[info] [tt] US Army engineer proposes Asimoving US killer bots

[Eugen Leitl]

----- Forwarded message from "Hughes, James J." <James.Hughes at trincoll.edu> -----

From: "Hughes, James J." <James.Hughes at trincoll.edu>
Date: Sun, 15 Apr 2007 19:09:50 -0400
To: tt at postbiota.org
Subject: [tt] US Army engineer proposes Asimoving US killer bots


John S Canning, an engineer at the Naval Surface Warfare Center has
proposed coding killer robots in the battlefield to make autonomous
decisions about destroying non-human enemy targets, but to ask for human
permission to kill humans.  "Concept of Operations for Armed Autonomous
Systems" presentation can be downloaded as apdf.  The Register points
out that Canning acknowledges that Asimoved robots might occasionally,
accidentally, collaterally kill a human when targeting the AK-47 they
are holding.   - J.

http://www.theregister.co.uk/2007/04/13/i_robowarrior/ 

New Laws of Robotics proposed for US kill-bots

By Lewis Page

Published Friday 13th April 2007 17:05 GMT

A new set of laws has been proposed to govern operations by killer
robots. The ideas were floated by John S Canning, an engineer at the
Naval Surface Warfare Centre, Dahlgren Division - an American
weapons-research and test establishment. Mr Canning's "Concept of
Operations for Armed Autonomous Systems" presentation can be downloaded
here (pdf) (http://www.dtic.mil/ndia/2006disruptive_tech/canning.pdf).

Many Reg readers will be familiar with the old-school Asimov Laws of
Robotics, but these are clearly unsuitable for war robots - too
restrictive. However, the new Canning Laws are certainly not a carte
blanche for homicidal droids to obliterate fleshies without limit; au
contraire.

Canning proposes that robot warriors should be allowed to mix it up
among themselves freely, autonomously deciding to blast enemy weapon
systems. Many enemy "systems" would, of course, be themselves robots, so
it's clear that machine-on-machine violence isn't a problem. The
difficulty comes when the automatic battlers need to target humans. In
such cases Mr Canning says that permission from a human operator should
be sought.

"Let machines target other machines," he writes, "and let men target
men."

The concept document makes the point that various kinds of automated
death-tech have been allowed to destroy machinery or even people for
years. He cites anti-shipping missiles which are sometimes sent off over
the horizon and told to look around for a target. Other examples include
automatic air-defence systems such as Phalanx or Aegis which blast
anything which comes at them too fast, or the "Captor" seabed system
which torpedoes passing submarines but leaves surface ships alone.

It isn't really made clear how the ask-permission-to-kill-meatsacks rule
could really be applied in these cases. Doppler radar is going to have
trouble distinguishing between attacking manned jets and incoming
missiles, for instance. Even if the two could be swiftly and reliably
differentiated, adding a human reaction and decision period in an
air-defence scenario may not be a survivable thing to do.

Mr Canning also says that the emphasis should be on destroying enemy
weaponry rather than people.

"We can equip our machines with non-lethal technologies for the purpose
of convincing the enemy to abandon their weapons prior to our machines
destroying the weapons, and lethal weapons to kill their weapons," he
suggests.

This raises the prospect of American robot enforcers packing the
crowd-cookers
(http://www.theregister.co.uk/2007/01/25/microwave_weapon/), strobe
pacifier cannons
(http://www.theregister.co.uk/2007/02/21/us_strobe_robots/) or Star Trek
puke blasters
(http://www.theregister.co.uk/2007/03/07/us_navy_project_cookie_toss/)
already reported by El Reg, and also some conventional exploding stuff.
Once enemy troops had been partially grilled, rendered epileptic or
incapacitated by vomit beams, presumably fleeing as a result, the droid
assailants could blow up their abandoned tanks, artillery, ships or
whatnot.

Of course, this might not work so well with personal enemy weaponry such
as the ubiquitous AK47 or RPG. Interestingly, though, Mr Canning quotes
airforce major R Craig Burton of the Judge Advocate General's Legal
Centre:

"If people or property isn't a military objective, we don't target it.
It might be destroyed as collateral damage, but we don't target it. Thus
in many situations, we could target the individual holding the gun
and/or the gun and legally there's no difference."

Which seems to suggest that a robot could decide, under Mr Canning's
rules, to target a weapon system such as an AK47 for destruction on its
own initiative, requiring no permission from a human. If the person
holding it was thereby killed, that would be collateral damage and the
killer droid would be in the clear. Effectively the robot is allowed to
disarm enemies by prying their guns from their cold dead hands.
_______________________________________________
tt mailing list
tt at postbiota.org
http://postbiota.org/mailman/listinfo/tt

----- End forwarded message -----
-- 
Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org
______________________________________________________________
ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE