal
And fast tracking your research sometimes goes leaps and bounds in generating value. So for example what I was what I was telling before that not thinking that large language models are primarily you just use it for writing an email or writing text.
00:52:04 Dr Mudasser Iqbal
Think about it. These are basically transformer models available as a service. What can you do with these models? Don't don't think about that. These are chat. GP kind of things. You have a transformer model neural network available as a service to you. Can you utilise this for pattern matching? Can you utilise this for identifying similarities classifications?
00:52:25 Dr Mudasser Iqbal
Fast tracking your research so thinking outside the box on the generative AI land.
00:52:29 Dr Mudasser Iqbal
OK. And really grounding your research into the pain points and possibilities, I think that will generate value faster.
00:52:36 Dr Genevieve Hayes
For listeners who want to learn more about you or get in contact, what can they do?
00:52:41 Dr Mudasser Iqbal
Yeah. So my LinkedIn profile, I think you can just go on LinkedIn search for Mudassar Iqbal, my name and I should come up and my email address mudassar@teamsolve.com very happy to connect with anybody who wants to and would love to learn. There is a lot of possibilities here. There is a lot that we're still learning. These are really early early days for AI.
00:53:01 Dr Mudasser Iqbal
Cash generate.
00:53:02 Dr Mudasser Iqbal
AI so would love to engage in more conversations with anybody who's interested. And yeah, let's let's do this together.
00:53:10 Dr Genevieve Hayes
OK. So thank you for joining me today.
00:53:12 Dr Mudasser Iqbal
Excellent note. Thank you, Genevieve. This was a fun conversation. Thank you so much.
00:53:17 Dr Mudasser Iqbal
For your time.
00:53:18 Dr Genevieve Hayes
And for those in the audience, thank you for listening. I'm doctor Genevieve Hayes, and this has been value driven data science brought to you by Genevieve Hayes Consulting.