FANUC Partners with NVIDIA to Advance Physical AI in Robotics - Mike Cicco, President & CEO of FANUC America

Announcer, this is the TechEd podcast, where we feature leaders who are shaping, innovating and disrupting technical education and the workforce. These are the stories of organizations leading the charge to change education, to rethink the workforce and to embrace emerging technology. You'll find us here every Tuesday on our mission to secure the American Dream for the next generation of STEM and workforce talent. And now here's your host, Matt Kirchner.

It's Matt Kirchner on the TechEd podcast, the number one podcast in STEM and technical education, where it seems like every single week, we have a conversation about what we call the edge to cloud continuum. We are huge believers here at the podcast, that the next iteration of AI and artificial intelligence and machine learning is all about embodied artificial intelligence, physical AI. Some people will call them applied artificial intelligence. It's how AI and machine learning manifest themselves in the physical world.

Our guest today is on board to talk about all of that. One of the quotes I love going back to regularly is one from Jensen Wong. I'll paraphrase it, but it was at the Consumer Electronics Show. Jensen, of course, the CEO of Nvidia, by market capitalization, the largest company on the planet, and Jensen said that the next iteration, the next big thing in AI, is indeed physical AI.

So we are 100% aligned on that. Speaking of Nvidia, today's guests just announced a huge partnership with Nvidia. The guest, by the way, is no stranger to the studio of the TechEd podcast. In fact, it is his fifth appearance on the podcast, and we keep inviting him back, and he keeps accepting because we have these incredible conversations about the future of robotics and automation and advanced manufacturing.

So someone who absolutely needs no introduction to our audience, but I will introduce him nonetheless, the President and CEO of FANUC America. Mike Chico, Mike, awesome to have you on the podcast again. Matt, as always. Thank you very much.

And and always a pleasure to be on there. Everything you do with this podcast is great, and being able to talk to leaders and you know kind of project, what's, what's the latest, I'm all for. So I'm happy to be here. And thanks.

And I'm building my lead. I like to stay in the lead of number one presenter on your podcast, so I'm gonna stay there. Awesome. Well, we'll keep making sure that happens.

There are some people that are that are on your heels, as you know, and they say, oh, so how many are? Is Chico at because I, you know, they're trying to catch you, but there's just always these great conversations Mike that we have in today, highlighting physical AI, highlighting Nvidia and that partnership, I should remind our audience, by the way, that I think they know FANUC America, largest robotics company in the in the world, huge, huge market share here in the United States and around the globe, also huge on the CNC side, really, the grandfathers, if you will, of the CNC control More CNC machine tools are controlled by fanuct controls and then any other control company, and it's not even close.

I won't, I won't share the percentages with our audience, but it's incredible, incredible market share, incredible company, and it's an honor have you on board. Speaking of incredible companies, Nvidia, I mean, that's pretty darn good company. Talk a little bit about this, this new partnership that you've got with Nvidia? Yeah, we're really proud of it.

We formally announced it at the IREX show in Japan late last year, and it really went off very well. A partnership with Nvidia is kind of one of those badges where I think that anyone that's working in the in the AI space, a physical or embodied AI, as you mentioned, really needs to be on board with what Nvidia is doing, and so that partnership is kind of the at the center of that. It's led by our corporate office in Japan, so it's a formal corporate partnership. And but given the geography of Nvidia and the work that's done in the United States, our team at FANUC America is playing a pretty key role in that.

So as you know, Nvidia is a huge company, the largest company by market cap in the world. But in terms of the partnership, it's really from our side, broken out into two major areas. You have, you have really the physical AI, part of it, from a hardware perspective, and that's somehow trying to utilize their GPUs in our space from a hardware perspective, and then you also have the digital side of things, and that gets into their Omniverse platform. And how do we bring assets into that platform as well?

So overall, it's really broken out, really into those two major areas. And for so many years, we've been talking about this idea of digital twins, the ability to innovate in the digital space before we bring those iterations or those innovations into the physical world in the form of robotics and automation and so on. Lots of lots of great things have been happening, really, for by my perspective, probably seven or eight years when some of these technologies really started. Started coming of age, you know, then you think about, I read the thinking machine.

Not sure if you've read that book yet, but it's outstanding. Stephen wood is the author. Open invitation to join us on the on the podcast, but that is all about the history of Nvidia, and it's all about Jensen Wong and where this organization started and how it grew. And you're exactly right.

I mean, it's really a badge of honor. And in some cases, I think, as we advance into this embodied and physical AI world, you know, almost a ticket to the dance, like we've got to have these types of partnerships with companies like Nvidia, because what they're what they're doing on the server side and on the processing side, super, super innovative. And so it's a really, really cool thing that you're involved with. One of the things that was highlighted, I think, in the press release that I saw, was this idea of Isaac sim, when we talk about digital twins, and we talk about manifesting what we can do in the physical world digitally first, and iterating and making changes and and making sure that that system is working the way we want it to before we we build it.

What should our audience know about Isaac sim and how that fits into this whole partnership? You know the term digital twin? It's one of those ones that's like widely used like AI, where it could mean so many different things, right? To achieve a true digital twin, you really need to have the simulation in the real world connected and then talking back to one another and having them improve each other.

That's a true digital twin and FANUC, we've had a digital twin software. We have a digital twin software. We've had it for a long time. It's called Robo guide, right our simulation platform.

It is a true virtual controller. It is a digital representation of our robot controller, and then it visualizes that with graphics, and it's used quite a bit. It's actually the largest used digital platform that that exists based on our market share. I had no idea that.

That's awesome. No idea that was the case. Very good. And a lot of people use it.

And then there's other companies that have other products. Rockwell has a product called emulate. 3d Siemens has a product called process simulate, and that chart that starts to bring, kind of the broader environment into it. So our software is very robot focused.

You have a robot, you maybe have a couple peripheral things around it. You could put multiple robots. You can kind of simulate a line, have a digital twin. As you get into some of those other platforms, you start to then bring in the rest of the plant.

So you might have other AGVs in there. You might have various components of a plant that are brought into that hevs would be an automated guided vehicle, or an Amr, an autonomous, automated mobile robot, or an AMR when you get to the omniverse, and then specifically Isaac sim. It's really a platform where it's a different type of structure, where now you can bring components from all of these things into that platform and start to really run various scenarios through there. Some of the demonstrations I've seen that are current exist in an auto plant where, unlike a true digital twin, where you have it in real life, and then you and you have it in the sim world.

What I like about the omniverse, it's my personal opinion, is you can build it digitally, and then you can run it through various scenarios. So you can have your whole digital factory set up and it's running, and you can say, What happens if that person doesn't show up and that AMR breaks down? And see what the result of that is in productivity, in the workflow. And to me, that's really some of the beauty of what this software is able to do is not so much on the digital twin side, where you're running it in real life and you're running it digitally, but this is what we would call for the virtual commissioning part of our business, where you want to do everything digitally before you even start to do the physical side, and you want to virtually commission it, so you almost have a digital runoff, so that when you want to actually physically commission it, it goes really fast and and so our support currently as part of our partnership, our support of the omniverse and Isaac sim, is that we created digital assets of every single robot product that we make, and that's Now available inside Isaac sim.

So you can bring in any FANUC product, and the NVIDIA Omniverse uses a file format called USD, universal scene description a USD. That's what it stands for, and it's the same technology that Pixar uses to make movies. And so we've created these USD representations of the robots that comes with a lot of information about the robot, and now you're able to import that and use those assets as you're creating a factory that's so exciting. And just to unpack a few things there, you know, first of all, we talk in on the podcast a lot about the convergence of different areas and how we can, like, learn.

And from maybe art in terms of iterating and manufacturing, and there's lessons that manufacturers can can teach, you know, artists and so this whole idea of, you know, if you think about the Medici effect, which kind of goes back to what was happening during the Renaissance, I'm going down a rabbit hole. I'll crawl back out of it. But you think about these different convergences of different, you know, whether it's science, whether it's art, whether it's law, and in the in the way that all of these different ideas come together to create innovation.

And here we are talking about something that Pixar is using so literally, like on the entertainment side, that is influencing what we're doing on the on the in the manufacturing world, which is really, really fascinating to me, but also just, we'll take a little bit of credit here on the TechEd podcast, you mentioned some great companies, Rockwell Automation, of course, and their digital twin software emulate 3d Blake Moret, who I know, you know, well, Chairman and CEO of Rockwell, has been on the on the podcast.

He's coming back, by the way, this summer, so we're already, we already have a date set to bring Blake Moret and Rockwell Automation, obviously, huge, huge company in the automation space. So Blake will be back. Barbara humpton, at the time, was the, the CEO of Siemens, and she's been on the podcast. And of course, neither one of them are getting anywhere close to the five episodes that Mike Chico has, but, but, but it comforts me to know that we're talking to all the people that we should be talking to in terms of how this innovation is happening.

And so am I understanding this right? I'll paraphrase and kind of just make sure I'm following everything you're saying. So if I am putting together an automated manufacturing process, and you know, for starters, before I start putting robots and conveyors and programmable logic controllers and sensors and devices into my into my plant, I want to say, All right, let's, let's really innovate this digitally, which is a lot less expensive and less impactful to make a mistake digitally than it is to do it in the in the real world.

So I can start to pull the benefits of some of these different softwares. And I can maybe take a fan x6 axis, traditional industrial robot, a fan X CR X collaborative robot. Maybe pull in a rock wall PLC, maybe pull in some sensors from Bosch, these kind of things, and integrate those virtually in Isaac sim, in the omniverse, to kind of do my virtual commissioning. I love that term, to do my initial runoffs before I start putting that to work out.

You know, in the in the physical world, am I? Am I getting that right? And what would you add to that? Yeah, so that's, that's the ultimate goal.

There's still some work. I mean, we're just starting this, this, this partnership really just started. So there's a lot of work to be done for that. One thing I'd add that's important, that's that's really impactful about the omniverse in general, is the the ability to start doing synthetic images and synthetic objects that have real world, physical properties around them.

So one of the one of the things that we've done. It's a confidential project, so I can't get into the specifics of it, but we were able to take in synthetic images of the product that we were handling, and then as the robot was picking one of the synthetic products, the rest of the environment reacted, not just like in a like in a fake way, but because the omniverse knew the the physical properties of these objects, they fell in the right way. They had surface finish in the right way.

And so we were able to run through millions of totes of these parts synthetically and proved out really what the reaction was going to be before we physically did it. So that's also one of the beauties of that is, is that that those images now we I really ran a real a real product. It wasn't about just moving the robot around and finding out the cycle time, but really exactly how the environment reacted. And then, as I mentioned, then you can start to inject problems.

You can say, oh, what happens if this breaks or that breaks, and see what the outcome is. So that's the future goal on a small scale. We've already done quite a bit of that, but I think the horizon is endless. On this horizon is endless indeed.

And I just think about, I was in contract manufacturing for a long time. So is, you well, know, and many members of our audience know, the idea is that you're not, you don't have your own specific product. You are manufacturing components for or assemblies for other companies. And the vast majority of manufacturing is actually in that space.

It's not necessarily the company that sells the end product, but it's the folks that are, you know, that are building the components. And that was the space that I was in. And so we would sit in a boardroom and say, All right, we're going to invest, pick a number, you know, $5 million in a new manufacturing process, and we're going to put in a new coatings line, for example, and it's going to be automated, and it's going to have all these different pieces and parts. And then you'd sit down with two dimensional schematics, usually, right?

So you'd have somebody that throughout this process, and you'd look at your whole value stream, and you'd figure out where your potential risks were going to be. And that was like pretty forward thinking 20 years ago, even 15 maybe 10 years ago. And now we're in this age where not only are we able to create 3d digital versions of the physical manufacturing plant, but also. The parts as they're coming in, and you're exactly right.

I mean, how that when you, when you said to me, we can, we can emulate, or we can create a digital version of how a part would fall on the floor, or how it would sit, I just think about, like, how it would sit on our production rack, like, so I rack it from a hook on a coatings line, and knowing exactly how that's going to sit and how, exactly how it's going to present itself to maybe the robot that's got a, you know, that's got a paint robot that's going to spray paint that part, or powder coat that part.

I don't want to miss, and I don't want our audience to miss how significant that is going to be in terms of driving so much waste out of manufacturing on the front end, and also being able to practice and kind of set up these scenarios to your point. And it's everything from somebody who doesn't come to work or this AMR freight fails, was the examples that you cited, to some of these really finite things that can happen in a manufacturing process that you don't even think about until sometimes they happen and all of a sudden you've got, you know, 5% of your parts are bad.

I mean, it really, is truly transformative in terms of what you're talking about. Agree? Yeah, absolutely, absolutely. And then, you know, I'll say one of the ongoing challenges that we're going to have that we need to continue to work on right now, you know, I have a bunch of really highly trained AI software developers kind of working on this stuff that where we sit right now, I've got The highest end programmers, and it's almost well beyond my knowledge space on how any of this really works.

I see kind of the resultant movie that someone created. But somehow we've got to continue to focus on making this easier and easier so that it can make its way more regularly on the factory floor, where some of these things right now take some really high end AI programmers to accomplish, and that's always our ultimate goal is, is to try to make it easier and easier so that everybody can can take advantage of these tools in the future. When you think about what's happening with generative AI, I had a huge project we were working on in one of the states, here in the US and had a whole bunch of information roll in all at once in the form of Adobe PDF files.

And I literally, like, just took those and dumped them into chat GPT and said, Build me. You know, these three different spreadsheets with these three columns comparing these three sets of data. And I'm not kidding. I mean, that would have been a six hour project for me prior to gpts, and it was literally, like four minutes, and it had them all built, and I had to massage them and make some changes.

I mean, that's the age we're living in in terms of just data analytics. Does it feel like we're kind of getting there on the computer programming? Should say the robot programming side of things as well. You know, right now, we're teaching students about, literally, like, how to write code to make a robot do what we wanted to do.

How far away are we from being able to do the same thing that I would be able to do with chat GPT to compare spreadsheets in terms of writing a robot program to have a robot do what we wanted to do on the shop floor? Yeah, I think what you said something important is, is that you did that. It saved a lot of time, and then you went in and tweaked it to be exactly what you wanted, right? So I think we're still maybe a little bit away from someone that doesn't know anything like zero about coding to be able to go in and do it, but we are already at the stage of massive time saving ability to generate large portions of code where you're just saying what you want to happen, and then some code gets written.

The other big part of that is, is now there are these big repositories of code that exist, that others have done. So this isn't necessarily GPT generated code, but GitHub is kind of one of those, one of those sites, you know, real story here that one of my applications engineers, we asked them to set up a demo where we're using some of this open source platform that we're working on with the NVIDIA partnership. Here, you know, the robot. We call it the keep away demo where the robot is doing something, picking up a box, moving it from side to side.

And there's a camera looking at the environment, a 3d camera looking at the environment. So that's the setup. And I asked them, you know, use this, use this interface, and go try to find some code that the robot wouldn't just keep doing what it's doing, but it would physically try to avoid you while, while it's doing it. And, you know, in in the past, like a few years ago, past, that would have taken a pretty massive software development to develop code to go do that.

But somebody had already written code where that 3d camera can recognize a human form and turn, you, like, into a stick figure. And so he took that code, and then there was some other code that had been written that creates, like an, I don't even know, like an opposite vector of which way you're moving, that then got put into a Ross two interface that then made the robot capable of doing that. And next thing you know, I had the demo done, and it wasn't but a couple days that he was working on it.

Now, this is a pretty talented engineer that which is. Why I said it's not like this person didn't know what they were doing. They were very talented engineer, but still, it got done, and it was really complicated, and we're seeing that a lot right now, in terms of being able to use these techniques to get to some pretty high end programming pretty quickly. Yeah, so I mean two things out of that.

Number one is just the fact that you were able to innovate in that way, right in the in the collision avoidance, accident avoidance, making a safer workplace and so on. And using the, you know, the kind of the predictive analytics that you mentioned of, all right, where is this human now? What are they doing? Where are they going to be?

And then kind of thinking through the same thing, but the robot, and making sure that they're working together in, you know, in a safe and collaborative fashion. And then the other part of it is just this whole idea of open source code, and that was, as you know, I traveled to China last year with Todd wanick, who's the CEO of Ashley furniture, and we actually did a really fun podcast together probably three or four months ago. And you've been on the podcast five times, I think he's probably leading on the view thing.

I mean that that podcast we did China was just getting downloaded all over the place. Part of that was the topic. But at any rate, I was fascinated by in, you know, in China, so much of the code is open source. I mean, they're innovating humanoid robots, and they're, you know, teaching humanoids how to do things, not necessarily manufacturing applications just yet, but I think it's coming, and it probably be here in the next couple of years.

But then once they innovate the software side of that, that's all open source. I mean, anybody has access to it, the code ends up on GitHub. So they're, you know, in terms of the innovation and the speed of innovation, rather than, you know, we use the old term reinventing the wheel. I mean, it's exactly the opposite.

If somebody figured out how to do some of this, you go to GitHub, you find the code that you know, and then you integrate it, you tweak it. You've got to have the knowledge that that software engineer, that programming or developer, that developer would have to be able to make it work the right way, but just the speed at which you can do some of this stuff. So So talk a little bit about this open source versus closed source, like all our AI here in the US is still closed, closed source.

I can't go into chat GPT and see how that algorithm, algorithm is written or operating, unless, of course, somebody accidentally lets it out that happened a couple weeks ago with one of them, not the one I just mentioned. But, you know, we can't look behind the wall necessarily there, but closed source versus open source, and then how do we think about protecting IP, which is really important, especially here in the United States, and for a company like FANUC that's got tremendous IP and still let people iterate in an open source format, so that we can, that we can innovate as quickly as we can.

What are your thoughts on how we do both of those? Yeah, I think the term, term open source is another one of those ones that gets broadly used. Sure, maybe let me tell you. You know, we're kind of playing in both sides of this environment now, in terms of open and closed, in one sense, if you're going to create an automated solution.

Obviously, you need the robot to move around and do stuff. The robot needs to move around and accomplish a task. And so in one way, where the robot goes is the open part of it, where with what AI is bringing the AI part of this, the brain that kind of sits and that maybe has a bunch of sensors hooked up to it, so multiple cameras, or multiple force sensors and or maybe a bunch of AI inputs, even like a GPT input into it that's deciding what the robot should do. That part is open, and we've created an interface, currently using ROS two to be able to bring that into a robot, an outside source.

It actually doesn't. It could be an Nvidia GPU, it could be a CPU, it could be someone else's GPU. It could be whatever is on the outside, sharing out something. Yep, the output of that, which, again, is the collection of, like, could be a bazillion sensors and inputs and things, but the output to that, it's actually, then, pretty simple.

It's just a stream of motion commands that's going to tell the robot where to go. And that might be very dynamic where, like, as I'm going somewhere, if that GPU sees something different, it might say, Oh, I'm gonna change my mind. I want now, I want the robot to do this other thing. And that's like the keep away demo, where the robot's going like this, put the box up and down, and then if someone comes in, I'm gonna still try to do it, but I'm gonna go around this way instead of this other way to do it.

So the open part is what the AI brain is figuring out, the fascinating, the part is important that's still closed isn't necessarily where the robot's going, but how it gets there, and that's kind of the secret sauce for us, is taking that kinematic control of the robot so that This open AI brain is just saying, go here, go here, go here, go here, go here, every couple milliseconds. And then what we have internally closed, private secret, you know, confidential is, once we get that command that says, go from here to here, we feel that we can make a fanic robot get there.

Say. Flee reliably, so you're not breaking your robots by having an AI brain tell it where to go, and then also, very easily get from point A to point B. And that, that, to me, is a good explanation of the open source programming, and then we still keep closed, maybe similar to like a chat GPT, where you can't you don't know how it gives you the answer that it gave you, but you're still going to ask it to give you the right answer, and you're going to take what it says and do that, and so that that's the way right now, from an open source programming perspective, we have it where our kinematic control module is still internal to FANUC.

We just allow you to take over the robot through that streaming motion command, through through Ross well, and you say, you know, you feel that you have that competitive advantage in terms of how to get the robot where it needs to go as efficiently as possible, I would just say unequivocally that you definitely have that advantage, and that as we talk about it, without naming any other automation or robot or robot brands, when people say, What's the difference Between a fan, xxrx and this or what's the difference?

The difference between a fan X's industrial robot and this other robot? And my answer is, it depends on what other robot you're talking about. But when you talk about the ability to handle incredible payloads, tremendous amounts of weight, in some cases, incredibly efficiently and with the precision, which is super important in manufacturing, right? To get that robot exactly where you need it, when you need it there, there's just, again, I'm not trying to do a commercial for your company, but maybe I am.

But the truth is that you do that better than anybody else, and so being able to keep that proprietary super important, right? I mean, you've spent tremendous amount of resources innovating in that way, and you should be able to enjoy the benefits of that in the same way. And I love the combination, or the comparison, I should say, to a to a generative, pre trained transformer, where it's like, Hey, I don't know how perplexity is getting there. We just know that it does.

And on the other hand, I can use it day in and day out to help me be more efficient at work. You know, all of these conversations, Mike, and as you know, I was a small to mid size manufacturing guy. I mean, I was serving, I was a tier one supplier to automate automotive for a number of years, meaning I was supplying, you know, one of the, in that case, one of the big three, or maybe it was a your Honda or Toyota or, I mean, all those were direct customers of mine. But my businesses weren't, you know, they weren't billion dollar companies, right?

You take a zero off of that, or a couple zeros in some cases, and that's kind of the space we were playing in. If I'm a small to mid size business or company in the manufacturing space, and I'm hearing all this stuff about physical AI, and I'm hearing about being able to create these really complex programs if I have the right people on my team doing it really, really quickly. And I'm recognizing now that, okay, I've got all these suppliers from a manufacturing technology space, and now I can start integrating some of this, and the ability to do that is going to increase for even more from where we are now.

But if I'm one of those, you know, 200 300 employee companies, I'm a 50 100 million dollar manufacturing company, how do I need to be thinking about physical AI, and is this like something I need to be on board with now, or do I, do I have some time? I think, I mean, I really do think we have a lot of time. Okay, if you look at the latest statistics from the International Federation of robotics, that's the IFR. That's the kind of the global overriding tracking robot knowledge base, the association there, the number of industrial robots in manufacturing in the United States, we only have 300 robots installed for every 10,000 people that are in manufacturing.

Wow. Now that, wow, that's, that's amazing. Yeah, and we're number, I think five on the list or four on the list, where Korea's number one. It's like 1200 robots for every 10,000 people.

Singapore is number two. You get to Germany, Japan, a couple others, but we're a little bit down. It's surprising, actually, just as a side note, China, China consumed 300,000 robots last year. In a year, 300,000 the United States only consumed 34,000 so they're almost they're nine times bigger than us.

Amazing. This just gives you the scope of how many people are still in manufacturing. China had only as 115 robots for every 10,000 people, 300,000 robots. Amazing.

And so I guess what I'm saying is, what physical AI is bringing us is it's opening up a lot more opportunities for the robots to handle some of the edge cases that can't be automated right now. So we're always looking at, what can we automate? What can't we automate? And bringing this physical AI into it really takes some of these edge cases where now the robot has to think and reason and make decisions, or maybe dynamically change and move.

But what I would say to small to medium sized businesses is, is that there's still a ton of non AI enabled applications that we can do, simple pick and places, simple. Machine tending, welding, painting, assembly work that doesn't necessarily need the really high end physical AI stuff that you see, like on the internet and things like that. Now, maybe utilizing some of the virtual commissioning side of it, where you were now there's, it's going to be a lot easier for you to kind of simulate what it's got, what it could look like in your factory ahead of time, just like now, if you were going to buy new furniture for your living room, you could go on Amazon or Ashley furniture and make a digital representation of what your family room might look like before.

I think a small to medium sized business could start taking advantage of the omniverse and some of these tools to do that, but I'm not sure you need a suite of 3d cameras set up in a small to medium sized business, you know, looking at everything, and having that robot dynamically move around and do things. I think there's still a lot of low hanging fruit that we can get to before we need to do that all over the place. So first of all, you're the king of the analogy on this episode of the podcast, because I love that.

Count that comparison to going in which a lot of us have done right, go on to a furniture manufacturer's website and build your own living room and then thinking about, okay, we can do the same thing in a manufacturing plant. So I'm going to steal that analogy from you, because it's perfect. So, great analogy. But what I'm hearing you say is, look, yeah, some of these tools that are coming in terms of 3d design, in terms of animation, in terms of digital twinning, and true digital twinning at some point, where we've got physical assets talking to, you know, talking to the digital world, and in innovating on their own and so on.

Yeah, that's, that's, that's real, that's coming. Some of it's here, some of it's on the way. But on the other hand, if you're not starting to automate. You know, you got tremendous opportunities to to the point that you made.

You know, look at how you can integrate a paint robot into a into a powder coat line, or look at how you can add a pick and place, little robot that's maybe doing machine tending in a machining company or in a manufacturing company. And to the folks that say, are the robots going to take all the jobs? I think that number that you cited, you know, where we're still at 300 robots per 10,000 people in manufacturing. People are still using robotics and automation to make jobs safer, make them more interesting, in a lot of cases, make them better paying, make them more efficient.

But that doesn't mean that the jobs for people are going away. The key it feels like is to get started on something right to be working on some version of automation in your plant. Because if you're intimidated by doing a complete digital transformation of a 300,000 square foot manufacturing facility, you don't have to start there. Let's start with looking at where we can use automation and improve processing yield and cycle time and safety and so on.

Is that, did I get that right? Yeah. I mean, you know, maybe I don't know why I'm in an analogy mood right now, but I am, but it's, it's very similar. I'm an early adopter, so I started using chat GBT, and we have co piloted work.

So I use that quite a bit, but sure, I talked to a lot of people that have never used it, and it's kind of like one of those things is just, just try, just, just see what it looks like, and just see what it could do for you. And I'd say the same thing about automation is, is just give, give someone a chance to come in and look at your plant and say, here's what I think you might be able to do. It may be one of those AI enabled things we might you know, there might be a reason why the plant didn't automate because it needed those kinds of skills that now exist.

But I think in the most part, most of what I hear right now is I don't have the skill really. This is back to what what we do primarily at work, is I don't have the skilled workforce to be able to handle automation. I don't have people trained in robotics, or I don't think I could. I could keep them running, because my workforce isn't trained in that way.

And so I think it's important for us to continue to work on the people aspect of it, and to make sure that we train on what is needed. And as we move into the future, we start weaving in some of these physical AI attributes to it, so that the kids coming up that are that that live on chat, GPT and those kinds of things. Really do understand how this plays in because I think it's going to change fast. I do think that some of this GitHub code, you know, open source programming, is going to is going to take root pretty quickly.

Absolutely, I've got an example. I mean, first of all, I want to mention that's why I'm such a fan, obviously, of the fan exert program and the work that you're doing across education in terms of preparing that next generation of automation and manufacturing and advanced manufacturing talent. I never want to have you on the podcast without a little bit of a plug for my dear friend, Paul Aiello and the and the education team at FANUC, because just so far ahead, speaking of kind of being, you know, out in front and in all the cert schools, you know, 1000s now, 1500 to 2000 of them across the country, really, really impressive the way that FANUC is invested in the next generation.

So at any rate, getting a little bit off, off of topic. But I had a conversation, Mike. It was really interesting. I hear because you mentioned people bringing on chat, you know, using chat, GPT or using copilot.

And, you know, you just got to get started. You talk to people who haven't gotten going yet. I'll talk to a. Lot of adults, people my age, and I think people know about how old I am, who are like, you know, I made a bullet point list with chat GPT yesterday.

It was so amazing, which is awesome. I had a conversation with there's a young lady in the consulting space that I've been mentoring for a few years, and she's about a year out of college or so, and is now doing consulting work and with one of the big consulting companies. And she said this to me a couple weeks ago. She said, I actually made a bullet point list on my own yesterday, instead of using AI.

And I was like, What a difference in terms of these two generations of, you know, here we've got one person who's saying, I used it to make a bullet point list. It was amazing. Somebody else is like, wow, it was so weird to actually make a bullet point list just with a piece of paper. You know, whatever.

So that's kind of that difference, right? So let's, let's think a little bit about, as we evolve, and we think about the next generation of of talent, what do we need to be doing, both in our companies and in on the education side, to make sure that we're preparing that next generation of talent for advanced manufacturing? Because you're right, it is going to move fast, I believe in and it is going to be a, in some ways, a similar world in manufacturing, but in a lot of ways a much different world.

Yeah. I mean, I think we need to continue to focus on the basics, the why, why? Why are we, you know, what's the core of what? Why we're doing it, even if the code becomes auto generated, or, you know, we use some some level of AI to generate it.

I think we still need to educate people on the core of foundationally, what it means and why it's there. And especially when you're talking about physical AI, when you're talking about a robot moving around, some of the things that get lost is some of the elegance that goes into robot programming where, you know, using speed and acceleration and position control, you can make a robot last longer, go faster, you know, generally look better so it doesn't have, like, really jerky, unattractive motion.

So that, you know, I remember when I was still caught, a long time ago now, when I actually coded robots and programmed robots, you could tell the difference between a seasoned programmer and a new programmer based on what the robot looked like. Both things may have accomplished the task, but one was done with more elegance, which in real dollars and cents means less maintenance, faster cycle times and things like that. And I've seen the same thing with some of these AI generated bullet point lists and stuff.

I can tell you this is, this is our interview outline. And I know that Melissa didn't do this using AI because of how it's written. But I get and I have papers behind me. I know for sure when I get something that passes across my desk, where someone just went into chat GPT and said, create this business outline for Mike to read that includes this, this, this, this, and this.

It's like vanilla toast to me, like I just buzz right through it. I barely read it because I know that someone didn't actually write it. It just pulled out a bunch of buzzwords from the internet and cobbled them all onto a piece of paper. And so I really think we're going to get to a point now where there's going to be so many AI generated documents that they're all there.

They basically all say the same thing. And we're going to we're going to need to focus on human written stuff that that really does grab your attention, and it's just maybe means something a little bit more. So that's a little bit of a rant, but that's, I think it back to coding. I think that's where it is.

You still need to train the basics of why we're doing that. 100% agree. And you know, I point a lot of times on the podcast to the book Genesis, which was written by Henry Kissinger, Craig Monday and Eric Schmidt, former Secretary of State, former VP of strategy for Microsoft and former CEO of Google, respectively. And this is one of the things that they talk about is, you know, we're going to get into this age of AI where it's like, yeah, it's going to be able to do a lot of things for us, but it actually begs some really cool questions about, what is the human's role?

What is the human in the loop? And what is it that makes us uniquely human? And I think you're getting on to some things that I spend a lot of time thinking about. So for anybody that is either thinking about writing a paper strictly using generative AI, or, for that matter, thinking about writing a robotics program strictly using some code generation, there's always going to be this person that I love the use of your word elegance, because I think there is an elegance in both of those that isn't going to go away.

Just let's dream five years into the future and a we used to dream like 10 or 20 or 30 years into the future. Things are moving so fast we'll be living in a different year, in a different world five years from now, explain or describe that world to us in a minute or so, and give our audience a sense of what's coming. Yeah, I think in that five year window, I think what we will see is a lot more things on the digital side, that virtual commissioning part of it, like I said, I started, I mean, Robo guide.

I've been using Robo guide throughout almost my 27 years at FANUC. I've been using Robo guide so creating digital twin kind of things, or visual. Organization tools are not new, but I think that the the level in which we use them to really create simulated scenarios and images and stuff like that, I think that is one of those things that's going to ramp up really fast, that's going to continue to get easier and easier to use. So I think that the time it takes for someone to come up with an idea of what they want to automate, and the time in which it's going to take for companies to kind of engineer that and get it onto their floor is going to be really, really fast, and that's going to kind of follow, like the Amazon model of is is I need it in a couple days, and we're already seeing that those kinds of requests from our customers.

So that's kind of on the digital side. And then I do think that this, this idea of physical AI, is going to ramp pretty quickly, that that we're going to have code written quickly. We're going to have plants expect that code to be able to be dynamic and changing to the environment. Maybe the biggest What if is, is kind of how safety plays into all of this.

Can Can we make it to where we can follow the proper safety guidelines, where some of this, AI generated dynamic code, can be safe and in a manufacturing environment, in and around the workers? That's something that there's a lot of folks working on right now, but, yeah, yeah, that's where I see the next, next couple few years. It's really exciting. I mean, it is outside of the fact that we have government support.

We didn't even touch on that part of it here, in terms of what the government's doing, but there's a lot of cool stuff going on there. A lot of support from the administration. So really excited about the future for sure. Well, the all time leading scorer on the TechEd podcast is Mike Chico now in his fifth episode, and we're going to have you back soon.

You're right. There were a whole bunch of things you're doing, some things at the national level, in Washington, DC, that we didn't have time to get to on this episode. Within the next I'm going to say, six months, Mike. Let's have you back to talk about some of this stuff.

In the meantime. Always a fascinating conversation with Mike Chico, the President and CEO of FANUC America, one of the, if not the most respected automation and robotics experts, subject matter experts and leaders in the entire country and on the globe. Thanks so much for taking some time. Mike, thank you very much.

I appreciate it. Matt, thanks, as always, always. A fascinating conversation with Mike Chico, the President and CEO of FANUC America, spent on with us five times as we mentioned, and every single time we learn something new, in this case, about generative AI about physical AI about the incredible work that FANUC is doing to protect its intellectual property and also make as many innovations in coding and robot elegance as we talked about come alive on the factory floor. So really glad that Mike was able to join us, really glad that our audience was here to listen in.

We're going to take the show notes a number of different resources that we mentioned. We'll put those at TechEd podcast.com/it called Chico five. So TechEd podcast.

com/c i, c, c, O, and then the number five. That's where you find the show notes. We, of course, do have the best show notes in the business. Also have the best social media presence in all of technical education when it comes to the number one podcast in STEM and TechEd.

So you'll find us on LinkedIn. You can find us on Instagram, check us out on YouTube, go to Facebook, go to Tiktok, wherever you consume your social media, we'll be hanging out there. So check us out. Say hi.

We'd love to hear from you. See you next week on the TechEd podcast. Until then, my name is Matt Kirkner. Appreciate you being with us.

You.