Mike Winter starts in an Alaska winter with little snow-going robots, opens Wired, and calls Mark Thorpe about Robot Wars — then builds the promo video, wins middleweight with a flipper, and finds a whole tribe of ILM and Silicon Valley people who just wanted to make dangerous fun machines.

We get into warehouse humanoids, googly eyes on combat robots, how people actually work with machines, and the Cannonball philosophy: never be bored, and always do stuff when you don’t know what you’re doing.

Transcript

Mike: In 1994, I built some little ‘snowmobile ones’—just out of batteries and little motors—that were very simple to go through the snow. Alaska: in the wintertime, it's gonna be very boring. You don't really do anything. I opened up Wired Magazine. There was a picture of a guy, and he had a robot with a chainsaw on it: ‘Come to San Francisco. We're going to have Robot Wars.’

Pablos: Mark Pauline?

Mike: Mark Thorpe. He did the demonstration of what a pulse jet will be like one foot from your ears, that you can hear several miles away. I phoned up Mark Thorpe and I said, ‘Are you really going to have Robot Wars? Are you going to make this? I'm very interested—if there's nothing fake about it. I want to test my abilities against other people's abilities, just have fun.’ He said, ‘Yep, that's exactly what it's gonna be.’

He said, ‘I'm looking for money.’ I said, ‘Do you need to make a video?’ He told me what he wanted. I made a promo video. He used that, took it around, and got an investor. He came out to Fort Mason Center, and there was a bunch of people that did the same thing that I did. They were just waiting for this. There were mainly a lot of people from ILM, Star Wars people, Hollywood special effects people—people that knew a million times more than I did. They knew how to design a practical thing and do it efficiently.

I won my middleweight class because I had made a flipper. It had the ability to take a block and throw it 20 feet into the air, which was highly entertaining in the traffic circle where I tested. The neighbors would bring out their drinks and watch me throwing stuff in the traffic circle. All that testing paid off. The big thing was I made a million fans.

A couple years later I met Will Wright. We started the fun club. John got Oscars off the wazoo. Him and his brother made Photoshop. He's awesome. Just tons of like-minded people. They like making stuff.

Pablos: I think that was one of the things I just loved about being in Silicon Valley in the '90s. It wasn't as frenzied around commerce as it is now. Everybody had time to also do some weird shit.

Mike: It was more—you're right—it was the people that were in Silicon Valley because they had done something. They were all over the place, doing cool stuff and willing to invest a lot of time into having fun with technology.

Pablos: That's pretty rare these days, unfortunately. It's become a major league game, and everybody's so busy trying to make their company successful or something.

Mike: I had revenge on the successful thing. Around the time of COVID, Amazon said, ‘You're going to have a contest with DeepRacer.’ Footlong RC cars had to go around the track, and they were going to have this all around the world. They were going to use their AI for it. You had to give them 500 or something, and then you got all the parts.

People all around the world did this. I started reading what they were doing, and I realized that AI didn't really work. I also realized that the RC car has very loose ball bearings. It was a cheap one, so the wheels were all wobbly. To go left, both wheels would go left in parallel.

I took my obsessive behavior, tightened up, and put high-quality bearings in there, high-quality this and that. Then I got rid of their AI, and I made my own AI, which is very simple.

I don't want to use math. I'm not that smart, but I know how to take a lot of pictures of black and white, of possible things on the ground. I took 50, maybe 100 pictures of just two white lines on a black background, and I labeled them. Then I used a little computer that Amazon gave us. I would just have the car look, and it would say, ‘I'm gonna turn the wheels 20 degrees,’ and it worked.

I slaughtered all the 16-year-olds.

Probably about 50 at the time—something like that. But not only that, I got a hold of Amazon's attention, and they said, “You want to come and work for us?” I said, “Your AI is kind of fakey.” He said, “We know. Don't tell anybody. But we have to get this to be a practical thing because we're going to do our warehouses.” The next day they were going to hire me, and that's when COVID struck.

Pablos: You want to work for Amazon? Might be able to help.

Mike: It's better just being a goof-off, doing it on your own.

I think their robots are not fast enough.

Pablos: The warehouse robots? I don't know. They actually built them quite early. They were definitely widely deployed before COVID, so I'm not sure.

Mike: They had shelves. The shelves moved, sometimes the robots moved.

Pablos: They have very extensive robotics teams now—the things they're building in-house. They're making their own silicon, and they're doing a lot.

Mike: I'm sure they're quite good.

Pablos: But they don't show off publicly too much. I think now is an exciting time because the computers are powerful enough to help. We're way beyond OpenCV now. I don't think there's much in the way of tech. You could have said we were missing lots of technologies to make robotics actually useful for a long time, but I just don't think that's true now. We have basically everything.

Mike: On every point, you're totally right. Everything is far better.

Pablos: I think they're going to start showing up in a lot of places in the next few years.

Mike: Are they humanoid robots?

Pablos: I think humanoids are going to be the primary thing. It's essentially because of economies of scale. Part of it is that they will fit the back of your Honda, and you can expect them to go everywhere humans can go. That logic holds. But it does seem like, why would you have a humanoid washing your dishes when you have a dishwasher that washes dishes? The truth is, if the humanoid can wash the dishes and do everything else, then you get rid of the dishwasher. You don't need a specialized machine anymore. Even if it takes twice as long, it's going to do it in the middle of the night.

Mike: But you only need a dishwasher if you have a human body. I'm just joking.

Pablos: Going back to it: you need a humanoid to load your dishwasher.

Maybe that's what will happen. But I think the one that won me over is the logic that you just get economies of scale. You can make one general-purpose robot. It can go everywhere. It can do everything. You have one thing to maintain, one thing to update, one thing to have spare parts for, one thing to get the cost curve down on instead of the 17 specialized vacuum robots that you need.

A vacuum robot has to be so cheap because it's got to be hundreds of dollars.

Mike: Amazingly—how not cheap they are.

Pablos: That's true, but they have to produce them with parts that are all a buck each, whereas if you have a thirty-thousand-dollar humanoid and it can just grab a vacuum and go, then it can be a higher-quality operation.

The humanoid is going to become so cheap—why wouldn't you do that?

Mike: The brushless motors and engineering—they have the tools to make it cheap. Maybe for a couple thousand dollars.

Pablos: I think the only thing really holding back the humanoid—the big kingpin—is hand. We have to solve hand. Once they have a good hand, basically everything else in humanoid is solved. The other part is it needs to be able to do real-time motion planning.

Mike: My daughter and a couple of people that she knows formed a company called Sensible. And they have the patent. I go visit her in her shop.

Pablos: They're working on pressure sensors?

Mike: They totally nailed it. Think it's 100 dots per fingerprint. Each one—not only does it have pressure—it has vectors. Their team has invented those, and she's making them. It is for the humanoid market, and it's kind of amazing.

If you're looking for venture capital, it's much easier if you do a humanoid robot because it's easy to explain: that's a hand, right? It's going to turn the doorknob, and it's going to open up this door. But from my point of view, you're in a warehouse, and you're going to have robots that are maybe going four or five hundred miles an hour, so fast that you can't even see them, and they're loading up trucks with little Amazonian goods.

Pablos: I don't know that.

I think there'll probably be a first version of everything that's not 500 miles an hour. It's going to be human speed because that's what people think in and what they're comfortable with, and from a safety perspective they can relate to. But then version three will be 500 miles an hour.

Mike: I work for a company called Kindred AI. I was the head of engineering there for robots. It taught me a lot. During the job interview, they asked me if I could modify a woman's sex toy to be a humanoid robot. I was absolutely shocked. This is wrong on so many levels. You don't want a sex doll assembling your iPhones.

I started making robots for them and doing testing. Marketing people didn't understand how fast and dangerous they were. We'd tell them there'd be signs all over the place: don't go anywhere near this. There's a fence. Don't touch it. Don't go petting the robot on the head. They would constantly do that.

I think in general people would just not think, “Why would the robot hurt me?” The robot's not thinking about that so much.

Pablos: You've described what I think is one case for the scary-looking headless robots that Boston Dynamics makes. You don't feel compelled to pat that thing on the head. It doesn't have a head. It looks fucking scary. Whereas if you make it look like a sex doll or make it cute and give it googly eyes, then it's going to attract people to interact with it.

Mike: We need to put googly eyes on all of our combat robots.

Pablos: I think googly eyes is the smartest thing to do to make a robot seem accessible to humans.

Mike: Have you ever heard of a seal robot?

Pablos: I've heard of seal robots.

Mike: They're wonderful. I thought, “Oh my god, this is a demeaning thing for people in nursing homes. The Japanese are going to fill them with these seals”—until somebody handed me one, and it felt just perfect. Very relaxed.

Pablos: Colin Angle's new robot is like that. I don't know if you've seen it. He calls it Familiar. It's halfway between the dog and the seal. It's very lifelike. It has lots of emotive capacity. I'll show you videos. It's really cool.

Mike: I would love to see it.

Pablos: The idea is to make a companion robot. I think he's going to start shipping it this year, maybe next year, at consumer scale.

Mike: I absolutely love the companion robot. It takes an interest in you. You're being friggin manipulated. But still.

The Boston Dynamics thing: you've been around their dogs?

I've been around real ones and fake ones. They shake the floor when they put their paw down. Boom. I think it would just smash my toe right through the bottom of the floor.

Pablos: It's really strange because the thing isn't actually biomimetic. The way it moves is not like a dog. It's a different thing that they've come up with that might make sense, but it doesn't emulate the movement of a dog, really. It's got a really strange set of choices like that: smack the ground. Nothing else smacks the ground.

I tried to argue with them about it. I said, “Why don't you guys just tear the head off a stuffed bear, put it on the robot, and people will think it's cute? They'll stop bitching about it on the internet as this terrifying robot army thing.” I think we like to be bad boys.

To me, it's a sad thing when your pioneers are scaring people off. That's what I think is going wrong with the frontier lab. They're scaring people off instead of providing a positive vision for how these things can be part of our world. With robots, it's incredibly important to do that. You've got to show people lovable robots.

Mike: I got to talk to Marvin Minsky once about MIT robots in general. He said that it really worried him a lot, too. The students have so much math and so many other topics that there's absolutely no time for taking even a single machining class.

How many courses in advanced calculus do you need to take?

Pablos: At MIT, the robots are these things that are predominantly software. Then you've got rare exceptions. Prosthetics—it's robot-adjacent.

I think a lot of the Boston Dynamics folks were from MIT, and they wanted to build robots.

Mike: I should look. I really don't know what I'm talking about.

Pablos: iRobot, unfortunately, I think they did what they could for as long as they could. Pioneering in that business is very difficult. They ended up in a financial dilemma, and the company essentially, I think, had to file a bankruptcy. But the founders have all been out for a little while now.

Colin, who's one of the founders, is the one making Familiar robots now. I don't know what Helen's doing. Rod Brooks, he's always got a few things going. I think the Boston Dynamics story is also convoluted. They built the thing, they sold it to Google, they spun it back out. Now they've split it up into different groups.

One of them ended up being funded by Hyundai to make a robotics research lab, and they were trying to do a lot of academic robotics research there. I think Hyundai wants it to be a little more practical.

Mike: Google bought seven robotics companies. When they found out that roboticists are just a hard bunch of people, the robots are just going to do whatever they want. They don't take guidance.

Pablos: I think it's a mismatch culturally for Google, unfortunately. Now is kind of a renaissance in robotics, both in defense and humanoids and all these other things. Do you have an interest in working on those things?

Mike: No. We did so much with robotics. We did hive mind stuff. We did hugging robots that patted people on the back, and we did remote control people with robots. People actually like not having free will. We could guide people around and have them shoplift in stores in Berkeley.

We would take people, put them on high-speed car seats, have them put flares on them, and have an opera singer with amplification flying around Berkeley. I got to have all the fun, but also find out: how do people work with machines?

But now I'm having a blast making dioramas.

Pablos: You're pioneering the diorama resurgence. Maybe that'll come after humanoids.

Mike: Me and a lot of second graders were out there.

Pablos: We were the second graders. And what happened with Stupid Fun Club? Did it just run its course?

Mike: It was about 15 years. We did 15 or 20 films. We did a film about a drunken astronaut going back to 15th-century Japan, stumbling around. It was a very high-quality village. It was done by somebody from ILM. They helped us, taught us how to make buildings. We did GoMotion, this five-pound robot that would just fly quickly through the set, and people in Japan just love it.

Pablos: Are these films somewhere?

Mike: They're on my website, robotmikew.com. And they probably have one view or two views. Or some more. I'm promoting them.

Pablos: You're promoting them finally.

Mike: But I just have too much fun making stuff.

Pablos: You might try TikTok because I think it really tries to reward unique, original content. A lot of people are regurgitating other stuff. It's only short-form stuff, depending on what you want to do.

Mike: 16 by 9, 9 by 16.

Pablos: You're limited in that way. But my girlfriend, for example—I don't think she's ever seen Star Wars, but she's from China, so she gets a pass, kind of. But she's seen so many Star Wars clips. She thinks she knows.

You ran the gumball race a few times? What's the story with gumball? It's something I only know through watching films.

Mike: This is part of the philosophy.

Pablos: Or the Cannonball?

Mike: Cannonball. Never be bored and always do stuff that you don't know what you're doing. I had been autocrossing, where you drive around a parking lot, and some tiny bit of racing. But I had a Honda CRX, which is a two-seater Honda. 110 horsepower.

But I replaced over a hundred parts on that because Mugen, the racing arm of Honda, had its lab right by our house: new suspension, new fuel injectors, the brakes, and really sticky tires on it.

I want to do this race. I showed my car: nine-inch Hella lights on it, and stuff all over. They said, “You have to join IMSA.” I said, “Oh my God, IMSA, that's where all the famous world-class drivers go.” But all IMSA wanted was money. I sent a check, so I got to go into this race. It was Brock Yates. All these cars had a minimum of 700 horsepower. Some of them had up to 2,000 horsepower. I think I had probably 225 horsepower with all my add-ons. And I had a five-point harness and all the safety stuff on it.

We were the joke, right? I put a cruise control on because you had to drive 12 hours straight; your co-driver drove 12 hours. I went to Honda and I said, “I need cruise control.” They said, “Okay, we can put it on there. A few hundred bucks.” I said, “I needed to go a hundred.” They said, “What the hell? Get out of our garage right now. We never want to hear from you again.” I found a garage. I said, “I want cruise control, but I only want full-on.” They installed it.

Pablos: Full throttle.

Mike: It turned out that was a really wise thing to do. There were people going 120 miles an hour on the shoulder of roads, on gravel. They were going through little towns; red lights didn't mean a thing to them. They were going on huge racetracks. We were going up hill climbs. We were going through paths in upstate New York that were gravel paths with trees. It's like you were in a state park where cars aren't supposed to go. These people had at least a thousand horsepower.

I talked to the guy behind me in line. I said, “You did this before?” He said in this Finnish accent, “Yes.” I said, “Are you a driver?” He got real mad. He says, “No, I'm a navigator. Never call me a driver.” Then the guy behind him says, “I'm a driver. They never call me a navigator.” They were rally drivers. They each had their station in life, and that's all they did. Their cars had no instrumentation. That had all been ripped out. They had maybe five stopwatches. They were doing the yelling: “Left, right, hard right, hard right, hard.” When they were driving, they would go up ramps at well over 100 miles an hour.

We were helping this one guy. We were going over the pass, and he said, “Turn on your lights.” We turn on our lights. You see the village. He's going fast, and his car blows up.

We see him the next day, and his car is fine. “What happened?” He says, “My engine blew up. We had to get a new Ford Cosworth engine from England. We had to rent a Concorde and fly it over. We flew the engineers and the engine over. They were ready early next morning.”

It's above my pay grade.

Pablos: Wow, seriously. It's boys and their toys.

Mike: It's just boys, too.

A guy from India had a Lotus 7, only one seat in the car. They had dim yellow lights on, and we drove seven days through the rain. He did it by himself. He was this 70-year-old guy.

Pablos: Shows you what humans can do, though. It's wild. You did that one time, or what?

Mike: Twice. It's seven days. We would go 700 or 800 miles a day, I think. Maybe more.

Pablos: Hold up?

Mike: It did well. It was nice having really sticky tires.

Pablos: Could the sticky tires last for seven days?

Mike: These were really gooey. I was always going as fast as the car would go. They just had to be replaced.

Pablos: After the race, though?

Mike: Yeah. We were going through towns, cruise control on. You'd have to be avoiding people, but the police were there. The highway patrol was there. A lot of the people before had used helicopters to spot the intersections.

Pablos: I was thinking a good way to do it now would be to build a drone that flies 20 feet above your car. You can see ahead.

Mike: I don't think so.

Pablos: All outlawed because there's too many cameras now.

Mike: Oh my God. People would be thrown in jail for this. You can't go 170 miles per hour.

Pablos: Now you couldn't get away. Traffic violation is one thing we're really good at enforcing at this point.

Mike: We're a more responsible people, I think. On average.

Pablos: I don't know about that. But we've shifted the irresponsibility into other areas.

Recorded on September 9, 2026