Stephen David Miller

Startup cofounder, AI researcher, podcaster, person, etc.

Sock Folding, or, Please Don’t Let This Be My Only Legacy

In 2010, my Berkeley colleagues Ping Chuan (“Ted”) Wang, Mario Fritz, Trevor Darrell, Pieter Abbeel, and I participated in the PR2 Quick Start Video Competition — a challenge to make a 1-minute research video with the PR2 which would maximally entertain Willow Garage founder Scott Hassan.

Ted and Mario had been working on a classifier which could detect inside-out vs rightside-out socks, and asked if I could help get the PR2 to do something with it. Of course, its grippers were too fat for such a dexterous motion. So we used a pole, and got this winning result:

Now, I was a 21 year-old male at the time. And despite the funnier story which circulated the internet (“Out-of-touch researchers don’t realize what this looks like!”), we all had a serious laugh throughout the sleep-deprived hackathon. So when the video went viral, at least 4 of us were thrilled. The pinnacle of virality, in my mind, happened when I turned on the TV on 08/25/2010 and saw this (mildly NSFW):

It’s immature, gross-out, Family-Guy-brow humor. And it made me laugh a lot. When I was presenting the [eventual paper] at IROS 2011, the Attack of the Show crew were kind enough to donate footage. In good faith I never put it online, but now that the show is no longer in existence I think the statute of limitations has safely expired. I hope this doesn’t go down as the most famous research contribution I was a part of, but hey, I’ll take what I can get.

Also, here’s the IROS talk with my recut, less NSFW video:

Domo Arigato, Mr. Roboto

Last fall, I took a trip to Tokyo on behalf of Open Perception, to do some work for MUJIN, an awesome robotics company who are trying to bring some much needed autonomy to the factory. With their easy-to-use, in-browser planning solution, they’re able to find optimal trajectories for pretty much any task you can think of: avoiding obstacles, bin picking, handing off open containers of liquid without spilling.

Of course, planning only makes sense if you know the state of your environment. We were able to partner with them to build the MUJIN Scanner; easy-to-use software for building metrically accurate models of an environment, and localizing known CAD models (e.g. robots) on the fly.

Last week they had a big presence at IREX2013. It’s awesome to see what they’re doing, and cool to know that something I programmed could become (part of) a real product.

At Berkeley: Coding in a Documentary

A few years ago when I was an undergrad at Berkeley, I received an odd request. There was a film crew making a documentary about life at Cal, and they wanted to know if they could watch me program a robot.

Now, my undergrad work with the PR2 had gotten a decent amount of media attention, so it wasn’t the first time I’d had a camera pointed at my face. But this experience was very different. A crew of maybe three people came in around 5pm, helmed by this guy:

Fred Wiseman

…and planted themselves there till about 10pm. They didn’t interview me, or ask to watch the robot do something exciting. They just sat there. After a silent evening (and hopefully not too many bodily noises), they left. I hadn’t heard of, or thought about them, since.

This past week trailers hit the internet for At Berkeley, Fred Wiseman’s new 4-hour epic documentary. Turns out he’s the real deal: a MacArthur grant awardee, influential social commentator, New York Times / AV Club praised type of real deal. A few friends pointed out to me that my pre-Stanford, hippie self makes a few appearances. I haven’t gotten to see the film yet, but sources tell me I’m in a pretty sizable chunk of it, and that it’s mercifully near the beginning.

Watch for a Jesus-lookalike about 38 and 48 seconds in.

Hello World

I’ve just switched my site to Octopress. Feels good! I’ll use this space for news. Or more accurately, I’ll probably use this twice then forget about it forever.