Our Robotic Future with Marc Theermann CSO Boston Dynamics- Part 2

What You’ll Learn in this Episode:

• The role that a Spot robot, can play in a manufacturing facility (01:28)
• The type of tasks that humanoids would be very good at (06:04)
• Atlas robot’s capabilities and level of autonomy (07:52)
• The future of robotics and how they’ll work with other existing technologies (12:54)

Connect with Marc: 

• LinkedIn
• Website

Connect with Ed Bernardon:

• LinkedIn
• Future Car: Driving a Lifestyle Revolution
• Motorsports is speeding the way to safer urban mobility
• Siemens Digital Industries Software

Ed Bernardon: Hello, I am Ed Bernardon, the host of the Future Car Podcast. Welcome back to Part 2 of my interview with Marc Theermann — Chief Strategy Officer at Boston Dynamics and a robot enthusiast from his earliest days growing up. Now, more than ever before, a future filled with robots is not only looking feasible, it is becoming reality in part thanks to the “magical” robots being built by Boston Dynamics.

Ed Bernardon: In Part 1, Marc and I discussed what sets Boston Dynamics apart from other robotics companies, the differences in how robots are perceived, and Boston Dynamics’ mission for their robots. In this episode, Marc and I continue our conversation on advanced robotics technology as he walks us through the features of Boston Dynamics’ robots and their various applications. Marc talks to me about their most lovable robot, their most magical robot, and everything in between. We even explore what it might be like 50-100 years from now in a world that is enriched by our robot companions.

Ed Bernardon: Join us as we look into the future of our industries and our lives where we will happily coexist with productive and lovable robots.

Ed Bernardon: There’s a theme that seems to be repeating here a little bit. Let’s say you’re in a processing plant — and it could be for beer or whatever it is — there’s going to be stairs that go down or up and things to avoid and a big pipe that you got to walk over. And you also said with Stretch that you can bring it in and not worry about the infrastructure; that I think seems to be a really important thing because some people don’t realize that you can’t just take a robot – not Boston Dynamic robots but other robots – just plop them down somewhere without training them. Here, you don’t have to worry about the stairs; “Hey, climb the stairs,” or like you said, your unloading robot doesn’t require special infrastructure. So, there’s a lot of cost in that aspect of trying to use robots successfully. It sounds like you are really overcoming with your machines.

Marc Theermann: Yeah, last week, I was at the largest chemical company in Germany – a very, very large facility. They were trying to get a more real-time view of their process and everything that happened in that manufacturing facility. And I said, “Well, why don’t you just put IoT sensors everywhere? Replace the existing nanometers that are there or the existing round instrument with IoT sensors that can measure the flow of a pump, or that can measure the temperature of the engine; then you don’t need a robot.” And they said, “Well, Marc, we would love to do that. Unfortunately, we have 150,000 round instruments. And in many parts of this factory, we don’t even have internet. So, to outfit this 25-year-old plant, it would cost us multi-millions of dollars and multiple years to really outfit this entire facility and make it Industry 4.0 IoT sensitive.” And so, you’re right, without touching any of the existing equipment, they can deploy Spot, and Spot can autonomously read all the instruments, measure the flow through the pipes, look at the engines to see if they run hot. So, we’re bridging the gap between old infrastructure — and in this country, much of the infrastructure, by the way, in processing plants is very old — we’re bridging the gap into this Industry 4.0 world without any fixed hardware.

Ed Bernardon: Would you say that Spot is your most lovable robot? Can you be more lovable than Spot as a robot? Most lovable robot in the world.

Marc Theermann: I think there are fictional robots that are probably more lovable and more loved. But maybe it is the most lovable real robot that exists today. It’s funny also that most of our clients, if not all of them, give our robot a name. So, they buy the robot as Spot, which is our brand name. But then every client that have visited, the robot is called Fifo or Mike.

Ed Bernardon: No different than naming your dog.

Marc Theermann: Yes, exactly. So, I think that’s true. But our most magical robot is probably Atlas, our humanoid, which is our Halo robot. That’s probably the most magical robot. And for everybody that comes through our doors here and gets to see an Atlas demo, their life changes a little bit because you are really looking a couple of decades into the future when you come face to face with Atlas.

Ed Bernardon: You can’t buy Atlas yet, so it’s still more of an R&D platform. But it is, you say, the most advanced humanoid robot that pushes the limits of whole-body mobility. Tell us about the performance: how fast can it run? How high can it jump? How many flips can it do in a row?

Marc Theermann: So, you’re right, at the moment, it’s an R&D platform. And we’re really using what we’re learning on Atlas to push it into our other robots. But we’re also using it to test and push the limits of mobile robots in general, and to push the limits of manipulation and athletic intelligence. So, the robot which is currently electrically powered but hydraulically actuated is an amazing robot to watch and interact with. At some point in the not-too-distant future, we think that humanoids might play a role in certain manufacturing environments. In the next two decades, I’m certain that humanoids will start walking into our daily lives. That’s maybe a little bit further away, but I think, certainly, in my tenure here, we will have humanoid robots deployed in factories.

Ed Bernardon: What do you think the first application might be? What’s the sweet spot here for a true humanoid robot like Atlas in a factory setting?

Marc Theermann: We’ve visited a couple of car manufacturing environments, and there appear to be two types of tasks in that environment that are particularly well-geared towards a humanoid where there’s a very repetitive task where you need to lift heavy things from one side and place them elsewhere. So, this is not necessarily on the manufacturing line itself – which, by the way, is a very hectic environment – but in other parts in the facility, and we think that that’s where the robot might go first.

Ed Bernardon: Can you give us an idea of how tall, how much Atlas weighs, how fast it can run? How does it compare to a human? I looked at one of the specs, it said a top speed of around 5.6 miles per hour. And then I looked up the average human running speed, which is six to eight – so, almost as fast as a regular human.

Marc Theermann: I think you will have to come visit to see all of that in person. I think, at the moment, he’s certainly more athletic than I am. I’ve been known to not be able to do an elegant somersault and certainly not a backflip, and Atlas can do both with a lot of elegance. The top speed – the robot is not really built at the moment for speed. A couple of years ago, the company built a robot called Cheetah, which also has some fun YouTube videos that could run 25 miles an hour. So, when we build commercial robots, they’re trying to fulfill something very specific. So, if there was a commercial application for a very fast robot, we could certainly build a very fast robot. But that’s not what Atlas was built for at the moment.

Ed Bernardon: We’ve talked a lot about how your robots – and Atlas included – can handle “the lower-level operations,” say, climbing stairs and that – probably not the right word. But they’re not deciding on their own, there’s somebody behind giving them, shall we say, the high-level instructions. How do you see that ability to do perception and make decisions more at a higher level? Is that something you think we’re going to start to see in your robots in the not-too-distant future? How will that evolve?

Marc Theermann: That’s a good question. So, it’s not true that the robots always get commands from humans. They get programmed and then they’re executing these commands. So, you’re right that they’re not sentient, and that they’re making their own decisions about their lives and what they should have for breakfast. But when Atlas runs the parkour, he’s executing on a set of programs that he can choose from. So, he can say, “Okay, this appears to be a box, so I’m going to execute the box jumping program.” They do have a certain level of autonomy. But of course, it’s true that they’re not sentient. It’s also true that for robots, they don’t have a semantic understanding of the world, per se. They can say, “Okay, this is a lamp, this is a garbage can, and this is a cup of coffee.” They see the world as shapes and objects. And over the next decade, that will change dramatically, I think. And if you look at the Amazons, Googles, Microsofts of the world; they are spending a lot of resources in building out that semantic understanding and building machine learning and vision programs to help. And quite frankly, I think that’s exciting. I think we will focus on building the most magical autonomous robots, but we’re eager to work with companies like that, that bring more functionality to our robots. Because I certainly believe that in the not-too-distant future Boston Dynamics will have an app store, and you will be able to download functionality to our robots that give them vision to understand; “This is a human or this is a coffee cup. This is a gauge or this is a specific piece of machinery I should be inspecting.”

Ed Bernardon: Oh, yeah. And then it would know what to do with it – “Oh, this is an empty coffee cup. Take an empty coffee cup and put it in the sink.” I know even around my house, sometimes I’ll leave a piece of trash on my desk and somebody wants to throw it away, I say, “Wait a minute, that’s not trash. I need that piece of paper.” So, there is that high-level understanding. One more question about Atlas, and then I just want to talk to you very briefly about the future. If you go online, you also see these robots with real facial features – the sad eyes, the lips, the mouth – very, very realistic. That’s not something that you’ve done in the past. How do you see that world of robotics melding with your world of humanoid robotics? It’s a different form of humanoid-ness if there is such a word.

Marc Theermann: I think at Boston Dynamics, we’re really focused on building robots that have specific functionality. And up to this point, since these robots don’t have a semantic understanding and we are not focused on building human-robot speech interfaces, they haven’t really needed a face, because in my opinion, you need a face if you want to interact with the robot as a human. And so if you need to interact with the human, you don’t want to talk to the robot’s knee, you want to talk to probably the robot’s head. And since our robots are currently mute, there wasn’t a need for that. I think, in the future, we might change our minds on that. But personally, I think – and I think I share this view with the rest of the executive team –I don’t think our robots are going in that direction that they will have highly realistic facial features because the uncanny valley, where it gets so real that it gets creepy, is a real thing. And I’d rather build robots that anybody can look at and say, “Okay, this is an industrial piece of machinery that’s exciting and magical and really cool, and has some anthropomorphism.” But we’re not trying to be a human or a real dog.

Ed Bernardon: Yeah, no use doing it just for the sake of doing it. I think what you say makes a lot of sense. If there’s a need for communication, there’s certainly the communication get with speech alone, but facial features also communicate; the sad eyes, the slight smile.

Marc Theermann: The next generation of HRI, Human-Robot Interaction, for us is actually much more about communicating intent. Because if you encounter Spot in a car manufacturing environment, and you are a line worker, and you see Spot walking towards you; you don’t exactly know what the robot is going to do next. And so we’re spending a lot of time thinking about that: How can the robot display intent?

Ed Bernardon: Well, that’s a problem with autonomous cars, too. There’s a driver, and we know that, we drive in Boston. So, you try and get, “Does that person see me or not?” You look for that eye contact. And how do you have what eye contact provides in an autonomous car? And I’d never thought of it but you need that in Spot when he’s bringing you a tool, or Atlas is bringing you a large car part. Let’s wrap up here a little bit. I’d love to ask someone who has been passionate about robots all their lives, and to boots working in Boston Dynamics about the future. One thought is a lot of talk about the metaverse these days, wouldn’t it be great if you plop Atlas down on Mars or who knows where, I put on my 3D goggles, I interact with Mars. Do you see something with the Metaverse and these humanoid robots maybe coming together somehow?

Marc Theermann: Yeah, for sure. I’m a big believer in robots in space; humans don’t really belong in space, robots do. So, yes to Atlas on Mars, for sure. Unclear if that will happen in my lifetime. But I see the clear and present overlap. For example, I was in Germany last week at the Hamburg Port Authority, which operates the third-largest port in Europe. They have a bridge that spans the mainland into the harbor, that is a 50-year-old bridge that is showing signs of structural integrity failure. And so they are currently using a Spot robot to inspect the bridge, inside the bridge, create an entire digital twin with a laser scanner and a color 3D camera, and then you can put either a VR headset on, which transports you into the bridge. And so if you’re a specialist that sits in North Carolina and you want to teleport into the bridge in Hamburg, you can do that with the VR headset. Or you can use a Microsoft HoloLens as the operator whose name is Martin, who has been working on the bridge for 30 years, and he walks through the bridge with the Microsoft HoloLens and the HoloLens displays where there might be cracks in the concrete so that he can inspect them further. So, sometimes, I think there’s good use cases for existing technology. But I also think that if you look into the future, certainly you could envision teleportation types of scenarios where you sit at home, you are a specialist at BMW Munich, and something is wrong in the car plant in South Carolina, you put a VR headset on and you operate Atlas to the problem, and you would fix the problem through a VR headset from your couch in Munich. So, I think that’s certainly coming.

Ed Bernardon: I think that’s fascinating. And if we talk about your expert, if you send someone in to inspect a bridge or a plant, you create this digital twin of that environment. If you had the expert there, when they approach a certain area, they might want to look around a different corner. You could always send Spot or Atlas, “Hey, give me a little bit more detail on the digital twin in this area.” You can bring in some of the best expertise in the world and have them, practically, on site as close as you could get.

Marc Theermann: That’s right.

Ed Bernardon: Final question. At the end, I always ask people that are in developing the future urban infrastructure or the next generation of autonomous cars; what’s it going to be like in 2050? So, I’m going to ask you that question, but yours is going to be a two-parter. What do you think it’s going to be like in 2050? And let your mind run wild – if we were sitting here, however we would communicate 100 years from now, how do you think robots would be interacting with us? So, the 50 and the 100, what’s the world look like where we have our companion, lovable robots with us?

Marc Theermann: Hopefully, the humanity has come to its senses and we’ve re-embraced nuclear, because without that we might run into bigger geopolitical and atmospheric issues. But let’s assume that we figured that out. I do think that robots will be a big part of our daily lives in 2050, for sure. Most of the cars will drive autonomously, car accidents will be less, certainly in Boston where drivers are absolutely lunatic. Everyone will have multiple robots at home, and we’ll certainly have a robotic companion robot that everybody in the family loves at home, and the existing pets will get along just as well with them, because every time we bring pets to the office here, they actually get along well with Spot, so I think that’s not going to be a problem. I do think that certainly in countries that can get out of their way with regulation, we will have flying vehicles and taxis of some sort. There are so many companies out there that are currently climbing that hill, and I think regulation will be the biggest hurdle here. But I think that countries like Singapore, who want to be at the forefront of this, will make it easier for these companies to succeed. And hopefully, at that point, the United States and Europe will also get out of their way with the regulation. So, I do think you will see things flying around, you will see lots of autonomous, electric or hydrogen cars, and robots will certainly be a big part of our daily lives.

Ed Bernardon: Do you think, in 100 years, that a robot could be indistinguishable from a human?

Marc Theermann: Yeah, for certain.

Ed Bernardon: Sort of like the Data on Star Trek?

Marc Theermann: That will be the case much sooner because as you mentioned earlier, there’s companies that focus exclusively on making faces that look relatively realistic. I’m a big Disney fan. If you go to Disney World, there’s a lot of animatronics at Disney today, that stationary robots basically without much intelligence, but then getting very good at creating things that look pretty realistic. And I think within the next five years, we will be very advanced with our humanoid already. The biggest issue there to figure out is actually software and safety. But yeah, for sure, in 100 years, you won’t be able to distinguish a human from a robot.

Ed Bernardon: There’s certainly the mechanical movement, you’re well on your way to that. The big unknown is the brain. Well, first of all, can it reason like we reason? Can it be emotional like we’re emotional? Who knows? Do we even want robots to be emotional?

Marc Theermann: Well, there I’m sort of more in the Ray Kurzweil camp that that’s probably only 30 years away and not 100 years away that you will have a conversation with a piece of technology and it will pass the Turing Test. So, I think that’s around the corner, maybe even for you and I. And then us creating very magical hardware that is ready to catch that AI revolution, I think is the largest single opportunity that’s out there in the world, and that’s why I joined Boston Dynamics.

Ed Bernardon: Create hardware to catch the AI revolution. What a great way to end our interview. Thank you so much, Marc. Last thing, rapid-fire section – real quick questions – I’ll fire a bunch of questions at you, you can give me a quick answer if you want. If you don’t like the question, just say “pass” and we’ll keep going. Are you ready?

Marc Theermann: Ready.

Ed Bernardon: All right. What was the first car you ever bought or owned?

Marc Theermann: A VW Golf.

Ed Bernardon: Did you pass your driver’s test on the first try?

Marc Theermann: You bet.

Ed Bernardon: Did you ever get a speeding ticket?

Marc Theermann: Oh, for sure.

Ed Bernardon: Living room on wheels – imagine an autonomous car, you don’t have to drive, five-hour trip, Boston to New York. It’s your living room on wheels, you can have whatever you want in that car. What’s inside your living room on wheels for that five-hour journey?

Marc Theermann: A great audiobook, and I want the entire outside of the vehicle to be LED screen so that I could change the environment to whatever I want it to be, because sometimes that drive is magical along the beautiful New England coast, but other times it’s absolutely disgusting. So, in those times, you want to be able to switch to a drive-in Tuscany or drive on the moon. So, I want LED screens all around.

Ed Bernardon: What person, living or not, would you want to spend that five-hour car ride with?

Marc Theermann: My dad.

Ed Bernardon: Greatest talent not related to anything you do at work?

Marc Theermann: I make a pretty mean baguette.

Ed Bernardon: Oh, wow. What’s the secret to a good baguette?

Marc Theermann: A good starter.

Ed Bernardon: If you could un-invent one thing, what would you uninvent?

Marc Theermann: Fracking.

Ed Bernardon: If you could invent one thing magically, what would that be?

Marc Theermann: Fusion.

Ed Bernardon: And the last question: Tell us something about yourself that would surprise your friends and family.

Marc Theermann: I’m pretty transparent. I think what you see is what you get. I often wear my emotions on my sleeve, so I don’t think there’s much that would surprise my friends and family because I’m a transparent person. I think they know all my dirty laundry.

Ed Bernardon: Marc, thank you so much for joining us on the Future Car podcast. And I’ll tell you what, I would love to get a photo, an image of that room in your house with all those robot toys. Maybe we could get you to do a tour of the Marc Theermann robot room at your house one day.

Marc Theermann: Yeah, you bet. And in the meantime, you should definitely come here and check out our magical office.

Ed Bernardon: Willl do. I’m looking forward to it. Maybe we can have a Sam Adams together too when we’re done with the tour. Marc, thank you so much for joining us on the Future Car podcast.

Marc Theermann: Ed, it was amazing to be here. Thank you so much.