Robots That Jump

Robot Bodies Needed Before Robot Minds

Monthly Archives: February 2012

Drexel basks in Asia’s glow

This article seemed exciting for a moment – looked like the US was finally doing something in humanoid robots.

Drexel Unveils Seven Humanoid Robotics on National

Engineers Week

“…The seven identical robots will be sent to Penn, Purdue, Ohio State, South California, Virginia Tech, Carnegie Mellon and MIT universities for research. The researchers of the corresponding universities will be provided adequate training to handle these robots. This step will take humanoid robotics to the next level…”

The catch? Not really, if you don’t notice that the robots were produced in Asia, and they are advanced HUBO robots. Robotic outsourcing? One more thing the US can’t make…It’s not that I think that the US should be first in humanoid robots. It’s more that it implies that we aren’t really thinking about the problems with Robots that Jump at all.


Everyone pretends not to shake hands with themselves

Here’s a post on NASA’s Robonaut 2, an update of the tele-operated space robot designed to facilitate work on the (defunct) Shuttle and International Space Station:

Link to Robonaut 2 Site:

In itself, Robonaut 2 is quite interesting. They still care about its looks – it is more like the Rockteer than Boba Fett now. It moves faster and demonstrates greater dexterity than the earlier model. Robonaut has some ability to move on its own, but to date, it has been used in a remote-control fashion. The near-term goal – send Robonaut out in a tele-operation fashion to work outside on the Space Station – is close. The long-term goal – have an autonomous Robonaut repair the Space Station – is decades away at least.

Here’s the video:

In the video, you can see Robonaut 2 extend its arm. It doesn’t go for the outstretched hand – it moves  down and possibly to the left. However, the human is willing to take its hand and shake. That’s a good sign – the problem with robot arms for decades has been that they are dangerous, even deadly to be around. One small miscalculation by the program or arm sensors and you have been thrown across the room.

But once again, the real show is not the slow progress no Robonaut, but how people react. NASA is in serious touble in early 2012, with the proposed budget zapping it even more. The exploration of space is hard to sell. So, let’s see NASA Space Station crew communicate with an Ai! Maybe the public will like the agency better if they look a little like Science Fiction Channel.

Everyone seems convinced, both journalists and Internet riff-raff commenting on the video, that this device is a real robot that wanted to shake hands.  If the press and public were told what’s actually going on, they would assume that (1) intelligent Robonaut is just around the corner, or (2) there’s a conspiracy to hide the real robot overlords which the goverment is cooking up to mate with us, Cylon-style.

The NASA site certainly doesn’t try to enlighten anyone – it’s a good practice for an agency worried about having its funding cut to make the public think you have a true Ai – rather than a humanoid tele-operated system that might be extremely useful in some environments. Apparently, it’s good practice to say that this device is just a tiny hop away from “machine overlord” status, just so long as it isn’t actually overlording yet. We really want it to have a mind of its own, properly docile and servicing the International Space Station, but a friendly pet that will extend its steely hand in a soft glove of friendship.

People like seeing The Prophecies coming true!

Seeing this, I can’t help but think how little we’ve progressed from Roman “temple tricks”. Back in the Roman city of Alexandria, temples could have automatic doors, and mechanical gods inside ratcheted on gears to wow the mob. This kind of thing already had a long history then – lots of older Egyptian statues had channels to let hidden priests “tele-operate” the god by speaking in a divine voice of overlord-ery.

We have to differentiate what Robonaut can do (useful tele-operation work) from what we want it to do.

The reality is that our current “machine overlords” are sensor packs, monitors and dashboards that feed back information to us on ourselves. The Quantified Self movement has a better sense of reality. Computers have failed massively at being intelligent, but they have worked well as Lickleider’s “Intelligence Amplifiers” – adding a punch to intelligent behavior. Now, quantified self lets us guide our own behavior more rationally.

A “robot” that monitors my body temperature, then acts by sending messages about my health to Facebook is the goal of Quantitfied Self.

At present, the most intelligence is seen in how Internet apps augment our social communication. Rather than shaking hands with a supposed intelligent robot, it makes more sens to shake hands with our friends, or, in the case of Quantified Self, ourselves. Of course, it has its own Doomer prophecy, and a little too much trust that “goodness” is the same as “optimization” (computer operation invading ordinary life and thought, again).

A little too optimistic at TED:

Reality bites:

But then, Robonaut 2 in its hand-shake cameo is nothing of the sort. It is a temple-trick. It is a big metal puppet that doesn’t even augment our own actions. We are content to see this as a future god.

But what if the future of robotics is in an advanced form of the Quantified Self? We’ll be communicating to ourselves, via our robot puppets, that we are our own gods. It’s lonely.

Is there an app for that robot? No.

I read with some amusement (tinged with irritation) the following report on “robots that can bring you a beer”, probably in response to the upcoming Superbowl:

The article considers a few robots that might be able to “bring you a beer”. First, it highlights the Asimo open a screw-top bottle – pretty cool, if it can actually do this for real, instead of in an industrial-robot style pattern.

The last part is more interesting. The article tells you how your smartphone can be even more useful than ever, if you stick it into a little robot body. What’s extraordinary is that the benefits of having a small, hobby-sized robot motoring around aren’t explained. There’s also no discussion of whether a cellphone is smart enough to create intelligent behavior.

But the last part goes right to the center of Robots That Jump. Part of my beef with the delusions of robotics – that we are about to create superhuman creatures that will revolt and enslave us – is the misunderstanding of computer power. There is also a misunderstanding of the power of individual programs. According to the article, to make your robot better in the future, you will just “download an app”.


The first reason web apps work is that the smartphone is a very standardized device, with a closed OS (yes, even Android) that is highly predictable in behavior. The second is that web apps are typically used for communication – extending the basic use of a cellphone – rather than number-crunching or creating intelligent actions. There’s a world of difference between making an app that allows people to share photos, and software than analyzes the same photo. True, apps like Shazam hit the cloud for pattern-recognition, but one wonders whether any “cloud” of computers could handle all the cellphones running intelligent software.

The delusion in this article is: Apps blew up the cellphone market. Now there is a way to make a cellphone run a robot body. Therefore, apps will blow up robotics.

This pop science article is far from the only place we’re seeing this idea. It is as if every time a new technology advance comes along, it will “finally” kick robotics into high gear.

The problem is, IMHO that we know exactly what will kick robots into high gear – fluid motion, natural walks, movement in human spaces without modification, an ability to pick and hold objects without special techniques. No mind, just a robot that could do an aerobics class if it wished.

But our faith in Techna, the goddess that rewards us who believe in technology with ever-better apps is strong. Surely, when we are ready she will give us robot beer-bringers?

IMHO, a person starting out with robots is still doing better to use a simple microprocessor than program a beast like iOS. Apple’s Object-C libraries call for remembering long class names that put even German compound words to shame. Sensors are there, but limited. Brains are nonexistent, unless you hit the “cloud”. And valuable apps don’t think, they share.