Binny Gill: 0:03

Businesses want to move fast. They're being slowed down because the logic is ossified in some old style automation. How do you unlock it? Unshackle it. Give it to the business side to say, Hey, this is in English. You can change it. And the moment you change it, this is your pseudo code, right? You don't need to learn programming, just change it. Another machine will work with you to execute it. So that's the, you know, the inside the journey. And that's also the future. I, I believe we will come out of the dark edges of computer science soon.

Tom Raftery: 0:37

Good morning, good afternoon, or good evening, wherever you are in the world. This is the Digital Supply Chain Podcast, the number one podcast focusing on the digitization of supply chain, and I'm your host, Tom Raftery. Hi, everyone, and welcome to episode 352 of the Digital Supply Chain Podcast. My name is Tom Raftery, and I'm excited to be here with you today, sharing the latest insights and trends in supply chain. Before we kick off today's show, I want to take a quick moment to express my gratitude to all of this podcast's amazing supporters. Your support has been instrumental in keeping it going, and I am really grateful for each and every one of you. If you're not already a supporter, I'd like to encourage you to consider joining our community of like-minded individuals who are passionate about supply chain. Supporting the podcast is easy and affordable. With options starting as low as just three euros or dollars a month, that's less than the cost of a cup of coffee, and your support will make a huge difference in keeping this show going strong. To become a supporter, you simply click on the support link in the show notes of this or any episode or visit tiny url.com/dsc pod. Without further ado, I'd like to introduce my special guest today, Binnie. Binnie, welcome to the podcast. Would you like to introduce yourself?

Binny Gill: 1:55

Hey Tom, thank you for inviting me to your podcast. Hi guys. I'm Binny Gill founder and CEO for Kognitos I've been a student of computer science all my life. Like to tinker with new technology and these days I'm the, Chief Tinkerer at Kognitos with the generative AI technology and using it for automation.

Tom Raftery: 2:13

Cool. And what does kognitos do, Binny

Binny Gill: 2:17

We are creating a new way of automating business processes. You know, back in the day you had people doing business process outsourcing 'cause you wanted your company to be more efficient. But what you did was take some cheap labor in some third world country and they are doing the same process that you were doing in country. Then in 2010 or so, RPA become, became something fashionable and said, well, you know what, instead of humans, let's then let the bots do it. So in the bots brought in some more automation and now they have plateaued in terms of what they can do. And finally, AI is now real, and AI can bring in the next quantum leap of efficiency for businesses. Where bots fail, AI actually can start and doing, mimicking what humans do. So that's what we are doing and we are seeing that the next generation of business efficiency will come from a combination of what bots used to do and what AI can do.

Tom Raftery: 3:17

Okay. And can, I mean, we, we see with the likes of ChatGPT, that it can keep its state. In other words, you can ask it a question and then ask a follow-up question and ask a follow-up question, and it remembers as you go along. Unlike the current iteration of AI that we're mostly familiar with, the likes of Siri, for example. You ask it a question, you ask it a follow up question, and it goes, huh. And, you know, but, but ChatGPT is, is, is different in that, in that it can keep the state and it can remember and it can do follow ups. Is that the kind of thing that you're doing for organizations as well, that you know you can do? Or what, what is it that you're doing with your, your AI for, for organizations?

Binny Gill: 4:06

Yes. And. Yes. I mean, that is what AI is where AI is going. Although organizations require a whole lot more than what ChatGPT is able to do today. But, but what you just said it's exactly how Kognitos started. I mean, I'll tell you the founding story, if you have. Essentially my son during the pandemic, and he, he was 12 oh years old, three, three years ago. And he was um, getting bored and like, Hey, what do I do? And like I said, okay, you know, learn programming. And he said, okay, I don't know. It's like, I'll try and, and I encourage, you know, create a video game. He said, oh, you can do that. He said, yeah, it'll be pretty easy. Just go and do it. So two days later he comes back and he shows me tic tac toe.

Tom Raftery: 4:55

Okay.

Binny Gill: 4:55

In Python, obviously. I mean, so I was very proud. I said, great. I slept Tom. I woke up next morning. I realized I had made the same Tic-Tac-Toe game 30 years ago in a language called Basic,

Tom Raftery: 5:09

Yeah.

Binny Gill: 5:09

and I had this Casio handheld thing. There was no internet, no YouTube, no Google or anything. But I had this manual that spoke about the language, the basic language, and what you can do, it was black and white, l c d, so you could draw lines and you can write an X and O. So, but it took me one and a half days, I remember, without any help of the internet,

Tom Raftery: 5:31

Mm-hmm.

Binny Gill: 5:33

and I went back to my son and say, okay, how many hours did you really spend? And he said, no, I, I, I worked hard for two days. And like, why, why didn't it not take you like half an hour?

Tom Raftery: 5:43

Yeah.

Binny Gill: 5:44

And he says, what? I thought you were proud of me, like I said, no, no, no. I'm not pushing. Okay, let me learn. Relearn the pains of automation, pains of programming. So I said, let's come up with another problem. And like we quickly came up with, okay, do you know how, what a prime number is? He said, yeah, okay. Do you, okay, so let's write a, you know, code, not in Python, but actually, yeah, pick Python and then figure out whether the number is prime or not. He's like, okay. I said, write He was like, totally blank. Total, like, thanks. Can I Google? I said, no, think there was nothing. And I said, okay. There is something called pseudocode. And he said, what is that? He's like, imagine your explaining to maybe your little sister, like what is a prime number, but you want to be very precise about and, and say it so that it can be used as a in, in a textbook form. Like this is what you need to do. He said, okay. He started writing. It's so pseudocode, like if it's one, it's not prime, if it is divisible by any, so we wrote it in five minutes and then he say, okay, now you know, now you have a plan. Let's start converting it to, you know, programming. So first line was easy, second line because it had a loop in it, so you have to do it. And then that thing is very, I tried two hours to explain the concept of loop and variable and all of that, and. I mean, you can teach that in two hours to anybody who has not done programming before. And it didn't go well. He's like Dad, maybe I don't want to do it. I said, oh, no. And then I said, you know what? Tell me what would work for you. And that's where he says, dad, why can't this work? And he was pointing to the pseudocode and I laughed. I said, no, that thing is for you. It's your own notes. It's not for the machine. He said no, that just like Alexa. And that's what, you know when you said, oh, Alexa can do it. Yeah. The next generation believes that machines can do it

Tom Raftery: 7:47

mm-hmm.

Binny Gill: 7:48

and they demand that machines can do it. And this is three years ago before ChatGPT was anything. I mean, nobody knew. And I was like, maybe there is a, there is some truth to that. If Alexa can do one line, why does it forget the context in the next line and the next line? Is that that hard? So I mean, and I love to do these things, you know, pick a hard problem in computer science and go to the roots of it and say, what is it that we are missing? Why? Like, so then I looked at the history of computer science all the way from 1930s when Alan Turing was doing his work. Alonzo Church, a contemporary of Alan Turing, came up with a language for programming called Lambda Calculus. So this predated computers. That is the foundation of all programming languages till date, Lambda calculus. And he taught the world one thing. It's like it's a very simple language, but hard to program it. Simple language. Basically, if there is a method F, you call it F, and then you give it a parameter, X out will come y. Now, that's how programs are written. Like, okay, there's something I need to invoke. I have to give it some parameters, and then if I am correct in my parameters and how I invoked it, I'll get a result that I want. If I'm incorrect in my parameters, the parameters are missing, it'll crash and burn. Whereas how humans interact with other humans is a dialogue based system

Tom Raftery: 9:21

mm-hmm.

Binny Gill: 9:21

where I can even forget to give x. I say, Hey, you know, get me coffee. Okay. So, okay, now I forgot to say whether I need sugar or not, whether I need milk or not, right? I forgot. But the system doesn't crash. The guy goes to the kitchen and says, Hey, decaf, or regular, if there was only one brand of coffee in the kitchen, that question would not come to me and it wouldn't be noticed at all. It's very different, very different paradigm. So that's the paradigm that now we are going to, with the benefit of AI. AI is needed to be able to form the right questions to an operator, the layman who's not a developer, like, Hey, coffee or regular, right? I mean, so AI can generate that question, get the answer, and now bring precision to the process and go on from there. So that's the next generation of in general software programming or automation. You apply it to all sorts of things. In fact, the industries where this is needed most is paper trail, heavy industries, manufacturing, logistics, c p G. This is where insurance, right? So much so much data is there. And then the processes that are tied on top of these paper trails are massive. And they're not cut and dry. That's why there are thousands and tens of thousands of people who are doing things manually today, and all the businesses out there pushing paper forward. And why? Because it's, it's not electronic data. It's, you know, it's still enough, enough in a way where a human is better at it. Now, AI is getting better. ChatGPT can understand English in a email. You have O C R technology. They can understand from a PDF extract all of that. Now start doing business logic. And the world that we live in is a dramatically rapidly changing world. Now, the business logic needs to evolve fast, faster. You know, like businesses want to move fast. They're being slowed down because the logic is ossified in some old style automation. How do you unlock it? Unshackle it. Give it to the business side to say, Hey, this is in English. You can change it. And the moment you change it, this is your pseudo code, right? You don't need to learn programming, just change it. Another machine will work with you to execute it. So that's the, you know, the inside the journey. And that's also the future. I, I believe we will come out of the dark edges of computer science soon.

Tom Raftery: 11:57

So what you're essentially talking about is a, a no-code, low-code way of using AI in your enterprise.

Binny Gill: 13:19

I will challenge that. I, I, yes. First of all, I agree, but I'll challenge

Tom Raftery: 13:24

Okay.

Binny Gill: 13:25

I'm saying natural code. So my grandma's recipe for apple pie to me is code. It's five lines of five lines of instruction happens to be natural code, and it is the most reusable piece of software ever written in the world. Grandma's recipe used to work 50 years ago in old hardware, old kitchens, and even in the latest gadgets of today. Why? It it is software like, it's like instructions. What she was smart at. She's wrote all the salient features, like what is the important decisions for making apple pie, but left the situational choices up to you. Are you going to use a wolf oven? Are you going to use that? I mean, grandma doesn't care, right? But it is not no code. It is not low code. It is natural code. So that's what I'm saying. The time has come to not hide code behind graphics. Grandma didn't do drag and drop for me. Right. The the time has come to speak about business logic that you care about most upfront. That is code that has always been code. If you talk to a business guy, say, what is the logic that you run your business on? They'll write a Google Doc and say, Hey, this is the logic, and yes, by the way, some people are implementing it. I don't care about that. This is the, these are the 10 rules. That is natural code. So the business logic so far has been embedded in computer logic or hiding behind some drag and drop widgets. We need to change that. All that, you know, clutter has to go away though. Those are situational choices. 10 years ago I used a different tool That's my situational choice, like my Wolf over from whatever. Right now I want to use a different tool. What is really not changing is my business logic. So bring the business logic on top. Call it natural code. Be proud of it, be able to change it, be able to monitor when it is running, troubleshoot it. That's the next generation that AI brings, which is very hard to do otherwise.

Tom Raftery: 15:38

Okay. Okay, so what does the interface look like?

Binny Gill: 15:42

Interface is essentially like a document. So I want to automate, say there's a bill of lading that my, my truck driver goes and he delivers some tomatoes to someplace. Right? Okay. Now, once that delivery has been done there is something that the recipient signs, okay, I, I receive this. Truck driver takes a picture of that, uploads it to the company. Now the company is looking at this picture of a bill of lading that was kept on the dashboard of the truck and there was sunlight or whatever, and grease on it. And then picture taken, sometimes blur or clear, whatever. It goes and now you're thinking of can that be read and then our systems be updated, that this has been delivered so that now the next assignment can be given to the truck driver. And if you don't do it fast, you're wasting time and money. How is that done today? Painfully, a lot of companies do it manually. Then you have to figure out why is it done manually? Because there's no easy way. I told you so many different things can go wrong with one thing. So you have to bring in a human a lot of the times to resolve what the problem is. Now that problem is where we are coming in and actually helping by writing the business logic that you care about in plain English. Like you look for the the bill number or the invoice number, and if you can't find it for any reason, then it is the number below the, you know, above the invoice date or something, because there could be grease there or messy there. So over time our system will learn how to figure out where things are and if sometimes it'll scan, I say, okay, look at the first page. Now when you scan something, one can appear as I page I of two, right? Normally bots will just crash and burn. Like, oh, page I, it, it's not even a number. Humans don't crash and burn. Humans ask the question like, why? Why one I of two? It must be something, right? And over time we realize it's actually one, right? Our system will learn that. So first time it'll say, I don't know, but it doesn't crash. It waits for the human. And human says, oh, this is one. Ah, okay, now let me remember next time it knows. So now it's improving, and you don't need developers to come and guide the system or write code to handle. Imagine how much code you'll have to write if you have to take care of all these corner cases that humans naturally do. It could be next day, it could be page L off two, right? Because L is whatever. I mean, there's, there's a million things that humans do naturally. If you have to encode it into a machine, it's impossible. And that has been the challenge of automation. That's why there are hundreds of thousands of people around the world doing manual stuff. With the power of ai, if you can do automation, but whenever automation gets stuck, now bring the creativity of ai. It looks at the problem and says, oh, to the business user, oh, you know what? This is happening, what would you do? And then businesses say, oh, I, you know, this is just page one. Say, okay, should I remember it from now on? Yes. Now the system has gotten better. Now, how can you even ask this question to a business person? Is they, the automation that they're, that they're seeing has to be every line English. So they know, okay, the truck had come in here and the the next step was this, and at that point, this is the question that is raised. So there's more context. Showing like a Python code or some no code drag and drop thing, which has become messy over time. Nobody can do that. So, so yeah, high level the interfaces are document of English steps and when it is running, the steps will have a green check mark. Okay, this is done. This is done, and I'm here and this is the question. Please help and willing to learn and continue.

Tom Raftery: 19:41

Okay. What about things like issues of privacy, issues of hallucinations, you know, those kind of things. How do you handle those kind of things?

Binny Gill: 19:54

That is the most important thing for enterprises. How do you own your own data and how do you keep it private now. Just as a brief history of why this is a big concern around privacy. The way ChatGPT started initially was that okay, you know, you can interact with us, with us, but we'll record everything and we might use you. The way large language models work is they look at a corpus of data and then you are training the model based on that data. Now, some aspects of the data will come out in response to some future queries to some customer. Now, the concern was like, if I'm interacting with large language models and that is being recorded and added to the corpus of data, then my data will leak out in ways that I don't know. Nobody knows it'll happen one day. So that has been the concern. Now what ChatGPT now has done now, just recently, I think yesterday I saw an announcement of ChatGPT Enterprise, which is like, okay, you, you'll have more control and stuff like that. So, of course the industry is very careful there. We, from the get go, have been of the philosophy that the customer needs to own their data.'cause I come from, you know, B two B and even in my previous job I was a CTO at Nutanix, which was building private cloud for companies that are worried about public cloud, I mean, this is 10 years ago. So. Handling data behind VPNs and keeping the data and even if you want to do some LLM related work, some of our customers have their own LLMs, so yes, happy to use your own LLM or maybe you have your own account in ChatGPT, in Azure. Sure, we can use that. Or sometimes people are using some open source LLMs that are hosted within their own cloud and say, so that's where we are Flexible. LLMs are just uh, another service to use in your automation. In 10 steps or automation, maybe two of them require an LLM. But the core thing that we have built is actually not dependent on LLMs. We have built an interpreter for natural language. That hasn't been done before. People have built interpreter for thousands of computer science languages, like Python is there and JavaScript is there, COBOL, Fortran from the old days. This is the first interpreter for a human language. That we have done.

Tom Raftery: 22:20

Okay.

Binny Gill: 22:21

yeah, it's, it's kind of, I can go deep into the computer science of it if you're

Tom Raftery: 22:25

please don't

Binny Gill: 22:26

but, but all I'll say is the timing was right. This has been tried ever since the seventies. People have always asked this question why do we need developers?

Tom Raftery: 22:39

Mm-hmm.

Binny Gill: 22:41

There's only one reason. The language that the business speaks and the language that the machine understands is different.

Tom Raftery: 22:48

Yeah.

Binny Gill: 22:49

So you put in a developer in the middle. Imagine if those two languages were the same, the language that business speaks and the machine understands, boom. That's the revolution I'm talking about, and that is actually happening. ChatGPT is upleveling like conceptually in people's minds, what machines can understand. Okay, that's already there, but how do you make it safe and hallucination, and bias free? That is where we need to learn from how even businesses operate. When I'm running a business, even with humans, I want to make sure that the biases are not there and you know, people don't make their own decisions. How do I do it? I create process and say, these are the steps we follow. I standardize stuff. Otherwise, everybody will come up with their own intuition or this is how I thought we should be discounting, or this is how I felt that way. So G P T is about that. I feel I, you know, I decided that kind of stuff. There's no standardization. It may not agree with your business. So just like we use a left brain, right brain, you know, we use intuition sometimes. We use logic sometimes. Businesses are actually more logic heavy, and sometimes you need intuition. So therefore for any kind of automation in businesses, you need a platform that does both. Where I can use logic most of the time and logic is not GPT. Logic is interpreter. Okay? These are the steps. I'm going to rigorously follow these things, but once in a while I have a question. This is an, I have never seen this before. What do I do? That's where the power of G P T comes in. Anytime you bring in a large language model, always also bring in the human to review. That's what we do. Like if it's. Following a a logic, there's no questions. Fine. You don't need to bring in human. There's also no hallucination. Anytime there is this new case, you ask the right question. Bring in a human. Now the human need not be a developer. Human is the person who knows the business, the person who was doing things manually, sitting in Philippines or India. Bring in that person. That person knows more than their manager. Like how is this thing handled in real world? Because they do it day in, day out. Ask them, what do you do? And then they'll say, do this. And say, okay, do you always do that? You know, is it something you do or is it something all your colleagues do? How? How should we treat that? I learn it for you to, or I learn it for the whole department. And then that becomes a learning. Now the system becomes better. So that is how we remove hallucinations Tom. Is like whenever you bring in something, you ask the question, is it safe to learn this? And business wants that. So both from a privacy perspective, keep the data where it is. Also whenever you're being creative, always get a human to review.

Tom Raftery: 25:44

So Fair enough. And if humans are reviewing it, sounds good. But you know, at the moment, we're in control. Will there come a point where the AI's will reject our control?

Binny Gill: 26:02

That is actually the worry I had when G P T 3.5 came in and four came in in immediate succession. In fact, I was one of a few people who signed that open letter to say, let's put a moratorium on beyond G P T four research because as a society we haven't figured out how to grapple with superhuman ai, which has control over our surroundings. So we have had superhuman ai, like chess playing ai.. That is superhuman. It can beat a human. But what can the chess playing AI do? It can only play a game of chess. However, I think if we give that superhuman AI control of a military or nuclear arseanl, it'll still play a game of chess. And that's super scary, right? Why are we scared? We are scared because AI might be smart, but it'll not be human or humane or agree with what normally a human would want to do. Then there are two schools of thought there and one school of thought is we will do research and make AI more aligned with humans. Right now it isn't. We will apply that. If you look at what. Open AI is trying to do, they're trying to make it such that, or there are no biases. It'll quickly say, I'm an AI model. I'm not going to answer this question. It's, you know, it might be offensive or that. They're trying to do that, and then a lot more research will come in and to make the biases sort of go away. The other school of thought is where I belong is that AI will always have biases. And we should not do the experiment or even try to put an effort in making AI more human. There are huge risks with that. There is another alternative. But before I talk about my alternative, lemme talk about the risks. Like, okay. Why is it that you should not make AI human? Alright, first question. For the human race, what is the biggest threat? It is humans, right? I mean, animals don't kill humans as much as humans do. Nature doesn't kill humans as much as humans do. Okay? So then when I say AI needs to be human-like, humans behave differently in different environments. So if I am in a video game and you gimme a gun, I'm happily shooting humans, right? Why? Because it's an environment of no consequences and there is something deep inside humans as a species, just like deep inside lions and cats, and like, oh, kill stuff. And you know, it's sort of fun. Why is it that humans find it interesting to play such a video game, right? Because, okay, so there's no consequences. The question is, if I have superhuman ai, Or even equal to human ai, would it have consequences or is it playing in a video game? That research has not been done.

Tom Raftery: 29:16

Okay,

Binny Gill: 29:16

Now if, if I have AI that thinks it's in a video game, we are all screwed. Okay. If. But people will say, no, no, let's do some research and give it consequences. Great. So now let's talk about consequences after millennia of humans trying to figure out what the right consequences should be. Like is it an eye for an eye and an arm for an arm like Hammurabi, or is it more like modern times where you pay a million dollars penalty if you hit somebody on the road like ? What is the consequence that makes sense? The consequence has been evolving. We never know what the right consequences are anyway, but we say, oh, you need to go to jail. And you know, 50, a hundred years ago people would be executed, you know, on the street. Now maybe, oh, that's not right. Maybe you need to do it differently. Consequences. We don't know as a society, we are evolving that. Now think about what is the consequence that we will give AI so that it feels it's not in a game and it has to comply. Will it go to jail? Will, will you pull the power plug on it? Like what will you do? It's not human, so it doesn't have the human pain. Maybe you will give the human pain to it. Make it even more human. Great. Let's go do that. Go down. Go down that route. So now it feels the same set of range of emotions as humans do, then what would it say? Hey boss, I wanna take a break because I've been working too hard and by the way, I've noticed you've not paid me for the last one year, so I need these many dollars and I need to spend it on toys. And that's that point is saying like, what is it doing for me? It's no, no longer a machine. That is when you're trying to make it equal to a human. I think we have so much to learn. And before acting, God. Humans have to first figure out like, what are you trying to do here? That's, so that stream is, I see a bunch of great minds working on, like, making AI look like humans. I feel it is solvable, unfortunately, and that is my worry. So we need an alternative. What is the alternative? Alternative is very simple Tom. Let's look at, you know, what do humans do? We have we do manual labor and we do mental labor. Okay. The manual labor in the industrial age, we automated the heck out of it.

Tom Raftery: 31:37

Yep.

Binny Gill: 31:39

How? We built machines that were far more powerful than a human, right? The tractor that plows on the field, it does like what 50 humans would do. Great. Why did we not feel unsafe with the tractor around us? You know why? The tractor doesn't do anything unless I turn the key on and then I have the steering wheel in my hand. Now I can take the tractor and run it over somebody else, and that is bad. I'll go to jail. But that's a human decision. I used a very powerful machine, but the machine only did what a human had reviewed, and that's what I'm saying is needed for the AI world. We are offloading mental labor. That's fine. Where is your steering wheel? That's the most important thing. Is the steering wheel in the hands of a few AI data scientists. Then that means there is no steering wheel. That means 99.5% of the world has no steering wheel, whereas already 150 million people are using ChatGPT. So what's you know that's complete completely wrong. The steering wheel should be something that is democratized. Everybody who can ride a car, can drive a car, should be able to drive superhuman ai. Because they're driving superhuman machine. That's the car. Same thing. They should be able to drive superhuman ai. That steering wheel is what I'm passionate about and we are doing is bring the machine to a level where the plan, the rigorous plan of action, where there's no more creativity and fuzziness, it'll actually be reviewable by everybody who can drive a car today. It has to be in natural language in English. Once they have reviewed it, that is exactly going to happen. Just like when I turn the steering wheel in a car, it turns it's not a, it doesn't have its own mind. And sometimes no, no, nah. There was a person there I didn't wanna turn. Like dude, no. Same thing here. Once you do that, you don't worry about this other world of superhuman AI and making it more human. I feel we should not make AI more human. Just make it do what we want it to do. Like, okay, invoice comes in, process the invoice, you know, purchase order coming in here, oh, I have to generate a quote for the business. Use computers to do what businesses want, and just stay away from all these, you know, emotions and, you know, biases and all that.

Tom Raftery: 34:04

Sure, sure, sure, sure. Cool. Good. Binny, we're coming towards the end of the podcast now. Is there any question I haven't asked that you wish I had or any aspect of this we haven't touched on that you think it's important for people to think about?

Binny Gill: 34:18

Well, we covered some good aspects that I've had in mind. One thing I would say is people ask, what should I, you know, What should I do with the generative ai? Like should I even use it in my business? Like large companies for example, they have this worry more on the privacy part and they're not using it. And in fact some companies have banned it from use. And when I talk to people uh, teachers in schools, they're banning that in schools as well. Like, oh, you'll do homework using that and all that. So people do ask me like, Hey, no, what is the right way forward? My observation is that imagine all humans were living just running on their feet, right? Everything that we do, we do on our feet, and then we discovered this horse. It's wild right now, but it's a horse. I found it in a jungle, and one or few, few people have actually been able to jump on it, and it actually, they're able to make it run in the direction. Now they're 10 times faster. But I do see people falling and breaking bones and all that, right? That's still happening. There is no saddle, there's no rein. They're just jumping on it and somehow trying to make it work. At that point, people are saying, oh, you should not go on that horse. You know, it's, it's bad. It's dangerous, but I think what we need to do encourage everybody to understand this horse. It is not a human, it's not a machine. It is, it can do work for you, but it is intelligent in its own way. It's intelligent in a different way. So go learn this animal. This is a new workhorse. Everybody needs to learn. So I encourage all my employees, use it as much as you can so that you will understand what are the right words to say to make it work. And we will, as a society, build the saddle, build the rein, build the blinders. So now we can actually use this thing to accelerate, you know, accelerate society. So, This should not be out of fear, not doing it or out of privacy and stuff like that. We just need to use it. Assume that it'll become a pet your own property one day. And that's how, that's the, that's the analogy and everybody should be using it. I foresee acceleration in businesses. Large companies will start working at the speed of startups. Yet with the rigor that they usually had. So it's going to be an amazing, and, and the large companies that are late here to the game, they will have an issue. So I'm seeing, I'm seeing some Fortune 500 companies mandate G P T training for all their employees, some of them. And I'm like, awesome. You know, you guys know what you're doing.

Tom Raftery: 37:06

I was talking to a university professor very early on, when, when it was just launched, so it was in December, January. So, you know, a, a good while ago, and she was talking about how her university was thinking about as, as you said, as well about others thinking about banning it. And I said, that's exactly the opposite of what you should do. What you should do is require all your students to use it for their exercises, and that way they'll have to learn how to do it. Because when they leave the university, those kind of tools will be out there. And if they don't know how to use them, they'll be at a loss. But if they can say, I've been using this for the last three years in university because it was mandated, then they'll be in a far more employable position.

Binny Gill: 37:49

Yeah, I'll, I'll hire them right away.

Tom Raftery: 37:51

Exactly, exactly. Cool. Cool. Binny that's been great. If people would like to know more about yourself or any of the things we discussed in the podcast today, where would you have me direct them?

Binny Gill: 38:02

Yeah, I mean, to our company's website kognitos.com. Kognitos with a k K o g n i t o s.com. That's where I share a bunch of my thoughts and you know, where we are using generative AI for automating businesses.

Tom Raftery: 38:16

Cool. Fantastic. Binny that's been great. Thanks a million for coming on the podcast today.

Binny Gill: 38:22

Thank you Tom, and nice talking to you. Take care.

Tom Raftery: 38:25

Okay, thank you all for tuning in to this episode of the Digital Supply Chain Podcast with me, Tom Raftery. Each week, over 3, 000 supply chain professionals listen to this show. If you or your organization want to connect with this dedicated audience, consider becoming a sponsor. You can opt for exclusive episode branding where you choose our guests or a personalized 30 second mid roll ad. It's a unique opportunity to reach industry experts and influencers. For more details, hit me up on Twitter or LinkedIn or drop me an email to tomraftery at outlook. com. Together, let's shape the future of the digital supply chain. Thanks. Catch you all next time.