31 comments

[ 3.4 ms ] story [ 77.7 ms ] thread
When I was briefly involved in doing research as an undergrad, Pieter Abbeel (the Berkeley professor behind this robot) was my grad student advisor. Back then I was just trying to pick up dry-erase markers off a table, and even that was extraordinarily difficult to do. To me, this video is nothing short of mind-blowing.
Anyone who's ever worked retail on the fitting-room -> shelf/hanger treadmill is now thinking feverish thoughts...

Edit: Oh. Swing and miss! Feverish was the wrong word. I was not at all implying that this is a "teh robots took our yooobbss" scenario, rather, as someone who has folded and restocked many shirts, I'm excited about the prospect that this kind of drudgework could be spared us by robots.

Philosophically, I think this is a very good thing. Sure robots do more important things than fold shirts, but this kind of AI is an important baby step to a much better world for people.

Really? robots already "do" a whole lot more important things than folding clothes and they have "threatened" industries and work sectors far more significant than that.

You must think the AIBO also threatens the ailing dog-pound, and animal shelter "industries" ..

(comment deleted)
People have had to fold their own textiles for as long as they have had shame. The more abundant technology is, and indeed any resource, the more we can free up the greatest of all resources--the human mind. Great comments noonespecial.

btw: "mahmud," the robot can do whatever it's programmed to do, not just fold clothing ;)

Every afternoon, I 'waste' 15 minutes of company time, leaving my desk as a highly paid software developer to do the dishes in the sink. No one told me to do it, we have a cleaning crew that comes at night, but I do it because it frees my mind -- the greatest of all resources.

Not all chores are "chores". Sometimes we just enjoy doing boring stuff.

It's so astonishing that what takes this robot over an hour and a half to do can be processed by our brains in a fraction of a second. I am looking forward to being witness to the tightening of that gap.
They are using just a single GPU (from the paper). So the robot can actually be sped up even today. Still nowhere near human pattern recognition though.
Since this was a male CS Phd student presumably the real research was discovering that towels:

a, needed folding

b, needed washing occasionally so they could be folded

I just want to point out that the robot's operating system is open-source and runs on heaps of different robot platforms:

http://www.ros.org/

The software is primary developed by Willow Garage, a start-up created by an ex-Googler that also has Larry Page on its board:

http://www.willowgarage.com/

You can actually simulate this specific robot and even write algorithms (in Python and C++ AFIAK) without having one:

http://www.ros.org/wiki/pr2_simulator

As you can probably tell, I really love how they're lowering the entry barrier for robotics programming!

That's pretty frickin' awesome... ideal for algorithm experimenting, and it could even help create a better-designed 'bot because you can experiment with different placements of tools.

Many thanks for the links! I'm definitely keeping those around...

My first reaction: How much time and money did this cost? How is this benefiting anything? Is this a test of some operating system that will be used on other, more useful things, or just a science project?
> Is this a test of some operating system that will be used on other, more useful things, or just a science project?

You say that like science projects don't lead to anything useful....

or

Everything is a science project until it ships.

Huh, how supremely uninteresting your first reaction is. I wish you hadn't shared your useless ungeekly feelings. But now that you have, I will share my first reaction: Cool!

Robotics is hard. We are just beginning to understand the amount of processing that goes into the simplest movement of living beings. We are still in the infancy of reverse engineering millions of years of evolution, and it is a wondrous new frontier of research!

Yes, universities do science. And cool projects. If they had to immediately sell whatever they invented, most research would never happen.
I'm just saying, I don't see how this is efficient outside of "can we make a robot that folds towels?" Sure, it's pretty cool, even putting aside its slowness, but it seems like a lot of effort towards something with little application. When you're just /learning/ robotics you're supposed to experiment and make useless stuff, but this seems like a professional project.

I guess I'm not familiar with the whole world of robots-- maybe this stuff is normal and fine I'm thinking too much at 3 AM..

The robot is picking up an object with a deformable geometry, analyzing that geometry, and then using that data to manipulate that geometry in a desired fashion. It does this with a series of objects of different sizes, shapes, and colorations. It was presumably not familiar with these particular objects ahead of time.

There's a lot going on here, and it's valuable to have done it.

A brain the size of a planet and all you people have me doing is folding towels. - MtPA
This illustrates an important way of staying competitive globally:

You don't want to be the country where people fold towels.

You want to be the country that designs & builds robots that fold towels.

Yes. In 20 or 30 years when a robot can actually fold a towel as quickly as a human, those towel-folding economies are going to be in trouble. ;)
More like 2 or 3 years. The singularity is at hand.
The singularity has been due Real Soon Now for a looooong time.
Well, let's revisit the state of the art in towel folding in 5 years.
The robot is, true to its species description to date, robotic. It is performing the exact same sequence of operations - almost like first trying to generate a 3D model of the towel (the spin) followed by some cross checks (the twists) and then proceeds to fold it. It looks like there is no learning apparent from one towel fold to the next, which would be thoroughly fascinating if it were done in even the littlest of tasks simpler than towel folding.
If it just looked a little better I would marry it. I guess this will be replacing women sooner that we can imagine.