Episode 88 – Robotics Panel Discussion at 2025 A3 Business Forum

Jimmy Carroll:
Hi, everybody. My name is Jimmy Carroll. I’m the vice president of operations at Tech B2B Marketing, and I’m here for a special edition of the “Manufacturing Matters” podcast. I have the pleasure of being joined by my colleague and friend Winn Hardin, and I have a great group here to discuss all things robotics today. I’d like to introduce SK Gupta from GrayMatter Robotics, Josh Cloer from Mujin, Suzy Teele from the ARM Institute, and Juan Aparicio from Reshape Automation. Thank you all so much for joining. Really appreciate it, especially here on Sunday, taking the time out of your day. It’s our pleasure to have you. So, I’ll jump in right away and ask a question that you’ve all been asked before and undoubtedly answered many times. Is the world finally starting to understand that robots don’t steal jobs? That they actually create them and they create more profitable companies and create new jobs, new interesting jobs? Suzy, if you’d like to start us off?

Suzy Teele:
I would say not as much as we would like. There still is a stigma. We were talking about this earlier today, what happened in Los Angeles with the dock workers, right? So the stigma still surrounds it. And what we had a good discussion about earlier today was about maybe taking it away from jobs and more focusing on increased productivity. So not making that a competition between a robot and a job. But instead, how can robots help people be more effective? How people can be safer, how they can have potentially more enjoyable jobs. If we can turn the conversation more in that direction, I think that will help a lot, because—— I’ve been doing this for eight years. That fear still exists, right?

Josh Cloer:
Yeah, I would completely agree with that. I think to the layperson out there that doesn’t think about robotics every day and see robotics every day or even think inside of the manufacturing space, they still have this idea that robots are taking jobs. It’s not just robots, but, you know, RPA-type automation and computers all over the place, right? Automation is taking jobs. And I think what they don’t understand, and what they need to learn more about, is the labor crisis that we are really kind of on a path towards. And I think Japan is a great example for us to look at and to see that, hey, if we are on that trend, we need to go ahead and take steps today to make sure that we don’t get to a place where we aren’t able to actually get things to people, you know, when they’re buying stuff online and that kind of thing, or even just the everyday interactions that you might have if we end up with a labor crisis similar to that. So, I think that’s more of the message that we need to get out there into the world.

Suzy Teele:
Yes.

Winn Hardin:
Well, you know, people are generally really, really good at anticipating problems coming down the road and preparing themselves in advance. That was all sarcasm!

Winn Hardin:
I wish that was the case. So can you guys talk a little bit about artificial intelligence? Always the big topic, right? And we’ve typically applied it more to machine vision, which is one of the technologies that we’ll deal with quite, quite often. But when it comes to robotics… Talk to us about AI and robotics. Juan, do you want to kick us off?

Juan Aparicio:
Yes. Definitely a hot topic and a lot of signal-to-noise similar to robots. What can AI do and what can AI not do? It’s very interesting how, if we think about AI a few years ago, we would all think that, okay, actually it will replace—— it will enhance robots and replace the manual task, the blue collar jobs, right? But in reality, what we are seeing is actually something different. How the technology is really changing the game is more replacing the entry jobs of the white collar jobs, right? The junior analyst kind of jobs, like the back office. And workflows can now be automated, which is very interesting, while things like an electrician that comes to my home are not going to be automated, probably in my lifetime. And things that require dexterity, that require autonomy, that require different ways of thinking? Humans will still outperform [robotics]—— and also on everything that is manual. It’s very counterintuitive. But to your point about what AI can do, it has really revolutionized machine vision and started to make a dent in dexterity tools for tasks where the output, even if it goes wrong, doesn’t really matter. Right? Like bean picking is a perfect application for AI because even if you don’t——.

Winn Hardin:
It seems like it, because you need that adjustable arm robot path correction, right? That seems to be a critical part of it, which has always been the challenge with bean picking.

Juan Aparicio:
But you can recover. Now the problem starts to be, what if you make a bad decision, like the algorithm because it’s stochastic? What will happen if the decision is not the right one? Then you don’t have that determinism that you used to have with automation, right? And so these are questions that we all as practitioners in this industry need to start asking ourselves.

SK Gupta:
Yeah, we are seeing many, many new ways in which AI is being used for robotics. So, one field that used to be there, learning from demonstration or imitation learning, where the idea was that a human will do a demonstration and a robot can learn from it.

Winn Hardin:
Right.

SK Gupta:
So the previous generation of AI technology was showing significant limits to how generalization, or how generalized the solution could be from a limited number of demos. But, you know, recent things that we are beginning to see with more generative AI type of techniques, such as diffusion models, generalization is dramatically better. So you can actually show very few demos and you can actually then learn significantly more generalizable solutions. We did talk about adaptability. So it allows the robot, when it’s doing processing or motion, to react in real time to whatever is going on. You have also seen now generative AI being used to synthesize test cases so that you can test the robot to better ensure that you know how reliable the solution is. So we are seeing that, beyond machine vision… I mean, again, there are many more modalities you have for sensing. You have tactile sensing. You have acoustic signals coming in. So all of those basically are benefiting from your cross-attention transformer technology, which is the backbone of your LLM (large language model). So, all of that is tremendously beginning to benefit the perception pipelines. And LLMs themselves are now being used to encode some common-sense knowledge for the robot, right? So some of the very hard-coded task planners? You can start replacing them with LLM-type of engines. So in the field of robotics, we are beginning to see many of these new generation AI technologies beginning to play a really important role.

Winn Hardin:
When we talk about LLMs, how close are we to voice programming? I’ve started to hear people make mentions of it. I think Neuros was talking a little bit about it…

Josh Cloer:
Yeah, I mean, I can start. From what I’ve seen in the kind of conversations with LLMs myself, it’s getting much, much better. I think there’s been tremendous gains over the past few months even, with some of the things that have come out in terms of voice recognition and being able to talk naturally to it and comprehension of what you’re saying, regardless of your dialect or accent or language that you’re speaking. I think it’s gotten a lot better.

Winn Hardin:
Siri still hates me.

Josh Cloer:
Well, I think that the challenge there is more, there’s still a big leap to take inside of the robotics area. In general, we’re still reliant on very legacy control systems, and I think that’s something that needs to be changed over the course of time. Before you can get there, you have to have an understanding of what is that you’re trying to do, what are the surroundings that might impact your goals? For you to be able to talk to something and tell it to achieve a task, it has to understand what’s going on around it, right? So I think there’s still a lot to do in that regard in terms of building intelligence into our systems. And overall, that’s what’s going to provide more accessibility to robotics in the end. And I think you’ve you’ve actually got a great panel here to talk about building more access for robotics. That’s, I think, kind of the core mission of everyone on this panel today. So I think that’s really where we should focus, from my perspective is: How do we allow more people to take robotics and do something valuable with it inside of their various manufacturing workflows?

Winn Hardin:
Right.

Juan Aparicio:
And I think the interesting thing, also, about programming a robot is that the trend of making it easier to program with no code tools has been going on for a while, and you still need to know what you want to do with that. It’s never the case of the simple examples of “okay, robot, please pick up the blue ball and put it in a bin.” Right? There are more complicated things going on in manufacturing, and it highlights more that you need to know more about the process —— and then probably relax the individual “how do you program this robot specifically?” That will be less important. But still having the notion of what you are trying to do and how variable is your process. What are the limitations of this automation system, the mechanical limitations? That’s still going to be the hardest part, while the programming will become more and more simple.

Suzy Teele:
And I would say, we talk about people’s fear of robotics stealing jobs? The fear of AI is tenfold, right? Because they can’t even define it or understand it. With the robot, you can see it. So you can get a sense of physicality, of what it is. But you can’t see AI, right? And so, what we’re finding is there’re a lot of organizations that are… Just don’t even know how to start to comprehend how they could use the benefits of AI in their environment. They don’t understand very basic things, like there’s software and there’s data, right? And they have to find a way to work together. And you have to be good at collecting data or finding the data in order to be able to make this stuff work. And so many small manufacturers are not good at those types of things. They don’t have sophisticated ERP systems or other systems, and they’re not doing the analytics that you need in order to be able to train the models to do the things that you want them to do. And so it’s even harder for these manufacturers who are trying to figure out how to bring robotics in to now try to lay AI-enabled robotics on top of that.

Winn Hardin:
It’s funny because it seems like the speed of change is —— and I know we say this a lot in technology sectors —— but that it’s literally exponentially growing. I mean, even with the AI adoption… We’ve been talking about deep learning, AI, neural networks for 20-plus years, right? And very little happened. And then suddenly —— about, what, 7 or 8 years ago? —— there was just an explosion. I’m thinking primarily in industrial. There’ve been smart agent tools working for data mining in a lot of different applications. But, it just blows my mind how fast things are going. And it seems to be a through-message, on almost everything we’ve talked about so far, that the base models are such a key point. I mean, as when we talk about doing demonstration and programming by demonstration, more complex assembly text, maybe using more complex end effectors or whatever the case is. The ability to show fewer demos, SK, that’s directly related to the quality of the foundational model and that contextual component you’re talking about, right? How much the robot can understand its environment and all the objects in it and how they interact. So it just seems like we won’t have to wait long before AI will actually evolve. And I don’t mean into general purpose, you know, artificial intelligence, but just… It seems like in five years, I can only imagine how much power industrial consumers are going to be using, what AI is going to be able to deliver for them.

Suzy Teele:
I would say yes if the data is there, and if we could either create the models—— we can create the data ourselves —— or whether we can extract the data. In the work that we’ve done at the ARM Institute, that’s one of the biggest challenges that we see. The software is there. It’s great. It does what we need it to do. But how do we get data on the scale where we can actually have an impact on multiple manufacturers? That’s one of the challenges that actually SK and Juan are helping us solve at the ARM Institute —— trying to determine how to connect all the dots, right? So that there are practical applications for AI that make a lot of sense in a manufacturing environment.

SK Gupta:
Yeah. But also just I want to put a little cautious note right on that, because, see, a lot of generative AI, the one we are seeing right now, which is doing quite well and perhaps reaching 99.5% accuracy. But for most of the manufacturing tasks, you are not going to be happy with 99.5% accuracy. You’re not likely to be happy with 99.99% accuracy. Remember, we are seeing signs of saturation in large language models with basically billions of data points, right?

Winn Hardin:
Right.

SK Gupta:
And they are saturating at 99.5% accuracy. Now it’s unlikely, first of all, that you’re going to have billions of data pieces in manufacturing. So if you purely extrapolate that, what we have seen as of today, is that five years down the road… For some application, sure. For some application, given the right context, given the right bounding boxes around the problem, AI will do spectacularly well, but there will be a whole set of problems where, at least with the current AI paradigm, where you throw more data and you squeeze out more efficiency– I mean, basically more reliability or reduce the error rate, the model is simply not going to skate in manufacturing. You’re just not going to skate. There is no way you’re going to get to 99.999% in my lifetime with the way we just acquire the data and basically train the model. It’s not going to happen. So, good news is that we will see a whole bunch of really great things in the next five years, but there will be some tough problems, which, at least the current generation of AI and anything that I know of, is not likely to solve in five years either.

Winn Hardin:
That’s good. We all have job security, right?

Juan Aparicio:
I’m much more optimistic about AI co-pilots than pilots really enhancing the capabilities of humans, in this case programmers. We are even using it ourselves. We have our own AI software engineers, that are agents that write code. They don’t know about software engineering. They know about coding. So they will create a very good code. This is the first amazing UI. But then on the back end, the best practice is you still have to do it. But it has accelerated our software development by 10x at least, right? It’s this copilot, this enhancement of human capabilities that we are experiencing with this technology.

Suzy Teele:
And I think it builds on what you both said. If we look back at the history of robots, the industrial robots took off because they could do the same task repeatedly, right? But cobots need to be more… People want them to be more customizable, more personalizable, more specific to their own environment. And we see that’s a harder problem to solve. We’re running into… We’ll go through the same evolution with AI, where there’ll be certain AI models that fit that industrial mode, right, because they’re repeatable process. But so many of U.S. manufacturers are custom manufacturer shops. They get an order and they have to change their process to build a product for some certain person that’s different from the thing that they’re doing the next day. And so how do you get good, 99.99999% in an environment like that? Is is a hard problem to solve. It’s worth trying to solve, but it’s a hard problem to solve.

Jimmy Carroll:
Just talking about LLMs in general, it’s interesting—— there are some big companies that are using —— like, Juan, what you were saying —— using it to do certain coding tasks or whatever. But I’ve repeated this story and I maybe even said it to you when we did a separate podcast together a while ago, but, there’s a gentleman I did a podcast with last summer, and he was the director of AI at a systems integration company. And he said, we’re finding we’re having great success with LLMs, but there’s still hallucinations that are causing issues. And to your point, Josh, ChatGPT has made some improvements, but he said, “tell me the names of Snow White’s 13 dwarfs.” So [ChatGPT] gave the seven names and then made up the rest that sounded similar. So there’s still the little problems that need to be worked out for it to add full value.

Suzy Teele:
That could be a new Disney movie. You know, we are in Orlando!

Juan Aparicio:
The difficulty of hallucination is that it’s not a bug, it’s a feature. Right? It’s something that, while the names are good because it can hallucinate and then that is a great feature, but when it comes to the physical world, then we can not allow this kind of hallucination because then things can break. I really like —— and Rodney Brooks will be one of the speakers here —— but what he mentions about that is that two things have to be true for any AI application out there. And I think it’s still true: Either there is a human in the loop or the risk of failure is low. Right? But if the risk of failure is high and there is no human in the loop, there are not many AI systems out there. Even in Waymo, when something happens, there is a person in the back that can take over, right?

Winn Hardin:
Yeah. And we’ve got all the Tesla YouTube videos to prove it.

Winn Hardin:
So we’ve talked a little bit about AI in terms of helping spread adoption of robotics, but are there, beyond the standard ease of use —— and you talked about optimization and customization of cobots specifically —— are there other things that robotic companies are doing right now that are helping to spread the adoption of this robotic technology, whether it’s AMR or whether it’s traditional industrial?

SK Gupta:
So basically, one of the ways in which robotics always seems to help is that it can improve your process dramatically over human operations. So in terms of speed, for example… If you’re doing sanding, the robot can sand four times faster than a human can. So obviously you get four times the throughput than if a human is sanding it. So this is a huge benefit to the robot doing it. And the reason for it is the robot can move much faster and the robot can apply significantly higher force. And you can spin the tool much faster in the robot’s hand than you can in a human’s hand, right? So once you take away human’s constraint, then robots can, in many applications, do it much, much faster. The second thing is quality. You can deliver quality, which is going to be, again, significantly better than what humans can deliver. So adoption of AI-powered robots offers those two benefits. If you are successful, then you get dramatically better throughput and you get dramatically better quality. And those then become business drivers for adopting it. Now the challenge has been in high mix. You need the right kind of AI to power the robotics so that you can have those gains. Without that, if you then come back and demand that a human programs a robot, then obviously this is not going to work.

Juan Aparicio:
The other trend, to your point: Algorithms are getting better for some applications. And things that were impossible, like doing something in a high mix, low volume, now it’s possible, thanks to companies like GrayMatter. Same thing with some applications in machine vision that used to be extremely hard with rule-based, now with AI they are possible. And at the end of the day, it’s also another trend —— that is less sexy, but that matters a lot because it’s all about ROI —— is that costs are going down. And if there is more competition, more systems, more standardization… Right now, if you want a palletizer, there are many options that you can choose. That’s driving the cost down. And for a lot of customers this is the moment they say, “okay. Now this makes economic sense.” Because even if the applications were amazing, it’s all about the ROI and the way that ROI is calculated. Going full circle, it’s not considering all the benefits that SK mentioned, it’s typically “how many people do I replace?” They don’t account for all the other benefits, that probably save 2x what the salary is.

SK Gupta:
So, building up on what Juan said and connect it to give you an example of robotics and machine vision and sustainability—— So, basically, what happens now when humans do sanding…. As the sandpaper starts wearing out, to compensate for it you have to apply more force. As a human, nobody wants to do that. So you’re better off just tossing the sandpaper after a little bit of use. Now, if you can get robots to do it, and then you can have machine vision that can monitor—— A machine vision problem is super hard because you have these sandpaper-wear detection patterns as you’re doing it. So you can do that with modern machine vision techniques. So if you can monitor it, you can figure out exactly how much life is left and how much the sandpaper has worn out. And the robot can compensate by applying more pressure and adjusting the velocity. So if you have a adaptive robot AI-powered machine vision, it turns out that you can actually reduce your sandpaper consumption on a sanding line by 40%. So that has now huge sustainability benefit as well, because you’re sending less sandpaper to landfills. And also if you are using expensive sandpaper, that is also a huge savings. Which, when people are doing traditional ROI calculation and seeing how many humans they are replacing, that wasn’t on anybody’s radar screen. When people start deploying it and say, “why are we not ordering as much sandpaper as we were ordering before?”—— That’s when they’re discovering, “oh, robots don’t need as much.” And then the question is, why not? Because the robot is really optimizing it.

Winn Hardin:
And since you’re tracking everything with force application and speed rotation and everything else, you can have a predictive model.

Suzy Teele:
Another angle to look at is just the amount of involvement now that you see from robotics companies or suppliers in the automation industry with robotics competitions. Another way to solve this problem is to make people feel more comfortable working with robots at an earlier age. And so what I’ve been encouraged to see is the more active participation by the industry as a whole in FIRST Robotics and VEX Robotics competitions and, you know, people volunteering and working with these students—— there’re thousands of teams across the United States right now. And so another way to solve the problem, beyond what’s really important —— what SK and Juan talked about, which is how well these work —— is letting people know that they can use them. They’re not intimidating. They’re not scary. If a bunch of 14-year-olds can get together and build a robot to do something, then maybe I can do that kind of thing, too, in my job.

Josh Cloer:
I think that’s another trend that you’ll see with AI as well, in terms of helping the operation use the very expensive —— in their minds —— robotic system that they just got. It’s usually non-skilled labor that has to go and figure out, okay, something went wrong. What do I do? I’m not sure I’ve seen this yet, but I think we will see this, in terms of trends in the software world. You see a lot of these co-pilots where you can kind of directly ask things inside of the software, like a CRM, and you can say, “hey, I’ve got a new deal, here it is.” And then they’ll just put it into the system for you. Well, you’ll see this in robotics as well. In the future, when we have more software-enabled robotics, you can go in and say, “oh, there’s an error.” Well, what does that error mean? And you can have this kind of conversation with the system. I think that’s going to be coming up as a trend for AI and robotics as well, to help the user experience within the robotic systems.

Winn Hardin:
Yeah. In words they understand.

Suzy Teele:
Exactly.

Winn Hardin:
Yeah. That’s fantastic.

SK Gupta:
We already have in our system PHM (advanced prognostics and health management) integrated into it. Every time, anytime robots enter the fault mode, it will give you a hypothesis as to why it went into that fault mode.

Jimmy Carroll:
Over the last several years —— and there’s some good examples in the companies that you guys represent here —— but over the last few years there’s been sort of a rise in the availability of these application-specific, purpose-built systems that are used for palletizing or, you know, sanding, grinding, whatever. What what what’s led to this, do you think, and what impact does it have on users. And and then part two of that is like could you see and Susie you mentioned robots earlier. And that’s a good you know, I’m glad that you did because part of this I’m thinking is like a lot of these systems are leveraging Cobots. Will that help the smaller and medium sized enterprise adopt automation faster.

Suzy Teele:
I’d say there’s one factor, and that’s lack of skilled labor. That’s pretty much everything that we see is pretty much driving an interest in bringing robotics in. So we have a couple of factors that are again, you mentioned Japan. So what’s happening here? Our manufacturing workforce is aging. Its average age is 55 and above right. Younger people are having less children. Yeah. And that’s actually a worldwide trend that they don’t they think it’s permanent. It’s not temporary. They see that in Japan already. Right. There are a lot of single women. There are women that are not choosing not to have children that are focusing on their careers. And so what we have is a shrinking labor pool in the United States. And when we look at and talk to the manufacturers in our environments, the number one thing we hear is, gee, I would like to bring a robot in if I can find somebody to actually manage it after the systems integrator or the supplier has left. And so I would say that that’s probably the biggest limitation that we have in the United States, which is why it’s important that we get younger people exposed to advanced manufacturing technologies earlier in their in their learning cycle. What we found is that you have to, about fifth grade, have to let kids know that they can be a doctor or a lawyer or a teacher or a manufacturer. Right. And we don’t do a very good job of that in the United States.

Speaker5:
Absolutely.

Juan Aparicio:
I think these are the two sides of the coin, right. The teaching them or exposing them to manufacturing to produce things to different processes and then automation. Right. Yeah. I will say that one of the reasons why this specialization exists is that you need to be very specialized in the process. The process becomes more even more important knowing about the process and knowing about the robots. Right? The same robot, the same white yellow, orange robot can be doing the ball or can be doing something with some differences, but it’s the knowledge of the industry that you can speak the same language. What differentiates one company versus other? If you want to do welding, you have to do it very, very well. If you want to do something, there are a few applications that are around that, but that requires a specialization and also that specialization, the go to market and the technology. Same thing for in your case for the end of line, right?

Speaker5:
100%.

Josh Cloer:
I think in terms of the the purpose built robotics, I think that’s where I see there’s a lot of talk about AI. I think it’s really more about the applicable intelligence and different buckets and different areas of the robotics environment that that’s creating these purpose driven things. So it’s been for for material handling. It’s been machine vision that’s primarily driven the capability of doing things like mixed palletizing, right. Picking anything that can just show up on a pallet where you have to really understand what’s going on with the combination of different sensor technologies and a combination of different algorithms. It’s not just machine learning that’s making that possible. It’s a lot of different things coming together. So I think that’s, you know, the hype is there with AI, but it’s more than just the the artificial intelligence that brings these things to life.

SK Gupta:
So yeah.

Juan Aparicio:
I was going to say just to be a little bit polemic here because otherwise we’ll be boring. The there is no such a thing as a general purpose robot, right? Like when a humanoid portrays itself as a general purpose, that doesn’t make sense. Like even humans, we are not general purpose, right? We do a few things very well and we are going to see that, right? That in most cases, the application, the best application or the best form factor for solving a specific application will not be a humanoid. It will be something else that is specifically built to solve that application. That’s how you get these 4 or 5 x ten x speed over a human.

SK Gupta:
Yeah. I mean, so decisions, you know, to to embrace new technology get made based upon usability and ROI, right? Am I going to get my ROI, which I’m hoping for, and how easy this is going to be to use this technology right now? You know, special purpose built robots often deliver very good ROI, right? Because they can run much faster, they can give you significantly better quality, and therefore you can make an ROI case and obviously some time the penalties on usability. Right. So the thing is that okay then you have to figure it out. What training do I use. How do I get the right kind of workforce. Right. But if ROI argument is there, people are likely to figure it out. The usability piece if the ROI argument is really compelling, right? On the other hand, if you approach the problem from the other way where robot is highly, highly usable, right? A child can use it, but your ROI is not favorable. It’s a toy at that point, right? Nobody’s going to deploy it, right? Right. So one has to think about, you know, from that perspective that, you know, if a, you know, very specialized robot can deliver you spectacular performance, right. Then chances are that you will figure it out how to utilize it, because business benefit is going to be so great. Therefore, you will figure it out, right.

Suzy Teele:
And you know, people tend to think in factors of I’ve got a problem to fix. They don’t say I want to automate my whole manufacturing plant. They say, I’m not getting this product out the door fast enough, or I’m having problems finding people to be able to work in the a certain area of, you know, because there’s chemicals or it’s, you know, it’s sanding. And that’s a very, very difficult job to do. So we’re oriented to think towards what can I get that? We’ll solve this one problem for me right now that meets the ROI factors and the usability factors. And I always say like when your heart when you think you’re having a heart attack, you don’t want a GP, you want a cardiologist, right? And so it’s kind of the same concept, like you want the specialist that knows how to solve that particular problem that you have in your environment, which is why you have special purpose systems. Ultimately, you want them all to talk to each other, right? But for now, you want to solve that one problem that’s holding up your production process.

Speaker5:
Gotcha.

Winn Hardin:
So you mentioned humanoid a little bit ago. So I’m really curious to get this panel’s input. And I was at the Humanoid Robotics conference earlier this year. It was really cool. And we don’t have a standard in place yet for it, right, in terms of safety standards. But where do you guys come on that and maybe compare and contrast that to to Amr with Arman movable pedestal mobile systems like that. Can you talk a little bit about both of those?

Josh Cloer:
Yeah, I think Juan hit the nail on the head for me. It’s it’s there’s no such thing as a general purpose robot. And that’s really what we’re trying to accomplish with with the humanoid is, I think one of the challenges that I see on my side in the distribution side of manufacturing is you have a lot of companies that have built these massive distribution centers all all around a human, and we think that we’re going to make a robot that is, you know, looks like and does things like a human, and then it just goes into those types of environments instead of thinking, why don’t we change the way the infrastructure looks to match what a robot is capable of, or even a new form factor that’s even more powerful and capable than the current robots that are out there today. So I think that humanoids, uh, over time, will evolve to do things that are really, really spectacular, that I think there’s other industries that they will play a big role in. I think they will find their place in manufacturing and distribution in specific areas. I’m not so sure it’s inside of the, the, the places where robots play today. Maybe there’s some assembly things that can happen, but overall I see that it’s still going to be something that is purpose built for a specific application.

Winn Hardin:
Which would seem to feed back into what Scott was saying about increasing the ROI. Right? Because if you if you keep things the way they are, and you simply just put a new automation tool right in the center of it, you’re not really going to be able to go faster, right? I mean, not not realistically speaking, you’re not massively.

Speaker5:
And I.

Winn Hardin:
Think whereas if you redo it with conveyors and everything else in purpose built robotics, you’re going to be able to go.

Speaker5:
Up and.

Josh Cloer:
And Americans have a specific way of looking at ROI as well. I know you guys all see this, but we have you know, you need a one year or two year, maybe even a three year payback. But when you go to other regions, when you go to the markets, they’re looking at ten year paybacks and they’re just ripping it all out and rebuilding it again, because they see it as the infrastructure that propels progress forward. Right. So when these systems can last decades and we’re expecting a three year payback, you’re not going to advance the adoption of this automation as fast as you could be, realistically, with the with the financing and the economics for for robotics. So we need to kind of change our mind a little bit about how we look at the financing of these things as well.

Speaker5:
Yeah.

Suzy Teele:
It also humanoids open up the possibility for other types of human beings to be involved in the manufacturing process. Uh, you know, so much of manufacturing relies on strength, right? But if you can have a humanoid that is being managed by a disabled person who maybe doesn’t have the strength but has the knowledge in order to be able to control that, then that that opens up more career opportunities. So that’s how I kind of think about kind of like just like Amrs and everything else. It does open up the opportunity for people who may not, you know, in their fathers or grandfather’s generation, thought about a manufacturing job because of the physicality of the job. Right?

Speaker5:
Yeah.

Winn Hardin:
Yeah. I mean, we see this and there’s been a lot of advances in exo systems and a lot of, you know. Exactly. And so I’m starting to hear about, um, that the Amr with the armed solution is is becoming especially at the end of the line, is becoming something that people are really, really excited for. Um, most large installations have to use a large number of different types of robots for different application spaces. And interoperability has been a challenge. I think a barrier getting these systems to talk to one another. Um, are you you want to you want to talk a little bit about that?

Josh Cloer:
Yeah.

Winn Hardin:
I think this is going to be the year of the of the pedestal.

Josh Cloer:
I think there’s similar to the robot. There’s there’s spots where you can create kind of spot solutions and have something like that drive around and do work all day. Maybe, maybe, maybe you only receive stuff into the warehouse, you know, for two hours in the morning, and then you need to go over here and you need to take these boxes and put it onto another pallet and put a label on it. You might be able to do some, some things like that, but you’re exactly right. It’s it’s the scale of this that doesn’t really go up. Right. You need this massive fleet orchestration thing all of a sudden with those types of solutions if you want to scale that up. So I think it works in some smaller areas. It’ll work in small scale operations. But to scale that up would be would be really difficult. And I think ultimately you would actually change a lot about your infrastructure. And once you start to do that, well, was there a better option for you?

Speaker5:
So might as.

Winn Hardin:
Well just done the greenfield development again.

Speaker5:
Right.

Winn Hardin:
Which brings us to the next point, which is great. And Susie, you were talking a little bit ago about China. Um, now, which makes me think of the rest of Asia, which makes me think about. I’m sorry you said Japan. I was thinking about China.

Speaker5:
Okay. And obviously.

Suzy Teele:
That’s a whole other area.

Speaker5:
Yeah.

Winn Hardin:
So, I mean, we’ve been hearing for years about China being the leader on robotic adoption, and but I’m curious, is that is that going to stay the case? Are we starting to see any momentum changes here in the US or in Europe? Is there maybe trying to catch up and and if it does stay that way, what does that mean for global the global manufacturing market. What is what does it mean for global manufacturing. China continues to run away. Or is this like a third rail question where.

Speaker5:
We yeah.

Josh Cloer:
I think I think they will continue to accelerate. I mean a lot of that is being subsidized by the government in China. I think a lot of lot of companies are starting there that are basically operating just off getting paid by the government to progress robotics forward. Right. Without even really paying, getting paid by their customers at the end of the day for, for for what’s realistic. So I think there’s a lot of acceleration that will still happen there. But there’s other markets as well. I mean, you see, South Korea is a massive adopter of robotics as well. Um, I think both the US and Europe will propel forward, will advance, but at the same pace. Uh, we don’t we don’t see that right now.

Juan Aparicio:
It’s very hard. Right. It’s 12 x. What we do in and what they do in a month. We do it in a year, right? Yeah. And it’s part of what you said is, is subsidize heavily. And it’s also not only robotics. Right. It’s not in isolation. It’s also the entire automation, the ecosystem. The fact that they decided they purposely decided we are going to be the makers of things in the world. So we are going to train the people. We’re going to have the infrastructure. It’s the mentality, right? I can send ask for a quote of a vendor in China. I get an answer in probably less than eight hours, sometimes overnight. I say, how are you awake? Like, why? How can you give me a quote and get it even faster than if I try to do it with an American counterpart? That may take two weeks.

SK Gupta:
Yeah. So, I mean, just to dissect the problem into two different buckets to just see why China seems to be doing so well. Right. So, so so of course, in order for a manufacturer to deploy a robot, there are two parties, right? Robot makers and manufacturers. Right? Right. So, I mean, if you look at it, us. Where are our robot makers? There are hardly any robot makers in the US, right? And it’s not likely to change in any time soon. Right now, to all of our surprise, China actually produces more than 50% of the robot, which they use themselves. Right? So China is not only becoming world’s biggest robot user, right? China is also becoming world’s biggest robot manufacturer. Right. So once you take a look at government helping both side of it. Right. So they are just, you know, making sure that they are robot makers. Other automation providers are also helping in building that ecosystem. And then their policies which are then enabling manufacturers to use those robots. Right. So that’s the government side of it. But Chinese made robot cost. When you look at it from the perspective of a Chinese manufacturer who is trying to deploy, it is very low.

SK Gupta:
Yeah. I mean, so we can compete on those terms, right? So same robot, which will cost you $50,000 in us if you were to get one of these Japanese or European robot, Chinese manufacturer will have access to a robot with comparable capability made in China at a fraction of that cost, right. So if you take a look at it, though, they have attacked the problem on both fronts. Not just that okay, we are going to encourage more people to use robot. They also said we will build the entire ecosystem for making really cheap robots, right? So they attack both problems. So I think I don’t think anytime we are going to see anybody in the world, which is going to be able to compete with the cost of the Chinese made robots, right. Even though, let’s say tomorrow, government may start subsidizing everybody that they should buy more robot, but the cost of the robot is not going to go down in us anytime soon. Right?

Speaker5:
Right.

SK Gupta:
So therefore, I mean, in the near term, we don’t see I mean, yeah, that the growth in China is slowing down.

Speaker5:
Right, right.

Winn Hardin:
But Nora’s picking up, it sounds.

Speaker5:
Like or catching up.

Suzy Teele:
I wish we all had better news on that. Right. I think, and a lot of it’s been addressed by the panel. It’s it’s a challenge, I think, in the United States, with the way that we’re structured to be singularly focused on a small number of things. Right. Um, you know, certainly during during the pandemic, Endemic. We all said, whoa, we’re not making enough things here in the United States. And that became very visible. But people don’t even think about that anymore because it’s not a crisis. And so we tend to operate more in crisis mode, I would say as a country than in long term strategic mode. Um, I mean, if I look at, you know, we’re part of a set of organizations called manufacturing USA that was created ten years ago. We have our ten year anniversary of all the institutes, um, to try and have some impact on this problem. And we’re trying and we’ve got great members that are doing great work. But you put us up with like the Fraunhofer organizations in, you know, the work that they’re doing in Germany. And we’re like this much of the type of thing that they’re doing there. And so we tend to I always say we tend to throw a little bit of money at a lot of problems in the United States because we have such a diverse population and we have such a culture of invention, which is Which is amazing and wonderful. But then we don’t focus after that. So we’re our own worst enemy. Sometimes we have the ability we invent. I always say we invent all the robotic technology, but we have no major robotic companies in the United States. Like we have a lot of great ones that are growing, but we don’t, you know, Fanuc isn’t here. Yaskawa is not here. Cook is not here. Right? Abb is not here. Universal is not here. Right? All the major robotics suppliers in industrial automation are not US headquarters. I mean, they’re not they weren’t started here. They’re all from other countries.

Speaker5:
So that’s.

Winn Hardin:
Interesting. You mentioned Germany because that was the first that was coming to mind. Whether, I mean, all the Asian nations tend to have a longer view. Point back to what Juan was talking about. Right. And Josh, you were talking you know, we need to have a longer vision when it comes to payback and and what our solution is. And then I was thinking, well, you know, then there’s cultural differences in some cases where certain or maybe in socialist countries where the government works more in hand in hand with commercial interests, and does a lot of the Fraunhofer kind of support and development. So I was wondering if maybe those as a matter of fact, I really don’t know what Germany’s uptake is when compared per capita to China in terms of using robotics. You probably don’t want.

Suzy Teele:
Yeah, they’re better than us. Better than us.

Juan Aparicio:
So but it’s also.

Speaker5:
Know.

Juan Aparicio:
That it’s also better than us. But also, uh, while robots is a proxy for automation, it doesn’t. It’s not just 1 to 1. So there’s a lot of automation that’s not necessarily robotic. Right. And so German is highly automated to the point that most of the tasks are automated. I think their strength is on the traits. Right. They are they were able to train people and say, okay, you don’t need to go to university. You can take the trades. And that’s institutionalized is rewarded. So that education and workforce development is probably top worldwide. Yes, we also have that problem in the US, right? We have the stigma too. Nobody wants to work in manufacturing, so we don’t have enough factories. Not enough. Then motivation for robotics companies to do products only for the US. So it’s a chicken and egg problem.

Speaker5:
Chicken and.

Juan Aparicio:
Egg problem. We need to industrialize. I mean, there is no there is no future. We cannot continue to grow if we don’t do it right.

Winn Hardin:
Well, at least we’ve given a board of directors a new list of agenda items.

Speaker5:
Yeah, the same ones. There’s always something to work on. Yes. Cool.

Suzy Teele:
But there are shining stars. I mean, if we are not moving backwards, we’re moving forward very slowly. That’s what I would say.

Speaker5:
I think.

Winn Hardin:
We need shining.

Speaker5:
Stars to be able to say.

Winn Hardin:
Okay, that worked.

Josh Cloer:
Well. And similar to the way we are in software development in the US. We have a lot of that trend here in North America, right? A lot of the software enabled robotics companies are starting here in North America. So there’s still a good trend there for us. Maybe not the makers or the adoption, but what we have for the future, which is building the intelligence into these systems.

Speaker5:
Yes.

SK Gupta:
Yeah. I mean, entire AI revolution is being pioneered by.

Suzy Teele:
Yes, yes. We are leading that. We are definitely leading that.

Speaker5:
Okay.

Winn Hardin:
All right. Cool. Because I really wasn’t sure between us and China right now in terms of AI, the number of GPUs that were developed or, you know, and their processing capability.

Juan Aparicio:
Definitely on the lead. Right.

Speaker5:
Beautiful.

Juan Aparicio:
With the interesting case of France that has also companies like anthropic that is also probably has two heads to OpenAI. But uh, the for different reasons probably. Right. Is the talent here. The software talent, as you were saying is is incredible, right?

Suzy Teele:
Yeah. We are excellent at technology innovation. Maybe not as excellent at technology adoption. China’s very good at technology adoption.

Speaker5:
Right, right, right.

Winn Hardin:
And so much of that is government regulatory, policy based macro level issues.

Jimmy Carroll:
Are you noticing any other notable trends in industrial automation in general for so long? Like, you know, automotive was huge is still huge for for robotics and automation. But but beyond the automotive like beyond the car floor, you know, beyond that and, and into some of these smaller or medium sized enterprises like are you noticing anything interesting or notable with with robot robots in general?

Josh Cloer:
I would say a trend in kind of the distribution side of manufacturing is that smaller to medium size enterprises, smaller warehouses are able to adopt more robotics technology these days because of the the software capability, the interoperability, the fleet systems that are out there. A lot of the QR code based platforms that that, you know, we’ve been kind of stuck because of the Kiva acquisition for a very long time and not able to deploy that Of that technology, but now that’s that’s over and we’re able to deploy that. And there’s a lot more uptick in adoption of that type of technology inside of the distribution area, along with robotics as well. Um, so I think you’ll see more small warehouses that don’t necessarily have all of the infrastructure adopt more flexible automation. Um, that’s that’s what I’ve seen at least.

Speaker5:
Yeah.

Juan Aparicio:
It’s very interesting to see and compare conferences. For example, you go to prom app or there’s more warehouse and then automate, right, and see how even the number of startups that target distribution instead of manufacturing, because manufacturing we put all under the same umbrella.

Speaker5:
But.

Juan Aparicio:
Factories look completely different. Well, warehouses have some commonalities, right? There are different. Pl may be different than your own warehouse, but still they are more adept to repeatability, therefore more adapt to automation. So I see a lot of very interesting startups in and new concepts for like also for unloading trailers in in distribution and less on manufacturing. I mean, I think scale is kind of the bright light on because that’s really innovative and targeting manufacturing, but they are not that many. And what I have seen in manufacturing is on the end of line. That was kind of not being looked at because that was more on the warehouse. Then it starts to propagate all these interesting and new innovations in distribution centers. They start to be applied to the end of line part of manufacturing. So you see more amrs. You see a lot more palletizing deeper, etc. and also machine vision, like more quality inspection is something that is also booming thanks to AI. Going back to.

Speaker5:
The.

Winn Hardin:
Totally aligns with what we were just having. We had a recent other panel on machine vision, and that was one of the biggest things they talked about. And we talked about the robotics machine vision relationship, you know. So by the way, so check out the machine vision panel when you get a chance.

Suzy Teele:
Yeah, I would just have to second all that because and in fact, people are becoming comfortable with it. Like they go to a restaurant and an Amr delivers their pizza. Right? And they’re like, that’s so like it doesn’t it doesn’t scare them as maybe it would have five years ago. And so I think that just the technology itself and its capabilities coupled with people are like, oh yeah, that’s okay. I, you know, like I expect something to be delivered to me by an Amr at some point in my life. Right. And so they’re comfortable with that. Um, and it’s easier to do, like you said, the processes are much more repeatable. So that’s where we’re seeing probably the most.

Speaker5:
Movement.

Winn Hardin:
Piece of that.

Speaker5:
Pizza.

Suzy Teele:
Yes, exactly.

Speaker5:
I mean, we.

SK Gupta:
Are beginning to see high mix manufacturers now taking, you know, beginning to take a serious look at robotics, right? So, I mean, even aerospace companies, right. All leading aerospace companies are taking a very serious look at robotics and see how they can use it. So I mean, so beyond automotive and electronics. I mean, there are marine industry, boat makers, yacht makers, aerospace, right? So, so many, many other sectors are beginning to take a serious look at robotics, what it can do for them.

Speaker5:
So it’s good to hear.

Winn Hardin:
Yeah that’s yeah. And that’s very difficult assembly.

Speaker5:
Task.

Suzy Teele:
Because we’re seeing movement. And Juan helped us with this in the robotics sewing area. So the whole apparel manufacturing recycling there’s you know a lot of large organizations now have recycling goals, right. I’m a I’m a producer of high end purses or high end shoes. I have I have certain goals for sustainability that are put on me by my board that I need to meet. Well, who wants to the job of ripping the laces out of a high end pair of tennis shoes, right. So that you can recycle those materials. So we’re definitely seeing movement in that whole apparel creation and apparel recycling as well.

Speaker5:
That’s interesting.

Winn Hardin:
That’s interesting. Um, how about um, And agtech things outside of the of the production floor. Um.

Juan Aparicio:
I mean, it’s driven by the the same problems that we see of labor crisis. You go to a farm, it’s even harder. You go to construction and it’s even harder there. So those are the industries that are they just cannot continue their business if they don’t automate. So they are forced to. And in manufacturing, there is still a I have been doing the same thing for 20 years. It hurts. But when you wake up with your and your leg hurts for 20 years, you get used to that, right? Yeah. But these other industries, they need to.

Josh Cloer:
And it’s a great place to to combine human intelligence and have a human in the loop type of approach, because there’s a lot of variables. Once you go outside and you have to deal with the sunlight and the weather and all the different things that that might happen, right? It’s a great place to apply AI along with that human in the.

Speaker5:
Loop.

SK Gupta:
And agriculture and construction. You cannot outsource it, right? So so manufacturing one of the strategies where you didn’t want to do automate and you didn’t want to make investment. People outsourced it, right? But construction, definitely, you cannot outsource it. You have to do it. Agriculture. Most people realize it. Outsourcing would be a really bad idea. Yeah.

Speaker5:
So yeah.

SK Gupta:
So then obviously people have to take a serious look at it, right?

Josh Cloer:
So then construction is definitely a trend for automation right now. I think the coolest one I’ve seen is this little, little mobile robot that goes and actually prints out your, your layouts on your floor and in the construction world. So you just have a nice blueprint for you to go and build everything on top of. So there’s a lot, a lot of cool stuff.

Speaker5:
Going on there.

Winn Hardin:
Yeah, I know there are already using robots to to pay and validate percent builds, right. And then, you know, along with some 3D scanning to be able to let people, their vendors, their, their, their contractors do pull downs from, from the whole project. So and there’s a whole lot more applications outside of the factory floor than there are inside it, right? So that’ll be once they’re able to tackle airframe assembly and complex operations and and outside environments that.

Josh Cloer:
That’ll be robots on.

Winn Hardin:
Mars. Fuel on the fire. Well, I’ve no doubt there’ll be robots. I’m sure they’ve already have been.

Juan Aparicio:
That have been done.

Winn Hardin:
That’s true. Yeah. Very cool. Well, thank you so much for your time today, panel. We can’t thank you enough for coming and joining us today. Thanks.

Suzy Teele:
You’re welcome. Thank you for having us.

Jimmy Carroll:
Thank you. Yeah. And if anybody has any questions or comments to the panel, we’d be happy to facilitate those, those messages and reach out to us at manufacturing. Com and thanks for tuning in.

Winn Hardin:
Yep. And if you want to see any past episodes of manufacturing. Com or any of your favorite podcast platforms, we’re all over the place. So until then, thanks again and thanks again for joining us today. Thank you, thank you, thank.

Suzy Teele:
You, thank.

Winn Hardin:
You.

Juan Aparicio:
Thank you.

Speaker8:
Thank you.

Sonix is the world’s most advanced automated transcription, translation, and subtitling platform. Fast, accurate, and affordable.

Automatically convert your mp3 files to text (txt file), Microsoft Word (docx file), and SubRip Subtitle (srt file) in minutes.

Sonix has many features that you’d love including automated translation, generate automated summaries powered by AI, collaboration tools, share transcripts, and easily transcribe your Zoom meetings. Try Sonix for free today.