In Machines We Trust: I Was There When... Robots Learned to Run

**Marc Raibert** (0:02)
This is MIT Technology Review.
It's true that I was there when robots learned to run. That's a robot walking by.

**Jennifer Strong** (0:17)
Is it?

**Marc Raibert** (0:19)
We can go over there later. There's the robot right there.

**Jennifer Strong** (0:21)
Oh, well, hello. That's Marc Raibert and his running robots. He's the founder and chairman of Boston Dynamics, a company spun out of MIT that builds agile and highly dynamic robots. You might have seen his early videos on YouTube, from big dog robots training in the forest to those stair-climbing robots for the military.
Or, more recently, the headless humanoid robot, Atlas, or Spot the Robot Dog, which can open doors and go lie down to recharge itself. It's used by SpaceX and others to do work in places that humans can't.
These days, Raibert is building what he describes as a kind of Bell Labs for robotics, an AI institute that aims to make progress on things like getting large language models to connect up with the physical functions of a robot. I'm Jennifer Strong, and this is I Was There When, an oral history project featuring the stories of breakthroughs and watershed moments in AI and computing, as told by those who witnessed them. This episode, we meet the man who taught robots to run as he gives us a tour of his new research lab in Cambridge, Massachusetts.

**David Cervick** (1:47)
In Machines We Trust.

**Jennifer Strong** (1:50)
I'm listening.

**David Cervick** (1:51)
A podcast about the automation of everything.

**Marc Raibert** (1:54)
You have reached your destination.
So when I was still a graduate student, I went to a conference on legged locomotion. It was mostly focused on animals and all kinds of scientists who study animals. But there was one presentation by a roboticist who showed a robot that had six legs and moved very slowly and was hugging the ground all the time.
And I looked at that robot and thought, wow, that's all wrong. People fly through the air, they bounce on their legs.
They don't have three legs on the ground as a minimum. They sometimes don't have any legs on the ground. So that motivated me to go in the opposite direction. And in the early versions of my lab, I made a robot that had just one leg and bounced like a pogo stick. And it had to balance itself. It had to manage all the energy of bouncing and collision with the ground and those things. And I think that set me and the people I've worked with on a course to look at more dynamic kind of robots.
You know, my father was a frustrated accountant. He wanted to be an aerospace engineer. And his mother thought that that sounded like being a grease monkey.
But the result was that our house was always full of tools and equipment and projects that he was building. So anyway, I've always been a builder and I don't think things have changed that much, even though most of the building doesn't happen with my hands anymore. I really loved being a professor. I had a lab that did interesting work. Later on, I'll show you the prototypes of robots that we built back in that time.
But one time, we decided to take the robots out to Hollywood for a movie set. We had them in a movie called Rising Sun, which was a big movie. It had Sean Connery and Wesley Snipes in it.
And we found how hard it was to take a laboratory robot and make it work in the real world. Nothing worked right. I could tell you about the grassy knoll scene. But by the end of the week, we got something working and you can see it if you watch the movie.
But it got us thinking about the difference between working in a lab where you had all these advanced students helping you, everything was under your control, as opposed to going out in the world. And so part of the result of that thinking was deciding to start a company that could work on more real-world products and applications. And that was really the birth of Boston Dynamics.
You know, the Big Dog project was really a milestone for us in robotics. And I think the place it got to be a milestone was it started working well enough. You know, we started out with this ambitious plan of making a robot that could go out in any kind of terrain. And there was a point where it started to work well enough that I looked at it and said, wow, we haven't solved everything, but you could imagine there was light at the end of the tunnel. You could imagine if you solved a set of problems that this thing could actually be out in the world and be useful. I mean, one of the signs of that is when the robot, when we were done with an experiment out in the woods, we didn't have to carry the robot up the hill. The robot walked back up the hill by itself. That was kind of exciting for us. Up to that point, it was like kicking the can down the road. You'd have some interesting technical idea or question, you'd work on it, you'd make something a little better. But there was no concept of getting to the point where the thing could actually come together and be useful. And I think that the Spot robots that Boston Dynamics is making now are the several generations, many generations later from Big Dog. But Spot is a quadruple, a four-legged robot. There's about a thousand of them out there in the world. Some being used by universities for research, about a hundred universities. And then some of them are used by companies to do inspection of their factories or their refineries or places like that where there's a lot of equipment that needs to be maintained and checked up on.