Speaker 0 00:00:00 At through band, we developed custom software connecting systems, machines, and robots to ensure to unleash the full potential of data and AI
Speaker 2 00:00:18 Hello everyone. And welcome to a three, the robot industry podcast. I'm happy to have Hugh Foltz from VU. Been here with me today. Hugh has been supporting companies in the development and execution of their digital transformation for more than 20 years in 2018, he became co-owner and executive vice-president of <inaudible> a firm recognized for its expertise in artificial intelligence and the development of custom software solutions. Hugh, welcome to the podcast. I'm glad you're here.
Speaker 0 00:00:48 Thank you, Jim. You got to be here to,
Speaker 2 00:00:49 Hey, can you tell our audience a bit about <inaudible> and how you got started?
Speaker 0 00:00:54 Sure. Sure. In fact, we started Lubin was founded 10 years ago. We started at the beginning building software custom software for national defense. We are born in the, the good field to learn how to build very strong and very neat software. That's why the migration and the adoption, uh, ban of, uh, the new technologies like the like AI, like artificial intelligence was kind of natural and building a team in AI was also something that almost was already on, on, on most, in the path of the company.
Speaker 2 00:01:30 And you're based in Quebec, Canada. And why is that kind of important to the AI conversation?
Speaker 0 00:01:36 It's interesting because Quebec and Canada around the world is the candidate itself is one of the, let's say between three and five, there is many opinions about it, but between the three and five major polls and ma major, uh, leaders in AI. So, so that's why we are very lucky to have so many universities in Quebec to hire various, very good data scientists and AI scientists so-so to build a team like we have. So right now, more than 80 people, it's for sure. A lot easier to be in Quebec Qudexy. Yes. But also with an office in Montreal.
Speaker 2 00:02:15 Yes. And the correct. City's beautiful for those listeners who have never been there. It is absolutely one of those places you've got to put on your bucket list. So I want to ask you a kind of a certain confusion on AI, artificial intelligence and ML machine learning. And how do you explain the difference between the two technologies?
Speaker 0 00:02:32 Very easy. In fact, artificial intelligence is the common name that we give to all our guardism all our go to them that won't and have as a mission to mimic human and or predict what can be predicted with human data. So if we want a human, if we're going to mimic human in a manufacturing environment, AI can, can be very helpful and can do a lot to do that. Meshing learning is just a subset of algorithm. That that is part of the big word of artificial engine intelligence today. And also also very common is the deep learning and machine learning. So all those subsets, and there is also aware of, uh, operational research that is another substance. So there, there is many different categories and our families of models under the very large umbrella that we, that we call artificial intelligence.
Speaker 2 00:03:29 Thank you for that. And what's happening you in artificial intelligence, especially in advanced manufacturing and robotics, what are some of the trends?
Speaker 0 00:03:38 There is a lot happening, uh, since a few years. In fact, we, we are facing kind of, uh, we, we are, I feel at the tipping point where AI N not only AI, but also everything related to data in our, in the manufacturing world is finally taking off. And it's, it's nothing. It's not nothing to say that because for, for many, many years, what do we can call also the digital transformation in the manufacturing world was not, there was not a good understanding, I would say, and probably not a rush in many, many companies to do that. So, uh, what, what do we see today is one very huge pressure that comes from the fact that there is no more manpower to produce and the manufacturing world is struggling. So they are looking to automate as much as possible and everything not as possible to automate. And so that's, that's also, uh, part of the reason then probably a big part of the reason why AI is so huge trend and so many projects are going on right now.
Speaker 2 00:04:45 Do you feel that one of the trends is kind of a FOMO? We say in English, it's like a fear of missing out. So is there a bit of a fear factor in AI?
Speaker 0 00:04:54 Yeah, probably, yeah. Right. I think there is probably a FOMO trend, also a are for sure the, they are afraid or they are scared. They are scared for, for many reasons because it's, uh, at one point and we, we are in late, there is nothing new about that in Quebec and Canada and around Canada and us many, few years ago, us was almost at the same level of automation that the Canada was. But the thing is, as we know Americans, they, they, you know, w when you, you, the giant, they normally stand up and fight back. So they got it exactly. Like we got it by the pandemic. The thing is their reaction was at the very beginning of the pandemic to stand up and invest massively information and include including AI and robots and everything. So in the few years, so in two to three years, we are getting in, we are, uh, we, we are in late now, so compared to America, so to, to us, so today, uh, I think there is clearly a fear of not being able to catch up. So, so probably something related to this movement and trend also.
Speaker 2 00:06:03 So who is buying artificial intelligence? Is it kind of the, C-suite like the president, or is it the it or the OT departments?
Speaker 0 00:06:11 I think, I think the first to be seduced by the potential of artificial intelligence is the C-suite yes. At the end of the day, the it department must be also convinced. So we always have a tour and many people to, to convince them, to bring on board for all the projects that we deliver. The good thing that I can say is, at least today, it's much more common and a lot better understand what can, what, what can be done with AI. And also what is probably too far fetched to look at with AI, because for many years it was the Sangra that the AI can do everything and pretending it, that, that it can do everything, but it's for sure not the case, obviously. Uh, so this better comprehension of what is really AI and also looking at data first, not only AI as a buzzword is a big change and it simplifies a lot of our life for, for, for sure.
Speaker 2 00:07:09 So what does the typical project look like for <inaudible>? And when you talk about AI in manufacturing
Speaker 0 00:07:15 For, it depends if the company or the manufacturer is, uh, already in, is what do we like to call digital transformations for many years or not? If the company is at the very beginning of his, of his ARIDE, we are looking for like the good old principal. And in project management, we look for small project with very high impact. And with also low-risk. So we have to manage all the risks, including technological risk. Of course, we are looking for if possible, a zero risk technologically, so small project and quick wins all of the projects that we are trying to tackle at the beginning for a company that didn't start, let's say that there's not automate that much for the rest. If, if, uh, and we, we have many large companies and the large clients that I have a ultimation in their factories since many years for them.
Speaker 0 00:08:13 It's another word, because at the beginning, when you look for quick wins, you think about predicting, uh, sales, you look for quality control, including cameras to, to replace human looking at things and trying to see if there is a defect or default and also sizing thing. So there is a lot of small projects and looking abroad. You can also look at automate the entire supply chain. So that's, that's the real end game for, for many manufacturer in the world. So that's where the replenishment that can be done in real time for gathering all the data that we can along the supply chain. So from the very beginning of the chain, if possible, in the factory of your suppliers to the end and the other end of the supply chain. So the customer itself, we can almost do magic, meaning that we w we can rebalance the supply chain in real time and ensure that the company would produce not only at a, at a good speed with a good efficiency and everything, but the right thing to maximize the profit then. So, so a lot of concepts that are not simply possible if you haven't tried to calculate them to make it by itself, by him, by himself. So if we, if the algorithm is train and well-trained to do that prediction, and the algorithm is giving the good advice and insights. So it's, it's almost like saying the factory is managed, uh, by the algorithm at the end of the day.
Speaker 2 00:09:38 Yeah, no, that's makes a lot of sense. So how long does a project take? I know that's kind of an unfair question because some project, how long is a piece of string, but maybe just give our audience an idea of how long this would take
Speaker 0 00:09:51 For the same two categories that I mentioned for, for our company that start, and for the quick wins, we should look at again, small projects that must take maximum three months to deliver three months. And if let's say that the algorithm and the project is delivered, there is a training time after that. So probably in six months maximum, the projects must be install and deploy in the engine, in the industry or in the factory for bigger project. When we attack a supply chain, it takes at least a year, probably to two years to finalize. And there is also a concept, very important in AI, it's the training and the retraining of the around. And so it's never over. So, so the algorithm, we get more performance and more accurate over time. So even if we deploy for first time to manage the supply chain of a factory, for sure, a year after with more data and more training, the algorithm would be again, more accurate, more performance, there's more gain. So those phases of maintaining the algorithm in the system, uh, are normally part of the process. And, uh, in the right,
Speaker 2 00:11:04 When we talk a little bit about some of the use cases that you might have for robotics in manufacturing.
Speaker 0 00:11:10 Sure, sure. But robots are always part of the equation for us. It's it's of course, if a company contact us and say, Hey, we want to automate software where we do at VU ban is software. So, so we know, uh, very well the players in the robotic word, but we are not neither an integrative robot, neither, uh, uh, robots specialist by itself. So, but we connect with robots. So, so the point is automate a factory needs, very large variety of solutions, and robots is almost always around there. So we always suggest to our customer to look at their digital transformation, again, as a, as an objective. And that must be very high in their, uh, priorities for the year or for, for the next 10 to 15 years. But which project do you pre-authorize? I think, I think if the priority is to put a robot, because there is no more manpower for this specific job, it can be the good thing to do before looking at AI and any other things. So it's always a question of, again, what is the, what is a project with the higher impact, with the less difficulty to implement and fast and size that can be delivered and implement in the company? So the question must include in AI is exactly like robots simply, uh, another tool in the toolbox, that's it?
Speaker 2 00:12:44 Yeah. That's a good way to put it. Thank you for that. I guess you may have already answered this about some of your prospects. Are they doing prototype projects, like small projects so they can show these quick wins?
Speaker 0 00:12:54 In fact, they have to do it every time. The thing is, even if I said earlier that it can be magic or close to magic, it's not magical at all. And it's an algorithm is simply mathematics. So once, once we say that, it means also that we have to do a proof of concept to ensure that we are not working for nothing. And we also have a project that is not only manageable in regards of the risk, but also that will deliver the performance and the accuracy that we are looking for. For instance, if, if we, if we want to automate the process and, or predict sales, for example, to keep it simple sales, predicting sales is something very common by the way. So in the manufacturing world, we have developed so many models, uh, for, for so many kinds of industries to predict their Salesforce for a year in advance.
Speaker 0 00:13:50 And that's all, we'll always the goal more or less, and the real name of a good, that good rhythm or the, the real end game of an algorithm per the predicting sales is the accuracy. If we predict sales within a courtesy of 60% at 50% is probably it's probabilistic at 50%. It's like flipping a coin. So clearly if the data is not correlating enough, we will get poor result in a poor algorithm predicting something without enough precision. So this question of doing a proof of concept to ensure that we can get, as let's say fast enough, the right precision is key. And it happens that we kill a project at even just doing the proof of concept we can say. And it's, uh, it happens in the past correlation was not strong enough. And we say, it's not gonna happen. We don't have enough data. We should look at another project. So yes, we have to do that all the time.
Speaker 2 00:14:48 And that's a good segue into my next question. What is ROI of AI? I guess it depends on the industry, but I'd be very interested to hear your comments on that.
Speaker 0 00:14:57 It depends if we are tackling or working on the right process, because, and it's always the same question, but one of the two, let's say the two questions that I like to ask before we start a project in AI with all of our customers, is where are you making a lot of money? For example, for this one is the first one. Then the other one is where are you losing a lot of money? If we tackle it, if we attack those processes or department, whatever it is, and focus on those department losing and very, very profitable or losing a lot of money and we optimize them three or 5% optimization where you lose or where you make a lot of money is of course, a very short ROI. So questions very easy, like that can change the game. And we are looking for, for game changers also. So the good answer is if the ROI for your first projects in AI, if the Yarra is longer than a year, probably it's not the right project. So that's normally what I say to my, to our customers.
Speaker 2 00:16:04 No, that's very interesting on the hardware side, like for the projects you're building, are you into like, do you prefer the new builds, like new equipment, or do you prefer retrofits where you're coming into it, maybe, uh, a system in a factory.
Speaker 0 00:16:18 Th this is a cool question because there is a very, a huge trend also to, to use those old machines that are in our factories and refurbish them. And I love this trend by the way. So for, for one reason, the actual robots that are available and so cheap today compared to 10 years ago, can, end-all the thing that are not possible with the actual machine, but with a few ad-ons on a machine, including cameras and training and the rigor to them to identify and do the QA and everything that we take care of food ban and working with a robot and the curator we can give to a machine and other 10 to 15 years with a pretty small investment. So, so yes, it's something fairly to consider for many companies. So Ratto to, to look for a brand new machine and only one can still do the job and with a good upgrade, not so expensive. It's crazy way we can do.
Speaker 2 00:17:21 That's interesting that you can give old machinery new lives by putting a new layer of AI. And like you say, cameras and sensors and such around it. So that's really interesting. And how much are people spending? I just, I know I don't, I'm keep giving you these questions that are really hard to answer, but maybe as a percentage of machinery. So if I'm going to maybe build a million dollar machine, should I put 10 or 15% around that machine for AI?
Speaker 0 00:17:47 I tried, I tried to not link the investment in AI and or in digital transformation to do machine too much because the right number is how much are you investing in technology by itself? Let's say all technologies and the right number was, uh, published by, by a few, a few, uh, companies, a few consulting firms that I want one good studio about it was done by Deloitte. And the right number is between three to 5% a year. That must be normally and not on so three to 5% of the revenues of the company must be reinvest in technologies. And it includes machine. It includes hammering AI and everything. So the, the, the play and the real end game, again, must be to disconnect the dependency between human and the operation and or the factory. So I want to say it's normally 10% to, to optimize and automate a machine and give a layer of AI on the top of a machine.
Speaker 0 00:18:52 It's not true. Uh, I prefer to say again, if you look at the right small projects and or at the projects that are providing a very good ROI with low-risk again, if, if possible, no risk and about this risk, by the way, an interesting parenthesis that I always say AI is at this stage since already many years, where there is no research, neither and clearly not fundamental research anymore that need to be done in four 95% of the projects. So that's something also very important. So the manufacturer or that are discovering AI are still a little bit in the concept that it is probably not mature. And the AI is, is it ready or not? It is really, and it is so ready that the algorithm that we use and that we implement it in almost all our projects are available out there. So, so the team that we have, even if they have their PhDs and post PhDs, the key is the, those guys have done their research when they, they were at universities and they were bored, the research, they were looking to apply AI. So it's clearly what the demand for manufacturers must understand. AI is mature and nothing about research on, or for the mentally research must need to be done for, for the vast majority of the projects.
Speaker 2 00:20:21 You probably hear this quote all the time, that data is the new oil or data. This is gold. Do you have any comments on that too? Is that wine you might use?
Speaker 0 00:20:30 Yeah. W we use this one effectually actually, and yes, data is a new oil and had if issue, if you know, the Genesis of it, but data is the new oil is, and there is also, uh, a link with AI that I can give you two to complete because data is a new oil is referring to the fact that there is can, can be, can be, I can make you rich as a company. If you, if you use your data and you empower your data. And we, we are, uh, as you, as you, as I'm sure you, you know, that we are very strong believer of it. But the thing about this quality is also saying comparing data to, to oil is also the point where oil, uh, so, uh, rough oil in a barrel. If, if we put a barrel in, in the room, the only thing that you can do with the, the Dior except putting fire in it is nothing.
Speaker 0 00:21:22 So it must be, uh, refined, must be treated, must be converted and everything. So it's exactly the same with data data, without the good effort to clean it, to, to take care of it, to label it and everything, to put it in the data warehouse in the data, right? So all those concepts are real and are very important. So raw data, yes, it's very important, but you cannot do that much with raw data only. And for AI, the next quote that we all always say after data is a new art is AI is the new electricity. So it's referring to the fact that exactly like electricity changed the world around a 100 years ago. It's happening right now with AI, the same, the same change, uh, in the business models and in everything surrounding us in our lives and our personal lives and jobs is changing.
Speaker 2 00:22:16 That's very exciting time. He was there anything that we haven't talked about that you'd like to a dimension?
Speaker 0 00:22:22 No, I think the message is clear. Uh, my main message is always the same. If we have something to hear about AI is it's really, and I think all companies must jump into training. And, uh, you mentioned it earlier. It's time. Uh, we, we have a powerful technologies available and automation is key if we want to still have a strong economy in Canada. And yes, AI is clearly part of the equation to get there.
Speaker 2 00:22:49 How can people get ahold of you if they're listening and they want to understand more about AI and what Ruben does.
Speaker 0 00:22:55 We have many, many, uh, documents, uh, white papers available on a website and, uh, um, for French people that are listening, I'm also writing blogs for less, uh, in Quebec every month, I'm writing a blog. So we, yeah, w we are communicating, uh, as much as we can offering conferences also to, uh, kind of, uh, democratize AI and also simplify AI for, for the concept at least. And of course my, my email, uh, uh, at van.com, uh, is, uh, easy to reach me.
Speaker 2 00:23:30 Great. Thank you very much for coming on. And, uh, we'll look forward to chatting more in the future.
Speaker 0 00:23:36 Thanks, Jim. Thank you for the invitation.
Speaker 2 00:23:38 Our sponsor for this episode is Earhart automation systems, Earhart builds and commissions turnkey solutions for their worldwide clients. With over 80 years of precision manufacturing, they understand the complex world of robotics, automated manufacturing, and project management, delivering world-class custom automation on time and on budget contact one of their sales engineers to see what Earhart can build for you. And they're at
[email protected] and Earhart is spelled E H R H a R D T. I'd like to thank our partner, a three, the association for advancing automation. They're the leading trade association in the world for robotics, vision and imaging motion control and motors, and the industrial artificial intelligence technologies. And you can visit
[email protected] to learn more. And I'd like to thank our partner painted robot painted robot builds and integrates digital solutions. They're a web development firm that offers SEO, digital, social marketing, and can set up and connect CRM and other ERP tools to unify marketing sales and operations. And there are painted robot.com. And if you'd like to get in touch with us at the robot industry podcast, you can find me Jim Beretta on LinkedIn. We'll see you next time. Thanks for listening. Be safe out there. Today's podcast was produced by customer traction, industrial marketing, and I'd like to thank my nephew, Chris gray for the music, my partner, Janet, and our partners <inaudible> and painted robot. And of course our sponsor Earhart automation system.
