Disturbing NYT article on robot soldiers

http://nytimes.com/2005/02/16/technology/16robots.html

The following is a rant.

i was struck by how optimistic the guys leading the research are. 10
years until groups of robots can autonomously travel inside “deep
woods” or “dense cities”? 25 years until we have one that “looks,
thinks, and fights like a soldier”? 1/3 of
deep-strike aircraft robotic in a decade? Yeah, right. I say 10 years
until a single autonomous prototype can quickly and
reliably navigate an empty desert with realistic “obstacles” like
hills and ditches.

Now, drones, i.e. guns and a camera on a remote-controlled truck, that
seems more realistic.

And I can guarantee you that a robot that can reliably distinguish an enemy
from a nearby civilian (the “tank or schoolbus” issue) is a very, very
long way off, especially since, after robots are on the battlefield,
the enemy will be actively changing their behavior and appearance to
try and fool the robots into thinking that they are civilians. I bet
we won’t see more than 95% accuracy in the next 30 years, even without
enemy attempts at confusing the robots (which means,
given an equal number of civilians and enemies nearby,
1 civilian target is killed for every 20 enemy targets; hardly an
acceptable ratio). This problem seems “AI-complete” to me. Hope they don’t unleash these
things near any civilians…

2 Responses to “Disturbing NYT article on robot soldiers”

  1. JC Says:

    ‘AI-complete’ needs to be a quotable term if it isn’t already (is it?).

  2. Neville Says:

    Looks like the NYT is really into this automated warfare topic:
    http://www.nytimes.com/2005/02/20/weekinreview/20john.html

    Here’s the relevant section:

    The Pentagon’s promotional material doesn’t mention whether the
    contract for the Future Combat Systems, said to be the biggest in
    American military history, includes a line item for philosophers. But
    they may be best equipped to judge whether computers, despite their
    faster speeds, greater bandwidth and bigger memories, are inherently
    different - less trustworthy than the gut feelings and hunches of their
    keepers.

    John Searle, the philosopher at the University of California at
    Berkeley whose most recent book, “Mind: A Brief Introduction,” came out
    last year, has argued for decades that the brain is not just a computer
    strung together from neurons. Whatever is happening in the head - and
    nobody really knows - it is not computation. Confuse reason with
    calculation, he argues, and disaster lies ahead.

    But that has become the minority view. Attend a conference of the
    Society for Neuroscience or the Cognitive Science Society and you would
    be hard-pressed to find anyone who doesn’t assume deep down that the
    brain is some kind of information processor. For the time-being, people
    excel at certain tasks, like recognizing faces and making sense of
    ambiguous data, but that may be only because of wiring details and
    variations in the algorithms - things that eventually could be
    simulated electronically. The result would be machinery that can do
    anything a person can, but faster and better.

    I think it’s an unfortunate situation that the popular media is setting up this artificial incompatibility between the idea of mind as computation and “gut feelings.” Of course, to a neuroscientist, gut feelings are just the result of some very fancy computations…

Leave a Reply