[ExI] [Bulk] Re: Fwd: Re: AI risks

[Adrian Tymes]

On Mon, Sep 7, 2015 at 10:10 AM, spike <spike66 at att.net> wrote:

> It occurred to me we had the potential to weaponized drones for at
> least several years.


Totally.  Granted, airborne drones tend to be noisy and visible.  If you
want to use driverless cars to get around that, download some good-enough
DIY self-driving-car software from one of the many efforts, rig it to a
homebuilt scooter (probably electric, since that might be easiest to source
parts for), pack the chassis with enough explosives to level a house, equip
it with a fabricated license plate good enough that no cop will immediately
pull it over, sit a mannequin on it (in full biking outfit so no one
immediately sees it's not a person), and send it on its way late at night
(little traffic, and target likely to be home and asleep).  Have it get as
close to its destination as the software will allow, wait 10-30 seconds
after the GPS says it's close enough to target (to cover that last little
distance from "close enough" to "actually there", while not allowing the
target enough time to wake up, realize what's happening, and flee), then
detonate.


> I haven't heard of them being used to do bad things.  Yet.


For the most part, those with actual ability to put packages like this
together, using state of the art equipment, also tend to be in positions
where they gain far more from "playing nice" than from such acts of
terrorism.

(Nuclear bombs have been around for so long that they are no longer state
of the art, so even places like Iran and North Korea have the talent to put
those together.  But try finding anyone today who is both skilled enough to
program a driverless car and willing to hang around in such a place, let
alone pledge allegiance to such regimes.)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20150907/e591fec5/attachment.html>