Some look at humanoid robots and only see Skynet, but Agility Robotics is building increasingly person-like machines with one goal: to serve mankind. From the one-legged Thumper to a two-armed, two-legged delivery robot named Digit, Founder Damion Shelton has long been at the cutting edge of robots meant to do our bidding. Damion, Bryan and Alex geek out about how to make these robots actually work, where we may see them used in the future, and how technology research funding is really used. Bryan and Damion agree there’s only one way to build smart, successful AI—but what does their consensus mean for the future of artificial intelligence and self-driving cars?

Listen On
Apple
Google
Spotify
iHeart Radio
Share

View Episode Transcript

Episode Transcript

Alex Roy
Welcome to the No Parking Podcast. As always, I’m Alex Roy, here with my friend and cohost, Bryan Salesky. Hello, Bryan.

Bryan Salesky
Hey, how’s it going, Alex?

Alex Roy
Pretty good. I find out almost every week that there was something I didn’t know about your past, and I thought I knew everything. I did not know how long the history was, the history of robotics, people actually working on wheeled and tracked and bipedal robots. It goes back decades, not new.

Bryan Salesky
Absolutely. It traces back even to work in the ’70s, absolutely.

Alex Roy
It’s really cool that this week we’ve got a guest named Damion Shelton, who’s the CEO of a company called Agility Robotics. These guys make a bipedal robot, delivery robot. It folds up; it emerges from a vehicle and will deliver a box from the vehicle to your front door. But this guy, like you, has been working on some of the most famous historic robots from decades ago.

Bryan Salesky
Yeah, he’s worked on legged robots, robots with multiple degrees of freedom, which just means that it has a bunch of joints and is able to do all kinds of really interesting work. It’s different than what you would see in the types of factory automation robots where, broadly speaking, that’s teach and playback. It’s to repeat the same motion over and over and over again. These are robots that have a lot more intelligence to them that are able to perform specific tasks, that have enough sensing that they’re able to grasp objects, and in smart ways, so they don’t crush the object, right? They don’t want to hold it too hard, but not so loose that it falls out. There’s quite a bit of technology and research that is being developed in this area, and Damion’s always been on the forefront.

Alex Roy
Again, I wish I had spent more time studying math in school. Let’s just get right into this episode.

Alex Roy
I’m told that this is a no-bs podcast, so I can’t ask about that science fiction film, right, Bryan?

Bryan Salesky
Why don’t we do a little business first and then we can get to pleasure.

Alex Roy
All right, Bryan, you can lead it off.

Bryan Salesky
Damion, tell us a little bit about yourself. How did you get into the whole robotics gig?

Damion Shelton
By accident, so my background was bioengineering as an undergrad, and ended up in grad school because I didn’t have anything else to do with my time. Went to Carnegie Mellon because it was close proximity to Pitt, which is where I was an undergrad, and ended up doing medical robotics for awhile, which is a completely unrelated field to what I’m doing now. Started and sold a company that made 3D scanners, and then…

Bryan Salesky
What was the name of the company?

Damion Shelton
threeRivers 3D. Imaginative.

Bryan Salesky
Got it.

Damion Shelton
As you might guess, it was in Pittsburgh.

Bryan Salesky
We’re not known to be the best at naming things, it turns out.

Damion Shelton
It was edgy because it was the word “three” with the T lowercase. Yeah. Let’s see, that wrapped up. We sold the company in 2012. I hung out and worked on some projects for awhile, and then was looking for the next thing to do. I had been friends with Jonathan Hearst, who was the other co-founder of the company since 2001. We knew each other from grad school and got into a lot of trouble together during grad school. He went out to Oregon State, got a faculty job there. Over the seven year period after he moved out, robots got good enough that it was not obvious what we would do with the ones that he was building, but it was obvious that on the path that he was on that we’d be having something that was available for commercialization around the 2016 timeframe.

Bryan Salesky
What type of system or robot are we talking about?

Damion Shelton
He started his career with a single leg robot called Thumper. That was his thesis project. Then that was turned into a biped at the University of Michigan called Mabel, with a long time collaborator of ours, Jessy Grizzle at Oregon State. He started back with a different single leg design called Atrius 1.0.

Alex Roy
A single leg design?

Damion Shelton
Yes, single legs.

Alex Roy
It seems suboptimal.

Damion Shelton
If you’re trying to learn about legs, it captures the minimalist leg. it was on a boom, it kind of bounce around in a circle.

Bryan Salesky
You got to remember these guys don’t have a lot of money.

Alex Roy
I could see how a single leg and a claw might be effective, but…

Damion Shelton
It was also attached so that it could just go in an endless circle. It was unapologetically a research robot. Then when the company came along was right at the transition between the Atrius series of robots and the Cassie.

Bryan Salesky
Going to Thumper, I’m not a huge expert when it comes to legged robots. School us a little bit. What are some of the innovations that came out of Thumper, and then follow that thread for us.

Damion Shelton
Right, so this actually goes back a ways. This predates my involvement at robotics by quite a ways. Ben Brown and Garth Zaglin at CMU, did some work in the early nineties I think with a monopod. There’s this long tradition of single leg bouncy robots adding compliance into the machine. Thumper as is actually a technology that did not get promoted up into the final version of Cassie, but a cable drive robot that on Thumper had mechanically adjustable compliance that ended up not being necessary. Then it turns out cable drives are a pain in the butt. That did not survive the transition from Atrius to Cassie, but the general idea of having compliant machines that don’t –

Alex Roy
Try to kill us.

Damion Shelton
They don’t kill us, but more importantly they don’t do position control. They do force impact control, and that allows you to move through the environment even when it’s unpredictable versus the other robots of the 90s and early two thousands that are awesome, like ASIMO. That really required very perfect knowledge of the world.

Bryan Salesky
Yeah, so you think flat concrete floor versus sand, mud, grass field.

Damion Shelton
Right.

Alex Roy
Bryan, you built a prototype that could flip over and could transverse any terrain, is that right?

Bryan Salesky
I didn’t build it. I was responsible for a very small piece of it, which was some of the safety work that was on that vehicle. But a team of really, really smart mechanical electrical engineers led by John Bears built a whole series of what I think is still probably the most terrain capable set of robots ever, which was affectionately named Crusher, then went on to become APD, which I won’t even tell you what that stands for. Really what it was about was a lot of times the research funding in the military world dictates use cases that you then try to solve with technology. At the time, there’s a big push to build these reconnaissance robots that could go off into really extreme terrain, sit, be silent and observe is the idea.

Damion Shelton
Which is actually where Cassie came from. Yeah, if you go to the kinds of parties that I go to and you want a fascinating piece of agility robotics trivia, the first name of Cassie was actually Silent Runner. The idea was a very lightweight, no arms, just legs, robot that could run around and do reconnaissance at ground level. Then Silent Running ended up for scifi fans-

Alex Roy
I’ve seen Silent Running.

Damion Shelton
Right, so yeah, we dropped that name because it’s A) very military sounding, and B) not as charismatic as Cassie.

Alex Roy
Right, and also the movie is depressing.

Damion Shelton
Yeah, but I think it’s actually a pretty realistic portrayal of robots. But yeah, it ends with everyone dying.

Alex Roy
Yeah, but what I love about the end of-

Bryan Salesky
Which is not what we’re after.

Alex Roy
What I love about the end of the movie is that the robots were designed to maintain the hydroponics on the ship and that even in the absence of humans, they would continue to do so faithfully. To me, that’s got this heartbreaking beautiful thing at the end is the last guy dies, and the robots are fixing each other and turning the vessel towards the sun. It’s beautiful actually.

Bryan Salesky
All right, so help me out here. I want to keep the thread going. History in the robotics field is important because we build on these innovations, right? We’ve got Thumper, Cassie. We’ve got-

Damion Shelton
Thumper, then Atrius and then Cassie.

Bryan Salesky
Then Cassie. We also have the concept of compliance, which allows these legged robots to navigate unstructured. Sorry, what’s the word for it?

Damion Shelton
Yeah, we call it unstructured, but it’s also cases where you do have some knowledge, but it’s not perfect.

Bryan Salesky
Yeah, exactly. There’s a lot of parallels, and we can talk some other time about parallels between this and ground vehicles, but then, now carry this thread for me because what’s interesting is I’m sure this is now influencing a lot of what you’re doing at Agility.

Damion Shelton
Yeah. There was a huge change between Thumper and Mabel, and then Atrius, and then a big change between Atrius and Cassie, smaller change between Cassie and Digit, other than the obvious changes of adding an upper torso with arms. Where that has really I think influenced business direction is it’s helped to narrow the applicability of what we’re working on. People always want to know are you going to use Digit to move stuff around warehouses. No, because those are engineered environments. That’s already the domain of forklifts. There are obvious automation past there that are not human modeled.

Bryan Salesky
Tell us about the founding of Agility Robotics because we interrupted you

Damion Shelton
Oh yeah. We took a pretty unconventional approach I think. We had an initial conversation with DARPA, and this was more than a year before the company was even formed, went down and met with at the time, I don’t even think he was a program manager. He was an up and coming program manager and said, “Hey, let’s talk about a pretty weird idea. The idea is we’ll put a grant proposal in. The grant proposal says that if it’s funded that we’ll get the infrastructure of a company in place so that if the grant is successful, we can roll directly from university into the company.” I’ve never heard anybody before or since structuring a grant proposal that way, and ended up working pretty well. That project was not even funded until, I think, the summer of 2015 post-DARPA Robotics Challenge. Then we started putting the company in place around the end of the year 2015.

Damion Shelton
At that point, Cassie still didn’t exist. We had no idea if it would be successful, but we took an option out on licensing the tech from the university. By that summer, it was obvious that Cassie probably was going to work. We had had all of the seed investment stuff lined up based on fundraising against this non-existent robot, and cashed our first seed investment checks October/November timeframe of 2016.

Bryan Salesky
Congratulations. That model, the government loves it because you’re basically stretching their dollar. Anytime they can put dollars in, but then there’s also industry that comes in. This is where a lot of the old school robot races, we talk about DARPA Grand Challenge, the Urban Challenge, the beauty of that was DARPA put in some amount of money for some of the teams, but it generated a lot of excitement. We were able to bring in money from industry, which allowed us to take on a whole lot more than if it was just the government money.

Damion Shelton
Right. Yeah, and similar with this, Agility has never a direct government contractor. We did a little bit of sub work designing Digit, but I think this is probably a matter of public record. Less than a half million of government money went into Digit. The nice thing with the partition is we’re able to really focus the government funding on higher risk speculative science projects and then turn around and use the investor dollars on actually building stuff and getting it out into the world.

Bryan Salesky
This is really important. I think folks don’t realize how much government money goes toward speculative research that we don’t know where it’s going to lead us, but it’s responsible for so many innovations of over the last many, many decades.

Damion Shelton
Right.

Bryan Salesky
This is another example.

Alex Roy
I’m curious though. The media covers these things in the abstract, like National Science Foundation budget being cut. Where are we today versus 10, 20 or 40 years ago in terms of government support?

Bryan Salesky
It tends to be really cyclic. It’s hard to describe that in exact terms, but I think there’s always some amount of money over the last many decades for robotics work. It’s changed; sometimes it’s gone really heavy on ground vehicles. There’s been times where it’s gone really heavy on these bipedal robots that are able to climb in really extreme terrain. It goes on and on, right? There’s a lot of money in this field. Robotics as a term is extremely broad, right? But I think the key is that the good people that pursue these truly “we don’t even know if it’s possible” type of ideas in a university setting, many of them are funded through government research dollars. If that dried up, we would no longer have those seeds of innovation that we can then take to a company and try to build a real business around it.

Damion Shelton
It’s amusing having been in the CMU community now for almost 20 years, that AVs in 2001 were coming off of the Hands Free Across America Project. It was super speculative. It was maybe with a minivan full of computers we can do what is now a product that’s in my wife’s, we don’t have it anymore, but she had a Subaru Impreza that had the eyesight cameras, and it’s virtually the same technology. The idea that in a relatively short couple of decades you would go from something that was out there newsmaking research to something that is in a Subaru as a $500 option is weird. Then what Ford has done with a lot of the tech deployment, yeah, then Argo, it’s such a rapid transition that we forget how recently this was out there research.

Bryan Salesky
For those of us who are spending our careers developing this stuff, it seems like an eternity, but in the scheme of things-

Damion Shelton
Right.

Bryan Salesky
It’s actually a very short time period.

Damion Shelton
Yeah, the DARPA Grand Challenge was what? 2005?

Bryan Salesky
Yeah, the initial one was ’04, ’05, yeah.

Alex Roy
Who won that? Who won that?

Bryan Salesky
The first two races I don’t want to talk about, but the third one, we

Alex Roy
We? You? Carnegie Mellon?

Bryan Salesky
Yeah. No, not me, the team.

Damion Shelton
Also CMU only lost the first one for a stupid reason, which was the way point stuff.

Bryan Salesky
Yeah, no, it’s true. Let’s not get into all of that.

Alex Roy
Yesterday I saw an article about Agility Robotics written by our friend Ed Niedermeyer, who was very favorable.

Bryan Salesky
It was a great article.

Alex Roy
Yeah, very favorable. There’ve been a lot of pieces about bipedal robots doing deliveries that have been skeptical and negative. Ed is a big skeptic of all of this coming out, really like what you’re doing. Why?

Damion Shelton
First off, technologically stuff has advanced quite a bit. Again, to get on the theme of how stuff has changed, five years ago for us, that would have been Atrius too, so a robot with no sensors that weighed 200 some pounds, couldn’t get up by itself. It’s hard to look at that and say, “Hey- “

Bryan Salesky
It had a tether, right?

Damion Shelton
It had a tether, was just barely off the boom, had no feet.

Alex Roy
Only 40 bullets and a claw.

Damion Shelton
No claws, but and then where was the rest of the industry at that time? That was early Atlas from Boston and Amex, which was way more capable than Atrius was, but weighed 500 pounds, was powered by hydraulics, a whole lot-

Bryan Salesky
It was extremely noisy. Did you enjoy when the YouTube video came out that mocked the BDI robot?

Damion Shelton
The swearing mod one?

Bryan Salesky
Yeah, it’s-

Damion Shelton
I’m a little bit…

Bryan Salesky
Did that annoy you or did you enjoy it?

Damion Shelton
Both. I would say I enjoyed the entertainment value of it, but I think it’s often viewed as or people try to suggest that there’s a competition between us and BDI that’s just non-existent.

Alex Roy
Why isn’t there?

Damion Shelton
Because we’re at a point where the competition is not robot versus robot in terms of social impact. This is very much an early cars versus the status quo of horses. It’s not that BDI is out there selling a million Atlas robots or a million Spot minis a year and is taking some dominant market share. In fact, because of the amount of money that’s been put into them, it’s very important to us that the space be viewed as investible and profitable. It would not be a dream scenario to have them go out of business. That would make it impossible for us to raise follow on funding.

Alex Roy
When I saw your video and I was reading about you, it seemed to me that a decision was clearly made that this robot would appear helpful, that it is friendly and helpful and supportive, and it is not a human replacement robot. It’s a human helper robot.

Damion Shelton
Right.

Alex Roy
Visually, visually.

Damion Shelton
We had announced the existence of Digit back in February, and actually intentionally there was an IEEE article that had some still photos, but we had intentionally withheld all video because we wanted the first appearance of the robot publicly to be in the context of doing a useful job, being shown in human environments, not just us walking a robot around the forest, which you know is how we do a lot of our testing. That doesn’t tell any story; it doesn’t show any vision. That’s why we had held virtually all of the media until yesterday was really the desire to show it positively. Then in terms of why is this being received positively, I would hope it’s because we’re willing to admit where we don’t have technological solutions. Why are we partnering with Ford? Because we don’t build self driving cars in the same way that we’re not a delivery service. There’s going to be necessarily some tech stack on top of that to schedule package pickups. We’re not going to turn ourselves into robot delivery service number 2000 or something. Hopefully what comes across is that we’re trying to integrate this into our rational business plan.

Bryan Salesky
You’re focusing on what you know.

Damion Shelton
Right.

Bryan Salesky
Which is the software, electromechanical pieces of what makes this robot. Can you speak a little bit to Digit and its capabilities?

Damion Shelton
Yeah, in terms of historical context, so we tend to build in threes. There was Atrius 1, 2, and then a 2.1. Cassie similarly went through V1, V2 and V3, and then V3 was what we sold. Currently what’s here is Digit one. There was actually a Digit 0. Digit 0 was so terrible that we just assembled it and never speak of it again. Digit 1 was really a functional design study. It’s shows us where the hard problems are. Among other things, we were cleaning up a lot of the cabling, a lot of the thermal management and so on, all stuff that’s already in progress. We’re going to be announcing shortly, and happy to include it here, that Digit Version 2 is actually already in production, be available mid-June also has an internal test unit.

Damion Shelton
In terms of the capabilities, it can carry about 40 pounds, about 20 kilograms plus or minus, weighs about 40 kilograms, so about 88 pounds, about what a very thin person of the same height would weigh, for somebody who’s 5’5″, 88 pounds would be pretty lightweight. It’s designed to be broad strokes about the same physical capabilities as someone who’s not a bodybuilder. This is not a robot that’s out there lifting cars off of people in car wrecks or carrying pallets around warehouses. It’s about what a person can comfortably do.

Bryan Salesky
But it is powerful enough to do meaningful tasks, lifting packages.

Damion Shelton
Yep, absolutely. Yep. That’s been a big project for us is to figure out mechanically how do you put together something that combines long run time, reasonable CPU power, actuators that are safe to be around, can do a useful task. What’s also funny is chemistry-wise, if you look at battery runtimes, if you do the calculation and you feed a person hamburgers that have the energy density of the battery, and they’re about the same size as the robot, then all of the physics balances out in this very surprising way that makes the universe actually make a certain amount of sense.

Bryan Salesky
I never heard it put that way before, but…

Alex Roy
Is Digit charging? Is it recharging in the vehicle? Is there a coupling?

Damion Shelton
Not right now. There will be by the end of the year. That hasn’t been a priority now because right now the battery is so large relative to the needs of the robot that there’s just been no requirement that we recharge during most days.

Bryan Salesky
Where would you say you’re really pushing the limits of the state of the art with Digit?

Damion Shelton
In terms of limits of state of the art, it’s more operationally than I would say technically. It’s that we’re trying to think through what does it really take to keep something up eight to 10 hours a day, make it so that if it rains you don’t have a

Bryan Salesky
That means you really are in the commercialization phase here. That’s really exciting.

Damion Shelton
Right. I would say by the end of the Cassie project, we’re actually in the process of delivering our final Cassie now. I think the customer is out training next Tuesday. That really answered the majority of the first phase of the science questions. We’re now into the phase where the science is more focused on making a useful product than it is on very abstract “how do we make a robot walk” questions.

Bryan Salesky
What are your early customers focused on doing with it? What are some of the tasks?

Damion Shelton
Exclusively research.

Bryan Salesky
What do you think are some of the things they’re trying out though? From a product standpoint, just going broad, what are the things that you could imagine them doing with it?

Damion Shelton
A lot of the research also parallels other fields. Jessy Grizzle at University of Michigan was our first customer. He also has a very active research program on prosthetics and assisted devices. The math transfers very well back and forth between robots that walk in assistance technologies. But most of the first 10 people who bought Cassies were unapologetically controls researchers because there wasn’t any sensing; there was no upper torso. People asked us, “What’s was the robot good for?” It was good at being a walking robot, but not much else.

Alex Roy
Is it good enough to pick up the pizza box if it is knocked out of its hands?

Damion Shelton
Right now with Digit? No, because with Digit Version 1, the sensing stack is limited to non-existent, with Version 2, probably if that was a desired behavior, sure.

Alex Roy
It’s desired.

Damion Shelton
Okay. Yeah. With the demo that was filmed, that’s a question people ask is how much of that was faked, and the answer is none of it was faked. The robot did do all of the things that you see in the video. It’s a matter of really expanding that out. We encourage a healthy skepticism in trying to not get people to read too far in between the lines, that the robot sees somebody and walks around them, doesn’t mean that it has an emotional awareness or anything like that. It gets arbitrarily complicated in these edge cases to start addressing the full complexity of the world. But yeah, I think people will be pleasantly surprised at where we’re at by the end of the year with Digit Version 3.

Bryan Salesky
That’s really exciting. Are you actually building all your own motors with the force feedback and everything?

Damion Shelton
We are assembling all of our own motors. We do buy the raw pieces in from a German company that we work with.

Bryan Salesky
You have to go pretty deep right into the design of those components in order to get the signals that the software would need to do intelligent tasks.

Damion Shelton
Yep. In fact, the only commercial pieces that we source whole in the entire robot are the circuit boards for the CPUs on the robot, and the amplifiers for the motors. Everything else is either custom manufactured or assembled with bits and pieces that we specked in.

Bryan Salesky
Yeah, this is huge, really huge appreciation for the level of systems engineering and component design that goes into building these types of robots. People don’t realize; it’s not like you’re buying off the shelf motors and stitching them together in some joints, and then put some software on top of it. It’s actually a struggle just to get from what I remember in my limited experience in doing this at Carnegie Mellon, just to get a one degree of freedom motor with the level of feedback that’s really accurate, really precise, that gives a software an opportunity, then understand what is it grasping? What is it feeling? Everything, all the sensing that goes into that is actually really sophisticated. There is no part that you can just buy off the shelf that does it today.

Damion Shelton
Right, yeah. That’s been a big issue with us. On a related note, simply purchasing and keeping track of all of those pieces is hard. Our very first full time hire after the founders was a PhD roboticist who was a generalist. But our second full time hire, Wendy, came out of the Navy and does purchasing for us because fortunately I had enough sense to not try to do it all myself.

Bryan Salesky
Good work.

Damion Shelton
In retrospect, that was a good decision. In terms of the support stuff, out of a 20 person company, we have a full time purchasing person, full time inventory, production manager, a test manager, an electronics technician, and mechanical technician. Those support people are absolutely critical to getting the robots built. On just the Cassies, those had I think something on the order of 600 unique parts, not including circuit board components.

Bryan Salesky
For all the kids listening, this is important, right? A lot of people think the only way to get involved in robotics design and development is to get a PhD from MIT, CMU or something.

Damion Shelton
Yeah, definitely.

Bryan Salesky
It’s totally not true. We need people who know how to work with their hands, who understand enough about electronics and electrical systems, but there is a significant amount of design, fabrication and assembly that happens at the lowest level to put these types of devices together.

Damion Shelton
Yeah, so one of our most recent hires, Corey, who’s our mechanical technician, before she joined us, was an aircraft mechanic, and then also a technical seamstress.

Bryan Salesky
I love that. What does a technical seamstress do?

Damion Shelton
People who work with fabrics that you have to sew using unconventional techniques. We—

Alex Roy
What? Like Kevlar? What?

Damion Shelton
We’ve actually talked about, with the robot, with very complicated high degree of freedom joints, how do you make it waterproof? One answer is to try to do bearings, seals and all of it, but those are high friction. They cost a fortune. We’ve actually been very seriously looking at doing fabric covers, like a CV boot like you’d find on a car. In that case, you have to work with Cordura and Gore-Tex, and there’s a very non-trivial amount of technical expertise in just working with those kinds of materials.

Bryan Salesky
It’s so cool. This is the thing I love about robotics is how you can bring people in from literally every walk of life, every discipline and we can find a job for you.

Damion Shelton
Yeah, absolutely.

Bryan Salesky
Isn’t that cool? I love it.

Alex Roy
We are running out of time here, gentlemen.

Bryan Salesky
I’m enjoying this.

Alex Roy
I can tell you are. That’s why I’ve kept quiet. A couple of questions. How do you educate people? Is it comfort and trust in technologies like Digit? It’s a big leap for some people that a thing might show up at their doorstep holding a pizza box. Think about the cultural and historical baggage that comes and pop culture is guilty. People like me are guilty, Cyberdyne Systems, Miles Dyson, of inventing all kinds of fears that may or may not be justified. How do you compel people to trust this?

Damion Shelton
I don’t think we can compel anybody to trust it. I think it’s an earned thing, not a compulsion thing. I also think that the recent history of robotics suggests that if anything, people are probably overly compelled to trust systems that maybe that they shouldn’t.

Alex Roy
What do you mean? What’s an example?

Damion Shelton
An example being the open questions with the Tesla autopilot.

Alex Roy
I was hoping you’d have another example, but…

Damion Shelton
Okay, I can probably come with another example.

Alex Roy
Nope, it’s Tesla.

Damion Shelton
Not understanding the tradeoff where a system has say 99% capability like you’d find in any modern jet where there is an appropriate time to take over from the automation and understanding when that time is. People are inclined to, as soon as they have any evidence that something is safe, just do it. This shows up across all sorts of things. People not wearing appropriate protective gear when they’re working with pesticides. Just because you’re probably not going to get cancer doesn’t mean that it’s a good idea to just go out in shorts, smoking a cigarette, spraying Round-Up or something—

Alex Roy
For you, a story; for me, Sunday.

Damion Shelton
But yeah, no, but to get back to us, first thing is to get the robot out there, not have this be something that sits in a lab and then suddenly appears overnight with thousands of them.

Alex Roy
You sound like Bryan, got to get the stuff out in the world and test it for real.

Bryan Salesky
Yeah. You have to get real learnings. You can’t build this stuff solely in CAD or in a virtual world. You have to get real world learnings.

Damion Shelton
Yeah, man, that’s manifested itself in some of our decisions that have nothing to do with the technology, but how big of a facility do we build? Do we want 50,000 square feet of indoor robot testing area? Then we ultimately decided that all that does is teach us how the robot behaves around roboticists who are building robots. That’s not actually a pretty interesting data point. What’s a lot more interesting is just to force yourself outdoors even though you’re uncomfortable with it. It runs the risk of breaking the robot. You do have to deal with the public, but you’re going to be dealing with that anyway.

Alex Roy
Yep, exactly right. We feel the same way. What’s the craziest thing you’ve seen on the road lately?

Damion Shelton
In Pittsburgh, there’s a lot of deer. Deer are very unpredictable and the number of people who don’t pay attention to the presence or absence of deer is very surprising. But I would say the craziest thing that pops into memory, the all time craziest thing I saw was a replica of the Ghostbusters car that just appeared in traffic. I was trying to decide whether I was over tired and hallucinating, but just a person on a highway driving and passing the Ghostbusters car, like it was not a big deal. But in terms of recent stuff, people don’t seem to remember their driver’s ed. I was coming up Route 8 from Aetna heading up towards my house in Pittsburgh, and traffic lights were out in Shaler. Of course, traffic law is when traffic lights are out, all traffic lights are stop signs.

Bryan Salesky
No one follows that.

Damion Shelton
No one follows it. In fact, what happened was I did it and then somebody blew around me, honking with the finger.

Alex Roy
They get real angry.

Damion Shelton
Almost killed someone in a crosswalk. Yeah, it’s the sort of thing where a robot has a rule set, it wouldn’t have any problem following it.

Bryan Salesky
I’ll tell you, we see this pretty often, is will be stopped waiting, yielding for a pedestrian. The people behind us are really irritated. Oh, it’s this robot car again. It stopped and they don’t realize we’re actually yielding to a pedestrian. They then very angrily try to pull off and go past us, and then they end up hammering on their brakes not realizing that, oh, actually there’s a person in the crosswalk. This happens pretty frequently in city traffic. Yeah. No, Pittsburgh has no shortage of really fun test cases. It’s the second animal related answer that we’ve gotten today. Now you’ve added another type shape of vehicle. We’ve known about the Oscar Meyer Wienermobile. We need to add the Ghostbuster replica then.

Damion Shelton
I have no idea where it came from.

Bryan Salesky
Was it in Pittsburgh?

Damion Shelton
It was in Pittsburgh. It was actually out towards my parents in Ligonier.

Alex Roy
All right, guys, we need to wrap this up. If we want to learn more about Agility Robotics and your social media website, where should we find you?

Damion Shelton
Yeah, I should say I’ll take responsibility. I am personally bad at keeping our website up to date, but we have a YouTube channel and a Twitter account that are both where we dump most of the stuff.

Alex Roy
At Cyberdyne.AI?

Damion Shelton
It’s Agility Robotics YouTube channel and @Agilityrobotics, all one word, on Twitter.

Bryan Salesky
Congratulations on the announcement.

Damion Shelton
Thanks.

Alex Roy
Quite cool.

Bryan Salesky
Really great talking to you.

Damion Shelton
Yeah, thanks.

Alex Roy
As you guys were talking and I was scrolling through his bio, looking for anything that I could hook into, I can’t help but thinking what percentage of the best people in the world in robotics went to Carnegie Mellon? Is it 70%? What is it?

Bryan Salesky
I’m biased. It’s all of the people. No, I’m kidding, but certainly, look, they say that it is the robotics institute, which is no longer the case. There’s other universities that now have a robotics institute, but they were the first and they had the first PhD program for roboticists. Look, there’s a lot of history there.

Alex Roy
When you go to Carnegie Mellon, is there anyone who just goes there who doesn’t intend to study robotics?

Bryan Salesky
Actually, CMU has a very serious and well known drama program.

Alex Roy
Really?

Bryan Salesky
Very, they’re world renowned for their arts program.

Alex Roy
When one looks at a map of Pittsburgh, before I started spending time there, I wasn’t even aware that entire neighborhoods are filled with startups, and that things like the Strip District, which at one time was just desolate, has become a really cool place to live.

Bryan Salesky
Yeah. It’s cool to have close proximity to Carnegie Mellon and have access to all of that knowledge.

Alex Roy
That seems like that’s the secret sauce of being in this sector today.

Bryan Salesky
Lots of cool roots that come out of there for sure.

Alex Roy
Good stuff. I’m also relieved to know that these robots are performing something for the public good and will not be like they’re depicted in science fiction. All right. If you want to learn more about Agility Robotics, you can go check them out at Agilityrobotics.com. Our podcast is www.noparkingpodcast.com. You can follow us on Twitter. You should, Noparkingpod. Bryan, on social yet?

Bryan Salesky
I got too much other stuff to do.

Alex Roy
That’s where he and I differ. I’m AlexRoy144 on all platforms, and if you want to recommend a guest or a topic or any feedback at all, you can find us. You can reach us at Guests@noparkingpodcast.com. We’ll see you next week.