How AI has Changed Manufacturing with Ranga Ramesh

Episode #12 January 18, 2022

Data science and machine learning are integral parts of most large-scale product manufacturing processes and are used to understand customer needs, detect quality issues, automate repetitive tasks and optimise supply chains.

It’s an invisible glue that helps us produce more things for less, and in a timely fashion.

To learn more about this fascinating topic, I recently spoke to Ranga Ramesh who is Senior Director, Quality Innovation and Transformation at Georgia-Pacific.

Georgia-Pacific is one of the world’s largest manufacturers of consumer paper products and uses AI technologies throughout their manufacturing process.

In this episode of Leaders of Analytics, we explore how computer vision and machine learning can be used to classify tissue paper softness and instantly detect quality issues that could otherwise render large volumes of product useless.

Ranga’s work is featured as a case study in our recently published book, Demystifying AI for the Enterprise.

Episode Transcript

Jonas Christensen 2:13
Ranga Ramesh, welcome to Leaders of Analytics. It is so good to have you on the show today.

Ranga Ramesh 2:21
Jonas, thank you so much for having me. It's my privilege and my pleasure to do this. Really appreciate the opportunity.

Jonas Christensen 2:23
Yes, of course, more than happy to have technological experts like you on the show. And we're going to talk about a topic today that is very new to me. So, I am so interested in learning all about how things work in a big factory environment. That is foreign to me and I'm sure listeners will also enjoy it. Before I get to talking about you too much, we want to hear straight from you. So, could you tell us a bit about yourself, your career background and what you do?

Ranga Ramesh 2:57
Sure. I was born in India and lived in India to like got my undergraduate degree in Chemical Engineering. Went to the National Institute of Technology, which is part of the University of Madras. Actually I should say, while I was doing my undergraduate degree, I was always looking for some different area to kind of expand out from chemical engineering to further my education and get my graduate degrees. At that time, my brother used to work for a paper company in India, even though he's not an engineer. He was in finance. But I got exposed to sort of paper technology and paper in general and I found that interesting. So, I actually pursued a Master's and PhD in Paper Science and Engineering, which is what I did. So, I came to the States and did my masters and PhD in Paper Science and Engineering from the State University of New York, College of Environmental Science and Forestry in Syracuse, New York. So, that's where I did my graduate degree. And after that, I've pretty much been in paper all my life in my work career. So, started in R&D and Product Development for maybe about a third of my career in the beginning and actually even early on one of the areas they focused on was tissue softness. You know, how to make soft tissue paper and how do you characterise softness? I know we will get into that a little bit more later. But in a sense, that was my first area of study and then I did a lot of product development, and then actually moved into manufacturing for a few years. Worked in a actual paper making plant. So, I've got a lot of great experience there and then I've been in kind of corporate roles for the rest of my career at Georgia-Pacific. I currently lead the Quality Innovation and Transformation Team within GP and we can talk about that a little bit. It's been a great ride so far. I have no complaints.

Jonas Christensen 4:45
This show is about data science but I have never heard about paper science before. What sits under that category?

Ranga Ramesh 4:52
It has everything to do with paper properties and paper manufacture, paper as in all paper. So, you look at printing paper. You don't want to use printing paper to use it as a toilet paper, right? So it's very different. And if you look at packaging, or you know, all these different paper products that you use every single day, the paper is designed, customised, engineered for its end use, the appropriate end use. So, there's a lot that goes into that. I mean, fundamentally it is still material science. Right? Paper is a cellulosic fibrous material. So, if you think about it from that standpoint, it is like any other material, like cloth, and you could look at it that way and then think about all the things that go into how would you engineer a paper product to perform at its best level. So that's what that is and we'll get into paper manufacturing a little bit.

Jonas Christensen 5:41
Before we get to that, I'm actually interested in how did you end up getting into the world of technology and data science, specifically, when it comes to this paper work?

Ranga Ramesh 5:51
You know, I've always been kind of a math, data nerd, if you want to say that. Engineering of kind of makes you that way in many ways. You always work with data. My PhD thesis, my doctoral thesis actually involved mathematical modelling part of a paper manufacturing process. So, I always enjoyed working with data, even before pretty much early part of my career. And even throughout my work career, I've used data analysis extensively, statistics, all of that. So, data science, per se, is not new. But I do want to say this: It has evolved a lot. Right? So if you think about back in the day, if I'm doing mathematical modelling of paper manufacturing for example, I have to be sort of a subject matter expert. So, you use subject matter expertise, along with established, maybe, science and engineering formulas and laws to come up with your mathematical models. But today, what's changed over the course of all these years is just the ability to handle big data. So, what's really cool is, you don't really need to even be an expert, subject matter expert on the specific area you are modelling, right. So, you're able to collect lots and lots of data and then kind of let the analytics sort it all out. That's the part that's pretty exciting. Because you don't always have well defined laws and equations to describe every aspect of everything you're doing. So, you kind of use some of the analytics that they help you out with putting together these relationships and correlations. So that's what's cool about it now.

Jonas Christensen 7:17
Yeah, so I think we should delve into that a little bit more. But before we do that, I just want to hear a little bit about your role, so that we established that for the audience. And then I think we want to get back to exactly how this has evolved in your time there, because you've been there for quite a while, in this industry.

Ranga Ramesh 7:33
Yes.

Jonas Christensen 7:34
So as you said, you're the Senior Director of Quality, Innovation and Transformation at Georgia-Pacific, which is a huge paper manufacturer in the US.

Ranga Ramesh 7:44
Right.

Jonas Christensen 7:44
Could you tell us what the company does more specifically and what your role is there?

Ranga Ramesh 7:50
So, as you mentioned, GP as we call it, is one of the largest paper building products companies in the world. It was founded almost 100 years ago, in 1927, in Augusta, Georgia. So, we have essentially three major divisions within the company. We have consumer products, building products and we have packaging. The three big buckets of work. My career at GP is pretty much almost exclusively been in the consumer products side of things. In this division, we make lots of consumer paper products like toilet paper. You may be familiar with brands like Angel Soft and Quilted Northern. I know we don't sell it in Australia but they're pretty big brands here in the US. We also make kitchen paper towels like Brawny and Sparkle. We make Dixie paper plates and cups. We make Vanity Fair napkins. Just to name a few of the brands and we're also big in the ''Away from Home''. What we call ''Away from Home'', which is restaurants, hotels. Where you stay and all the toilet paper and towels that go into those. We're pretty big into that as well. So, it's a fairly large paper and building products company and I am sort of part of the quality capability and I lead the team called Quality Innovation and Transformation. It's been about three years since we established this team. I mean, if you think about quality traditionally, right. It's sort of always an after the fact activities. It's traditionally how is thought out, right. So, first you make some products and then you do some inspections and some testing and then you evaluate if the product is any good or not. If it meets specifications, expectations. I mean, it's all sort of after the fact. And many cases with the advanced technology these days, you know, we can end up making lots and lots of products in the time you're actually doing your testing and evaluation. So, it really is not conducive. The traditional quality methods are not really conducive to be more proactive and more real time, right. So, that's what the main mission of my team is. It's to come up with tools and technologies that will enable us to either build in the right quality to start with or at least be able to manage the quality in real time and this is done through technology and analytics, predictive analytics. So that's that's what my team is focused on.

Jonas Christensen 10:03
So, when you say those things, Ranga, I'm imagining things like computer vision or the light scanning real time, what is the client outputting, so, you can stop the press and change before you've produced thousands of toilet rolls. Is that the sort of thing?

Ranga Ramesh 10:19
Exactly right. Yep. So, you do vision inspections, computer vision, and all of that, to be able to inspect as you go. But more importantly, then you're able to predict failure modes, for when things don't go well, because you have all this data now to be able to connect machine conditions to the product quality side of it. You are able to make those connections to say, ''Okay, these may be the failure modes''. We want to run the process away from those failure modes, right. So, that's what you want to do. So, you're basically building in quality that way. So, you know your outputs much more predictable. That's the goal of the team, it's to be able to do that.

Jonas Christensen 10:54
Really interesting. For most listeners, I think, soft tissue, paper manufacturing is as foreign as building rocket ships. So, could you tell us a bit more about the actual data science process that goes into producing these perfect bath tissues and I suppose also how you're doing things to change the quality of the manufacturing then and there when you discover quality falls on the line?

Ranga Ramesh 11:23
So, let me start with talking about a little bit about paper making, in general, the paper manufacturing process. So, typically, all papers starts with a combination, right. Mixed up fibres and other functional additives, right. You may have softeners for soft tissue. Strength agents for papers that need strength. So, you start with like, sort of, a big soup of all of this, right. So,fibres and all of the additives and then you basically shoot it over a paper machine that actually makes the paper. Paper machines are huge, right. They could be 50-100 metres long. It could be 7 - 8 metres wide. So, these are very large, high capital machines. You basically start with a mix that's almost 99%+ water and the rest of the ingredients only make up less than 1%. So, the entire process involves taking this water out of the sheet and creating a uniform consistent product, right. So, that's what the goal is. That's what takes all of this size and capital to be able to do. And then with tissue paper, particularly, there's one fundamental difference. It's that in order to make the paper stretchy or more absorbent, soft, there is a interim step in the process called craping, which is essentially creating micro corrugations in the sheet to create a supple, thicker, absorbent paper. That's what makes tissue tissue. So, you trade off some of the strength for softness and then you make the product. So, from the paper making process, we end up with these giant rolls of paper, which then we take to our converting process, where it's cut down to the consumer size. Plus, we also can add some other functional things in converting. There's been like some lotionized tissues, where they actually add like aloe and other type of lotions in tissue. Could be scented bath tissue. Could be 3 ply, 3 ply. You name it. It could be embossed. So, all of these things can be done in converting once is all put together. So, that's what really the whole tissue paper making process involves. Then now if you're talk about soft tissue paper, obviously softness is a subjective parameter. A consumer says something is soft. What do they really mean? Right? So, you've got to understand the fundamental science behind softness and what one consumer may define to be soft may not be another consumers perception of softness. So, there also there's a lot of subjectivity involved in terms of you know, someone may like velvety softness, others may like satiny softness. So, you have different types of softness you can have and so all of these things go into designing the product and the process of how you do all this, right. So as I mentioned earlier, part of the work involves breaking consumer words and subjectivity down to material science and figuring out ''Okay, what do I measure, the physical attributes in the product that then we can correlate back to softness?''. So when that step is done, now you can generate lots and lots of data with that and then you can then tie those back to what the process conditions on the paper machine or all the knobs, right. So, what settings were used and all of that could be then put together and come up with predictive models that says, ''You're now making soft tissue''. And just having that kind of feedback available as the tissue is being produced without actually doing the testing is huge for managing real time the quality of the product being produced. I hope that made sense. If you have any questions on that, I'll be happy to answer.

Jonas Christensen 14:45
Well, let's start with the first part to that, which is how do you figure out what softness means in the first place. So, you've calibrated your models to identify when something is not soft, but how did you identify what that mean in the first place?

Ranga Ramesh 15:01
Lots of consumer research and subjective panels. I mean, obviously, at the end of the day, as I mentioned, it's a consumer defined attribute, right? Interestingly, what you will find is, in North America, consumers generally like their tissues softer and maybe less strong than if you go to Europe. Tissues tend to be stronger and less soft. Again, some of it relates back to how many sheets they want to use or these consumer behaviours are established over time, right? So, the different attributes are sort of traded off in different ways. I believe Australian paper is sort of in between the two, if I'm not mistaken. It's been a while since I've actually familiarise myself with some of the Australian products. Basically you have to start with what does your consumer want. Where are we marketing this? What does your consumer want? And then of course, then it's a lot of research. Trust me, we have lots of research. Details on people's toilet behaviour that you don't want to even know. It's one of those. It's like, ''Okay, TMI. I don't want to know more about this''. It's basically through, you know, whether it's homeostasis or you could do subject to panels and then that's kind of how you try to get at this phenomena. It's not any different really than taste, for example, in food industry, or fragrance industry if somebody made like a smell. I mean, these are all subjective parameters. What one person likes, the other person may not like. So, you're really trying to hit certain target segments with your product design. Again, that's why we have different products, right. So, we have products that are targeted towards the real, ultra soft, super soft segment of the population, which wants nothing but absolutely the most plush toilet paper they can buy and then folks that they want to go more trade off, strength and softness, right. For whatever use they have, they may want that. So, it's all those things that go into the product design.

Jonas Christensen 16:50
I'll tell you what, Ranga. I'm sitting here thinking that next time I go for an overseas holiday, I'll definitely be thinking about another tourist experience, which is figuring out how Australian toilet paper compares to whichever country I'm in.

Ranga Ramesh 17:05
There you go.

Jonas Christensen 17:06
Yeah,

Ranga Ramesh 17:07
My family always makes fun of me because that's the first thing I always notice. Whether I'm in a supermarket somewhere or staying in a hotel, it's like the the toilet paper is the first thing that come in a boat. So, it's hard to take it away.

Jonas Christensen 17:20
You know what? COVID-19 pandemic has definitely made us aware that toilet paper is one of the most important products that we have. I don't know what it was like in the US. But in Australia, we had fights breaking out at supermarkets over the last pack of tissue.

Ranga Ramesh 17:36
Same here, same here. There are YouTube videos about it. You can go google and see. Probably it's just as bad here as it was in Australia, i bet.

Jonas Christensen 17:44
Yeah, that's right. We never ran out really, other than the fact that people hoard it. But that's a whole different story. I'm sure you do demand and supply forecasting as well and that would have been hard during that period.

Ranga Ramesh 17:56
Very, very challenging, very challenging. Especially with all these variants coming in and out and you just don't know what do you have out there.

Jonas Christensen 18:06
I'm sure we can cover that a bit later. But before we get to that, I'm still interested in this manufacturing process. So, how hard is it to get good quality paper out of the machine? How often is the quality not quite right and perhaps how's that evolution evolved over time? So, I think you've been working in this industry for about 20-25 years. You must have seen this technological evolution really take effect.

Ranga Ramesh 18:33
Yep, absolutely. What's changed a lot is all of the process control automation. Things like that, right. So, you're able to control the machines a whole lot better than you used to do before, right. So, definitely, from that standpoint, unless you have some other type of upset that you don't know about, for the most part you can design a process and let it run there and be able to produce good product, right. So, that has not been a problem. Our focus has always been about making sure the consumer experience with our products is as good as it can be, right. So you start with a good design and then you try to minimise variation in the process through all of these technologies and things and tools available to you. So, that you can give a consistent user experience with your product. That's where a lot of advances have happened, which is a great thing. So, we're able to get much more consistent with the products we produce. Same thing applies even on the converting side of technology. I mean, the controls are phenomenal these days in terms of how tightly you can manage the process, so you don't have unexpected failures, if you will.

Jonas Christensen 18:38
So, Ranga, this process is actually something that has been documented in the book that I'm a co-author on, ''Demystifying AI For The Enterprise'', and you have some case study into this book, which we're extremely grateful for.

Ranga Ramesh 19:50
No, Thank you. It was my pleasure to do it.

Jonas Christensen 19:52
And in this case study you described how you got started with this whole exercise of using AI and machine learning to actually start predicting softness. So, at the point here, when you're starting to use these technologies, how did you set that up in the first place, because there must be a lot of trial and error in that process?

Ranga Ramesh 20:14
Sure, yeah. I mean, you have to go through several iterations of this, right, in order to get to where you are. Again, any model is only as good as the quality of the data that goes into it. S,o you have to ensure the quality of data is very good to start with. And then it's just a question of - You know, you have so many different choices on how you can model. Data scientists will tell you. So many different approaches to modelling something. So, we just go through an iterative process to figure out exactly what type of model would would suit us the best and be able to predict what we need to predict. And then, of course, you go through extensive validation. Right? So, to make sure that what you're predicting is, first of all, within what's expected and then how is it actionable for the folks on the floor, right, that are actually running the machines? I think part of the benefit I have in the sense of having been in manufacturing - I worked in manufacturing for a few years. I really understand how operators kind of think and operate. They've got so many screens they have to look at, normally on a given day. How do you provide the information you need in a way that still gets their attention, when their attention is needed, but doesn't over clutter them with different things that are going on. So, you know, it's kind of a good balance, you have to strike there, in order to make sure that whatever the output is, it's presented in a way that's actionable, immediately accessible for the operator and then be able to work on that. So, that goes into that. So, yeah, I mean, every model is unique. Every machine has to be modelled separately because each one is a little bit different and depending on the type of product you're producing, it could be a different model also for that. It's one of those things that takes time but once you establish a good system, then you can start kind of duplicating that process over time. That's kind of been our approach,

Jonas Christensen 21:59
You're highlighting something that I often say to my stakeholders, which is ''Data science is a team sport and we have to make something with data science that works for the end user''. It can be an accurate model. But if you cannot consume it with ease...

Ranga Ramesh 22:15
Change management is so critical. Right? So, they're used to running the machine a certain way. They look at certain things and without all these models. They've been running payper machines for years. So it's not like you're teaching them how to run the machine. It's just, you're giving them tools to be able to highlight problems before they happen. That's really where the key is. When something is trending the wrong way, you're able to tell them that, ''Okay, it's trending the wrong way. You may want to focus on this kind of thing''. If it is not actionable, the model is not worth anything. So, there's no use in having a model that's sitting in a screen somewhere in the control room and no one's really paying attention to it or actually using it. That's a key focus of mine. It's making sure something is usable and actionable from whatever you're putting out there.

Jonas Christensen 22:59
So, one of the things I picked up on there was the implementation of models and changing of process. So, how long did it take or does it take for you to build, test, validate, and implement a new model into a manufacturing process? And the second part of that question is, if you're comfortable sharing, what kind of performance uplift have you had in your manufacturing process from using machine learning to do these things vis-à-vis what you had before?

Ranga Ramesh 23:31
Right. I mean, the timing can vary, because of the number of data scientists you can throw at it. So, it's a resource issue, right. So, if you have a good team of data scientists, you could crank out a model in fairly short period of time. But it's more the validation, you know, the iterations you go through, depending on the quality of the data and how long it takes to get to that clean state.That can change machine to machine, depending on the quality of the data. So, you could do this in a month. It could take six months for one model, for example, just to come up with one model for one machine or it could be once you get it to a point, it could even be a week. Right? I mean, if you have the data ready to go, you could connect the dots and create the model in about a week. I mean, it all varies with where you are in the maturity state of your data science programme and that's what really affects it. Performance wise, I mean, it is all about our consumer experiences, I mentioned to you. So, the lower the variability of the softness of the output, - Actually, I mean, we still do validation with our traditional test methods, just to make sure. As we keep reducing that variability down, then we know we're getting there. So, that what we've been measuring. It's how much that variable is going down in terms of the product variation, so that you get a much more consistent product that is going to be much more readily accepted by our consumers. That is the goal. So, that's how we measure that performance. It's been good enough that they want to do more of this, which that's always a good statement when you say, ''Okay, keep going because this is exactly what we need, approach'' right? That's always a good thing when things are moving in that direction. It does take a period of time really to truly know the value of the model and so forth. But at least everybody realises it's going in the right direction.

Jonas Christensen 25:13
Yeah, and one of the things, we talked about the pandemic earlier, one of the things that I have been impressed with, overall, there will be cases where it's not true, of course, but one of the things I've been impressed with overall, is how local and global supply chains have managed to just keep up with fluctuations in demand, borders closing, etc. And I imagined these sorts of real time technologies are really important to keeping production going, regardless of challenges. I'm interested in your point of view: What are the biggest opportunities for AI machine learning, that category of new technology, predictive technology, to improve the way we manufacture consumer staples more generally?

Ranga Ramesh 25:59
Yeah, I really believe we're just scratching the surface. We're definitely leveraging machine learning AI in things like safety. Knowing where people are supposed to be and not supposed to be and using AI for that. Keep people from getting hurt. You know, it's a very important aspect of any manufacturing operation. Secondly, reliability. So, we can measure, monitor vibrations on machines, motors. You name it and then be able to predict when the failure might happen. So, it allows us to then kind of service the equipment before it actually fails. Preventing these failures. Right? So, that is huge in terms of keeping a reliable operation going. And then in quality, of course, we already talked about vision technology. You know, vision and AI combination, creating all these predictive modelling and inspection technologies is huge. I mean, I think we have so much more room to expand in that area. Same thing with material handling for supply chain purposes. You know, how we take the products and getting it on a truck and getting it out the door. There is just so much still to be tapped in this arena, I think. You're really going to see more and more applications of these. Because you know, at the end of the day, it creates a safer, more reliable, better quality operation. So that's exactly what you're looking for.

Jonas Christensen 27:18
So if you had to pick one out of all those, where would you start today?

Ranga Ramesh 27:23
I think the two big areas to me are still reliability, I believe, and quality, are the two big areas, I think. Because reliability kind of relies on equipment and preventing equipment failures. Equipment can be censored up so well that you can generate all kinds of data. So, the combination sensor technology with predictive analytics can help you with reliability quite a bit. So, you can run an operation reliably and not have failures and that leads to a lot of good things. Right? When you have a reliable operation, that means people are interacting with the process less because the machine is running all the time. They're not getting into the machine trying to fix something that's broken. So, from a safety standpoint, it automatically has a positive impact in terms of people-equipment interactions, right. And then on top of that reliable operations also tend to produce, in general, good quality product, because most of the time upsets happen when you have machine upsets that cause quality upsets in the product. So, definitely, I think reliability is an area where it's huge. I think the opportunities are quite a bit. And same with quality. I mean, you know, we talked about machine vision and all of that. Robotics too, right? I mean, that's another area where we look for material handling and all of that. You know, Automated Guided Vehicles (AGVs) and all that. AGVs are starting to become very prevalent and commonplace now to be able to move material around. And a lot of that involves technology, in most cases AI, to know what you're picking and what you're loading and things like that. So, I think there's tremendous opportunity there.

Jonas Christensen 27:26
Yeah, I think in my last few episodes here, I've really grown my own appreciation for just how the supply chain is made up of lots of disparate, very clever production technologies, transportation, logistics, analogies. And then what happens in retail stores, for instance. The difficult part in many cases is how do you connect all that up,, so that all these systems speak to each other and coordinate across the full value chain? Is that something that is worked on in the industry?

Ranga Ramesh 29:34
Absolutely. I think the demand management side and the supply management, tying those together along with inventory management, you're looking at all of these pieces. They all have to fit nicely. First, not so that you don't short the customer. So, you still have product when the product is needed on shelf. But secondly, also you don't end up with a lot of inventory. So, the other side of it is just warehouses filled with product or raw materials that you haven't produced with. So, all of these things are really important for overall economics of the whole thing and making sure your customer gets the product when they need it and then you're able to produce it just in time and be able to get it there efficiently. So, yeah, I think that's a huge area of work that's very active and happening, as we speak.

Jonas Christensen 30:17
So, Ranga, let's talk about the great toilet paper crisis of 2020. You had, I'm sure, huge supply demand shocks from consumers wanting to buy toilet paper now and all of it and then of course, that stockpile has to get used. And you had to, I'm assuming, really plan out your production and manage that effectively, throughout that period. How did you and the company go through that and how did you use technology to optimise how the business went through that time?

Ranga Ramesh 30:51
I probably won't be able to speak intelligently for all the technology involved in demand and supply planning, because that's not my area of expertise. But I can tell you this, at least just as an employee looking into what's happening. So, so many people worked so hard during that period. Basically, it was just all hands on deck, nothing holds us back. Just run as fast as you can just because of how much of a dire crisis it was in the US at least. I don't know about Australia, but definitely here, I mean, there were times when people were resorting to using some of the sort of commercial grade paper to meet their needs. So, it was definitely a pretty big challenge. So many teams had to work together hard and then, of course, our manufacturing partners, our employees had to do it safely and do it in a way that still kind of keeps the supply going. So, it was a pretty challenging time, without a doubt.

Jonas Christensen 31:46
Yeah, I can imagine the logistics of not just keeping up with demand and supply from a consumer point of view, but also your staff might get exposed to COVID-19. And you might have locked down departments that they need to go in quarantine or what have you. And all that would have been a big challenge for a production environment like yours.

Ranga Ramesh 32:06
Very tight safety controls and precautions were taken. People's health were top priority. So, want to make sure we provide a safe environment for them to work in and at the same time, still keep the lines running so that the customer demand is met. So yeah, that's definitely quite a challenging time. Not sure we're still the out of it yet but we'll see. Hopefully, we are.

Jonas Christensen 32:28
Yeah, it's coming and going a bit at the moment. This is recorded in December 2021. We're not far away from just learning about this new Omicron variant that we don't know that much about at this time. Maybe when you're listening in the future, you'll know all about it and what happened since. But at this point in time, it could be leading to the second toilet paper crisis of 2022. We don't know yet.

Ranga Ramesh 32:54
Yeah, I hope not. We'll see.

Jonas Christensen 32:56
Hopefully not. Now, Ranga, we're sort of coming towards the end of the conversation, which is so far been really interesting for me to learn so much in such a short time. Is there anything else that you would like to get across to the audience or anything else we should know about this subject area?

Ranga Ramesh 33:13
I mean, I think it's a fascinating subject area in terms of what we can do with AI and how things are advancing so fast. You're going to see more and more of this kind of technology really helping everyday lives. Right? If you think about it, you have manufacturing, doing all this AI and predictive analytics and all of that, then you combine that with like, internet of Things, to connect all of this communication wise, all these, how you manage the demand side of it and things we can do with that. The possibilities is kind of endless. I mean, I'm sure you've seen already Amazon things that that are there, for like, when you run out or something, you press something or you click something or you getting close to something, it automatically sends a signal to replenish all of these internet of Things, that technologies combined with what the manufacturers can do. I think it's a pretty exciting time. It's to revolutionise the whole marketplace think.

I agree and we're just getting started. I mean, these technologies are getting smarter, faster, more capable of handling bigger amounts of data and we're learning how to do it better, if I can say it like that. There is a natural evolution of these things. The internet today is not what it was in the beginning and AI technologies, I have to be careful putting a timeline on it, but it's definitely in the early spring of the technology evolution of this area. So, very interesting to see where we're at in, say, 10 years. Now, Ranga I have a couple of questions left for you. First one, which I always ask the visitors to the show is I ask you to pay it forward and that means asking you who would you like to see as the next guest of Leaders of analytics and why?

Well, it kind of runs along the same topic we just talked about. Advancement of technology and AI. One of the areas I'm always fascinated by is natural language and application of AI and natural language. Because it's a pretty fascinating area, whether you're doing consumer research or doing this or doing that, you know, having the ability to parse, I mean, natural language and being able to interpret all that to me, it's a pretty amazing thing. So, I ran into this company called DeepMind. You may have heard of them. And I just read recently about a new language type or model they have called Gopher. I don't know any specific people in that company. But I would love to have someone from DeepMind, maybe come on the show, and talk about that whole world of natural language application for AI. I think it's a fascinating area. Personally, it's just my personal interest. I'd love to see someone from there come in and do this

Jonas Christensen 36:00
Great suggestion. And I was playing around with some of the GPT-3 models the other day myself. It's pretty cool what it can do. I was sort of trying to compare it to version two, second generation of this and it's really come a long way. And in some instances, it's hard to see that it's not a human writing it. It still can come up with some really random stuff. But on the whole, it's very cool for generating sort of basic texts or emails or factual information. So, this whole speeding up of language is a huge topic that I'm very interested in, too. So, I love your suggestion. I will definitely find someone. We'll see if they're from DeepMind or wherever else. But that will be in the future. So, everyone, keep your ears peeled for that. Now, Ranga, where can people find out more about you get a hold of your content?

Ranga Ramesh 36:50
I mean, I'm on LinkedIn. So, definitely anyone interested in this area, I'd love to connect with and chat more about it. So that's probably your best bet. Connecting through LinkedIn. I'm on Facebook, too. But it's just I don't go there much. But LinkedIn is probably your best option.

Jonas Christensen 37:06
Yeah, I agree. I have recently cancelled my Facebook account. That's bad AI, in my opinion. It's not giving me a very positive...

Ranga Ramesh 37:15
I'm very close to that. Yeah.

Jonas Christensen 37:17
But the LinkedIn is such a nice positive environment. I find everyone's very supportive. So, everyone out there, go and look up Ranga's profile and hit the connect button. Going to be a good connection for you, I'm sure. Ranga Ramesh, thank you so much for being a leaders of analytics. You have really opened our eyes into something that I think for many people they take for granted, but it's a highly scientific process. I know that, maybe I shouldn't say like this, but I will be thinking of you next time.

Ranga Ramesh 37:50
I'm really happy to be on the show and talk about it. One joke we have in the family is like you know, all my education is going down the drain, can I say so. They all make fun of me for that. But it's kind of nice to come and talk about something different that maybe your listeners don't encounter all the time. So happy to do it. Thanks for having me.

Jonas Christensen 38:08
You're most welcome. Enjoy your day.

Ranga Ramesh 38:10
Thanks, you too.

How AI has Changed Manufacturing with Ranga Ramesh

Join Our Free Trial