*Robot Futures* (and fan out)

This is a very good short book by Illah Reza Nourbakhsh.  Here is one excerpt:

In USAR, the effective number of robots controlled by a single human operator has a formal term: fan out…Ironically, fielded robots have very low fan out scores today.  For instance, the Predator-class drones, unmanned aerial vehicles that fight proxy battles for the United States in distant lands, have a fan out of less than 0.2.  That is, more than five people are required at all times, just to manage a single robot.  In USAR, researchers have begun to demonstrate ever-increasing fan out — exceeding 6.0 — by providing the robots with more and more autonomy so that the human operator is only responsible for the most strategic decisions, with robots making every tactical choice.  Critical to this success is the ability of robots to decide when they need to ask for human help — when they face a survivor, or are stuck in the rubble in a way that the robot cannot extract itself, or when the robot has suffered a serious hardware of software error.  This “intelligent reasoning” for deciding when to ask for help means that one human can manage even more robots to achieve a higher fan out…They do not need true autonomy so much as a willingness to call for help whenever required.  This alleviates the pressure to create perfect robots, and instead good-enough robots can play meaningful roles in a USAR team because humans will bridge the gap between the robot’s capabilities and what the situation demands.

File under “meta-rational robots,” and buy the book here.


Yeah, it's called Starcraft or its predecessor Warcraft not to be confused with WoW.

Robots perform routine functions unless and until commanded to do otherwise.

The game could be easily improved by adding commands that tell the robots to resume a routine function after a specific function is complete, or to develop new routine functions from specialized functions.

One human player can control dozens of robots, most of which have a single function and a few that have dual functions.

The issue with unmanned aerial vehicles is that their use can violate laws or the law of war if they are not monitored by humans responsible for their operations. Even land mines are falling out of favor as lawful weapons of war for precisely that reason. Their usefullness and need are beyond question; but the unintended effects are intolerable.

I'd say a broader issue with unmanned weapons systems in general is that they involve trusting your software with the power to do a lot of damage. That might be done to innocent people in violation of the laws of war, but it might also be done to *your* people, or might carry out a kind of accidental attack that has major political consequences. And if the software can be tampered with, then someone else can intentionally use your unmanned weapons to kill your people or your other unmanned weapons systems, or to target allies to cause the alliance to fall apart, or to massacre civilians to give you a black eye. (That assumes that anyone reports the massacred civilians, which assumes that there's anyone around doing the reporting and that anyone pays attention.)

Sounds like a cliched Tom Clancy novel I just read: Vector Threat.

The most souped-up RTS is open source, and it's been out for over a decade: http://springrts.com/
Game recording (sped up and zoomed out: www.youtube.com/watch?v=JKCEvYNPbeM)

*It's a very diverse playerbase worldwide. I played against at least 1 Russian cosmonaut. :)

Ah, Total Annihilation. Love that game. The original, the one the Spring engine is based on, is available for cheap on Good Old Games, although I bet the Spring community is much larger by now. There's also another open-source engine version of it that focuses on the single-player campaign rather than the multi-player, one called TA3D.

The only serious proposal I ever heard for cutting the F35 involved building more F22s and a new UCAV, such that missions could involve an F22 escorting a group of the UCAVs. The idea was that drone technology really wasn't up to the task of replacing an F35 in general, but that if you could guarantee a human on scene, maybe you could do it.

I believe the correct link, modulo your affiliate id, is


Foolishness! Have we learned nothing from the Butlerian jihad?

It seems to me that one natural endpoint for this increasing fan-out is something like a call-center of humans who deal with problems that the software on the robots can't deal with. When there's a problem, the robot signals for help and gives an urgency level, and one of the humans in the call center deals with it or bumps it up the chain to someone more capable. This strikes me as a plausible future, but not an especially pleasant one. You, personally, are part of an organization that has a fan out of 30, and last week you gave commands to a couple hundred robots, but none of them are yours, and you get subsistence + epsilon per hour and work in a modern reimagining of a sweatshop, effectively a servant of the robots who themselves are servants of rich and powerful people.

An example of high fan-out, with personal robots calling out to Mechanical Turk for help:


This was research I conducted at Willow Garage (a robotics research lab.)

Since you worked at Willow Garage, can you answer a quick question?

Is the PR2 so expensive because the actual marginal cost is close to being that high, or is it mostly because building the PR2 involved a very high fixed cost and there are still relatively few customers?

I tend to think that the answer is yes -- there is very little reason why the marginal cost of a PR2 should be more than an ordinary car, especially since the software (ROS) is free -- but I would appreciate an answer from someone who is on the ground so to speak.

Also the Heaphy software looks cool -- is it going to be the same permissive licence as most of the ROS software?

The Kindle edition is rather expensive, considering it's only 160 pages long.

I too have seen the trailer for Iron Man III.

Comments for this post are closed