Filip Popov (00:00)
Hi there, and welcome to another episode of Espresso 4 .0. Today, we have a very special guest who needs no introduction. We will introduce him anyway. It is number one on the Lytika list of influencers on LinkedIn. Number one influencer industry 4.0 on LinkedIn. You all know him. He is the king of Industry 4.0, the prince of digitalization, and the Duke of IoT — all the way from the Windy City, Jeff Winter. Jeff, welcome and thank you for grabbing a cup of coffee with me.
Jeff Winter (00:38)
Yes, thank you for having me here.
Filip Popov (00:40)
Absolutely. Jeff, why don't you tell us? Why don't you start us off and tell us what you do, actually?
Jeff Winter (00:47)
Sure, so I am the senior director of industry strategy for manufacturing with Hitachi Solutions. What that basically means is my job to point my company in the right direction to be a world-class digital transformation solution provider. In order to be successful, I've got to be intimately involved with what's happening in the industry, which is why I participate in so many associations, advisory boards, academic groups, and even research teams so I can stay current on the latest and greatest and then I just love sharing that with the world.
Filip Popov (01:19)
Excellent. And share what you do and thank you for that on behalf of the industry 4 .0 community that's on LinkedIn. Okay. Thank you very much. I am not going to coddle you because you're a veteran at this. You've been at this for a while. So, let's take a cold plunge with a question like, is Industry 4.0 a scam? Is it just a marketing ploy, or is there some substance to it? And what is that?
Jeff Winter (01:47)
I think that this is a fun question. No one has ever asked me this before. So, I'm going to kind of start it a little differently. So, imagine stepping back into a time machine, traveling back to the early days of the first Industrial Revolution. If you were to ask a local mill worker in the 1770s whether the steam engine was a revolutionary leap, they might scratch their head, just unaware that they were standing at this dawn of an era that would stretch decades and completely change everything. Now fast forward to today's discussion about Industry 4.0, and just so it took a long time for the steam engine's impact to be fully realized, Industry 4.0's revolutionary changes like artificial intelligence, the internet of things, digital twins, advanced robotics, they're all still unfolding. We may not see the dramatic overnight changes that a single invention like the steam engine presented overnight, but we are definitely in the midst of a transformation that's integrating the digital world with the physical world.
Now, it's true; the term industry 4.0 has been around for over a decade now, and some may now see it more as a buzzword than a revolution. However, like the roughly 80-year span of the first industrial revolution from the 1760s to the 1840s, at least according to National Geographic, these things take time. The technologies defining Industry 4.0 are increasingly becoming integral to how industries operate, showing that while it's potentially a slower burn, it's by no means just smoke. So, no, Industry 4.0 isn't a scam. It's a longer play. It's a more sophisticated game that requires a bit more patience to appreciate fully. After all, it's a revolution in the way the whole world works.
Now, one notable aspect that distinguishes Industry 4.0 from previous Industrial Revolutions is its strategic institutionalization. And that was really spearheaded by Germany. In 2011, as part of its high-tech strategy for 2020, Germany introduced the world to the term and the concept of Industry 4.0. And since then, it's taken hold all around the world, causing dozens of governments to make similar declarations and investments. And its evolution is driven not just by technological breakthroughs but also by concerted governmental forces. So, while the term may have been catalyzed in part by policy and economic interests, making it potentially seem a little like a constructed concept, the resultant focus and investment in technologies and advanced manufacturing are very real and progressively reshaping our industrial landscape.
Filip Popov (04:44)
Got it, got it. In fact, if I can summarize it clumsily, it's a group of different technologies that aggregated together, make our, bringing about the change and the revolution in the industry as opposed to the time machines and steam engine that you've mentioned in combination with other external factors that are pushing it as well.
Jeff Winter (05:08)
It really describes the era that we're living in right now. And so, the vision of Industry 4.0 is different today than it was ten years ago, but it's all around the same fundamental concepts of what we see changing.
Filip Popov (05:19)
Gotcha. Gotcha. Thank you very much. So, in the past few years that you've been very active on LinkedIn, you've actually garnered a lot of followers. You've recently reached 100K. Congratulations on that. And I'm sure you've had a lot of conversations as a consequence. I'm sure you have come across a similar kind of objection as I did in my puny amount of followers, whatever that may be. Talking to some seasoned process engineers, I've encountered people saying stuff like, in my time, we used to call those statistics. And nowadays, young people are calling that machine learning and AI. What do you say about that? Is there a fundamental misunderstanding? Are they related? How do you answer to someone who's not a data scientist?
Jeff Winter (06:16)
So, it's true; I would say the roots of machine learning are firmly planted in statistical theory. And if you look at both machine learning and if you look at statistics, one of the things that unite them is they both deal with data, and they both solve problems, which makes it easy to see them as being the same and confusing. And yes, quite a few people who especially have exposure to statistics but are not statisticians, like engineers; they're the ones, in my opinion, who often view this as getting a little confused because they know enough to be dangerous or just be a little wrong. So, let's go over the difference. So, statistics is primarily concerned with inference, which involves drawing conclusions about a population based on sample data. Machine learning and specifically supervised learning has a focus on prediction. So, it involves training a model on a labeled data set where the correct output is known.
The goal is to learn mapping from inputs to outputs to make predictions on new unseen data. This type of machine learning is the one that is often confused with statistics just because of how it's working. But that doesn't mean that machine learning is all the same everywhere because unsupervised learning doesn't really focus on prediction in the traditional sense. It aims to discover patterns, groupings, or structures without predefined labels or outcomes. So, when we talk about confusion, it mainly deals with supervised learning. Now, many of these supervised machine learning techniques are drawn from statistics, linear regression, and logistical regression, for example, but they also pull from other disciplines like calculus, linear algebra, or computer science. Now, interestingly enough, with the abstraction of machine learning from statistics with the use of libraries, that's often why some individuals even make the argument that the knowledge of statistics is not really needed anymore to do machine learning. So, they're starting to diverge more with time.
Let me give you an example for manufacturing to help highlight the difference. So, let's say you have a manufacturing company that produces electronic components, and the company has a goal to reduce the rate of defective products and improve overall production quality. So, the statistical approach is to understand the factors that contribute to defects in the manufacturing process. So, they're going to use statistical models like logistic regression to look for significant factors that will increase the likelihood of defects. Now, with machine learning, the objective is to predict whether a new batch of products will be defective or not based on similar or historical data. Here, they may use a decision tree classifier or more complex models like random forests or neural networks if the relationships are not linear or involve interactions. But here, the model can predict defects in real-time for or from production lines. Now remember, that example was chosen where machine learning and statistics could be easily confused. If we were to have a totally different type of machine learning, like deep reinforcement learning, there should be no confusion because it's fundamentally different from statistical methods. So, I hope that helps, but I can understand the confusion.
Filip Popov (09:43)
Indeed. Thank you. Thank you for enlightening us, and it does help a lot. Thank you also for simplifying it for myself and some other viewers who may not have data science backgrounds. That is one of the obstacles and challenges that I face when talking about machine learning in Industry 4.0. But what are some common objections to Industry 4.0, and how valid do you think these concerns are other than the one you've just mentioned?
Jeff Winter (10:16)
So, across the board, I would say everyone knows Industry 4.0 isn't just knocking on the door. It's halfway through the entryway. The discussion in boardrooms and even break rooms isn't about whether this will impact the workplace but more about the urgency and the timing. Do we have to dive in headfirst, or can we dip our toes and slowly go in? It's not about if we need to adapt to Industry 4.0. It's really about when. Industry 4.0 is a big deal, which means it's expensive. And it isn't just expensive; it's also transformative. Because if it's done right, it fundamentally changes the entire company. It changes every department and nearly everyone's job within. Oftentimes, how does the company even make money? Now, the value, the benefit, and the reasons why people do it for this transformative power are also, I would say, one of the biggest reasons why companies don't do it or are hesitant because transformative to a lot of people means disruptive, which can be good or bad depending on how you look at it.
What is interesting is I actually just read a study just a couple of days ago that was done in 2023 by Tsinghua University aimed to figure out and explore the motivations behind the adoption of industry 4.0 technologies, but they aimed at exploring the motivations behind the adoption of industry 4 .0 technologies among manufacturers in developing economies. And they were trying to find out if their technology pushed if their market pulled, or if they're government driven. They analyzed something like 215 companies and found that the adoption is not predominantly driven by the inherent technological advantages alone. So, like technology pushed, which is interesting because what they found is that the market pulled factors such as competitive pressures and government driven incentives such as supportive policies and other incentives, they play the more critical roles in influencing firms to adopt industry 4.0 technologies. Now, if we look at the barriers, I have two different ways of answering this.
According to Rockwell Automation's 2024 State of Manufacturing report, which is a survey-based report, the top reason was cost. The second was the skills needed to implement. The third was a lack of skills to optimize. And the fourth reason was employee resistance to change. So, three of the top four had to do with people. And in the manufacturing industry, especially in the United States, there is a shortage of manufacturing people. Deloitte this year predicts that just in the US alone, the manufacturing could be short by, or sorry, need as many as 3.8 million new workers by 2033, and they could be short by roughly 1.9 million workers if something doesn't change. Now, the second way I would answer this is there was a study done by Helion in December of 2023 where the researchers prioritized barriers to industry 4.0 adoption through the application of something called TISM, total interpretive structural modeling based on literature review. At the heart of their study, they found ten major barriers to Industry 4.0 implementation that were identified, ranging from IT infrastructure to training deficiencies. Yet when the dust settled, it became unmistakably clear that the commitment from the top, the top management, and the top leaders of the company wasn't just another barrier; it was the barrier from which all other challenges stemmed. And this finding underlines a critical truth. The success of Industry 4.0 hinges not just on the sophistication of technologies alone but on the strategic alignment and visionary leadership at the top. And I would argue this answer supports why people fill out cost as the top barrier from the Rockwell Report.
Filip Popov (14:37)
Gotcha. Wow. Thank you. That's very interesting insights, actually. I've taken some mental notes on that as well. Thank you very much. In terms of the value and impact of AI, how can companies understand the impact of AI when it comes to process optimization outside of just sales and marketing?
Jeff Winter (15:01)
So, one of the ways I like people to think about AI is you need to remember that, at its heart, it's a decision-making tool. And so you can either use it to augment or automate decision-making. So, in areas like demand forecasting and predictive maintenance, AI supplements human judgment with data-driven insights, enhancing decision accuracy or the speed of decisions. However, in process optimization or any other control uses, AI's role expands to automating decisions entirely. You're basically letting AI take the wheel. So, the AI would actively manage and adjust processes and parameters, analyzing real-time data to optimize production, and that can lead to reduced inefficiencies and improve overall operational effectiveness. But this shift from augmenting to automating marks a significant transformation in how companies leverage AI to achieve greater precision and efficiency in their operations. And it's one of the areas I think that makes manufacturing more unique than other industries is the amount of AI that can be applied to the automating of decisions in the actual production processes.
Filip Popov (16:15)
Gotcha. And you've mentioned some of the buzzwords earlier. There's another one, scaling AI, right? Now, can AI really scale in manufacturing, or is it just another one of those buzzwords?
Jeff Winter (16:30)
Well, can it scale? Yes. Is it easy? No. Actually, it's quite hard. And I would say there are four things that make AI hard and kind of unique compared to other attempts to scale technology. The first is talent. Simply put, it's hard to find AI talent. There's often a shortage of skilled personnel who understand AI in general, let alone AI and manufacturing processes deeply enough to implement and scale AI solutions effectively. Typically, AI isn't something you can just pick up in a one-week training course. It requires years of training to be good. And since AI kind of took off in the past few years, it's been hard for the world to catch up from a talent level.
The second is data. So, AI can't work without data. And oftentimes, if not all the time, it needs a lot of it and more than other technologies. Unlike IoT and digital twins, and other technologies that are really doing processing in real-time, AI often needs historical data. And with that, manufacturing processes generate vast amounts of data, often noisy and unstructured. So, ensuring this data is clean, well-organized, and useful for training AI models is a significant hurdle.
The third, I would say, is explainability. AI models can often be easily explained, the model itself, but individual answers produced by AI models can be quite difficult. In manufacturing, where safety and precision are paramount, this black box approach of many AI systems, at least the way that people think about it, can hinder a broader adoption and scalability because people may be fearful of being able to explain how exactly that decision was made. And the last I would say would be maintenance, or more specifically for AI would be model calibration. So, AI models require extensive calibration to accurately reflect the complex realities of the changing manufacturing environments. This process can be labor intensive and requires iterative adjustments that are more complex and require different skills than for other static or traditional technologies. You can't have your maintenance teams maintain an AI model. You have to have data scientists maintain your AI models.
Filip Popov (19:01)
Indeed, I think you hit the nail on the head right there. Now, you mentioned just a moment ago that one of the driving factors or one of the biggest hurdles in deploying or having a successful deployment of AI projects in manufacturing is the top management level and leadership. So, how do you convince stakeholders and leadership teams that investing in Industry 4.0 will bring them positive AI? And what has worked for you in your experience?
Jeff Winter (19:36)
So, that's a good question. It's kind of changed over time, and I'll kind of answer it in two different ways. So first, especially over the past few years, post-pandemic, I now fly all over the world speaking at public conferences and privately for leadership teams at large companies. I've done a few of them in just the past couple of months. And the most effective way I've found to, if you want to call it convince, is often to show these leaders what happens if they don't do anything. When I do my standard Industry 4.0 presentation to companies, I typically take that and tailor it specifically to that company's industry. And I use public case studies from their peers as examples. That gets executives' attention, seeing what their competitors are doing; what they most want to know is benchmarking. How do I compare against my competitors? So, in my opinion, even though we're still early in our industry 4.0 journey as a society, there are enough companies that have been doing great things that put a lot of market pressure on everyone else.
The World Economic Forum Lighthouse program is a great example. They have 153 well-documented public case studies. And if you start to look at those or you start to share them with companies, it'll really start to make the leadership wonder if they're going to be around in five years if they don't start doing things differently. So, I would describe it, it lights a fire under their butt, and it's by showing them what others are doing out there because they need to have an answer to respond when they directly see their competitors doing something massively different, not just a small incremental improvement.
The second thing I would say is that I like to focus on future-solving instead of problem-solving. And that's a concept I got from a book called Autonomous Transformation by Brian Evergreen. So, most companies, most manufacturers, and even most vendors that are selling solutions always talk about the problems that they're solving. Few talk about future solving. So, problem-solving typically focuses on addressing current challenges and inefficiencies, leading to incremental improvements within existing frameworks. There's nothing wrong with that, but it's very different than future solving because, in contrast, future solving involves envisioning entirely new possibilities and pathways, aiming to create transformative changes rather than just optimize or modernize or improve the status quo. This forward-thinking approach not only solves existing issues but also redefines what's possible, paving the way for radical innovation and progress. So, a good example that I like to bring up when I talk to manufacturers, I talk about the amount of data that is generated by the manufacturing industry, and how it's more than other industries, and how that is a huge source of value. And most companies aren't taking advantage of the value of their data created. Something as simple as, are you selling that data or using that data as a competitive advantage against your competitors? And that usually has a lot of light bulbs go off for people to go; we didn't think about that. We're just thinking about doing what we do better rather than doing things differently. And so that's a different way to shift the thinking as well. So, those would be the two ways I would answer that question.
Filip Popov (23:09)
Thank you very much. I'm going to probably use at least the first one in my addressing the ROI question for the future for sure. And I think it perfectly echoes what you said or rather what the investigation and research shows is that it's the market forces that are driving mostly the industry 4.0 or shift to Industry 4.0. And it's an arms race to an extent. Thank you, Jeff, for taking the time to come and have coffee with me. Really appreciate it. Thanks for sharing your experience, your knowledge, and insights. I hope the audience finds it useful. Congratulations again on reaching 100k followers, and I hope to see you in the future. I hope I can have another coffee with you.
Jeff Winter (24:00)
Well, thanks for having me here.
Filip Popov (24:01)
Absolutely. Bye.