Robots That Jump

Robot Bodies Needed Before Robot Minds

Some people understand limits

A nice article in the NY Times today, on the DARPA Atlas robot. The author emphasizes the gap between fantasy and reality.

A few points the article makes:

The robot has to be tethered, since a slight mistake (e.g. a fall) would destroy it

The robot requires huge amounts of energy to operate, impossible to supply via current battery technology

The robot takes all day to move a couple of boards

Now, this is not an attack on the designers and programmers. Instead, it is a tribute to just how hard it is to make any robot do anything at all. Kudos to John Markoff, senior writer for The New York Times for getting it right.

Another kind of limit is discussed by the following article:

Here, the “killer robots” are mostly mis-interpreted as self-actuated “terminators” – when in face, they are almost all tele-operated drones. In short, the article, and many of those commenting in the article, act as if the problem was robots, rather than the people puppeting said metal puppets.

The final link is to a practical robot, primped to act more like an autonomous robot of imagination. After all, automatic milking machines have been around for many decades. Like the Mars rovers, there has been a deliberate attempt to make the machine seem more humanoid than it actually is. Sigh.

But, OMG what if these robotic machines get loose. What if they don’t just milk cows, but EVERYONE? Mayhaps we need a worldwide ban on autonomous milk-bots…before it’s too late!

This seems to be the theme of (yet another) Wired magazine article on how the robots will take over…

Silly, because Marshall Brain discussed this in his “Robotic Nation” articles a decade ago:


One response to “Some people understand limits

  1. John Nagle, Silicon Valley, CA November 2, 2013 at 7:57 pm

    What Markoff saw was one of the university projects controlling the Atlas robot. They’ve only had a real robot for about two months, and previously they were developing against a simulator with an unrealistic physics engine. (The foot/ground coefficient of friction has to be cranked up to 10 to get it to work, I hear.) So the motion control is awful. Boston Dynamics has better motion control in-house, but the entrants in the DARPA Robotics Challenge don’t have access to that code. (At least not officially). The entrants just get a dumb subroutine library (it’s a binary-only .so file) which can do simple walking and can stand still to allow arm usage.

    By this time next year, there should be a physics simulator that doesn’t suck. Then the machine learning people can get into the act and start making the motion algorithms self-tuning. There’s also a project underway to build a small Diesel powerplant for the Legged Squad Support System robot which will be quieter and more efficient than the present fixed-speed go-kart engine. If that, or a variant of it, can be crammed into Atlas, they’ll have a self-contained robot.

    DARPA has poured about $120 million into Boston Dynamics so far, and it’s working. One thing we learned from the first DARPA Grand Challenge in 2004-2007 was that throwing money at robotics, combined with fear of failure, works. Before that, the typical robotics project was one professor and 3 to 5 grad students. It took years to get anything done. Then DARPA put the screws on academia to get results or have their funding turned off, and entire CS departments were focused on automatic driving. In 18 months, success.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: