So when will we see the rise of Skynet?

Al Fin has a very interesting post about autonomous, armed micro drone assassins:

The size of these tiny killer-drones is quickly shrinking, so that it will not be long before lethal drone packages can be delivered under stealth in the form of birds, bats, or insects -- right in the lap of an enemy base or command and control centre.

It has become clear to militaries that remote controlled drones are subject to being hacked, hijacked, or having their communications jammed. Consequently, there is a rush to develop semi-autonomous or fully-autonomous drones capable of carrying out a mission after communications is lost.

This is, of course, a horrifyingly Bad Idea.  After all, what could possibly go wrong?  I have predictions:

1. The people who design this will understand what they're doing less than they think they do.  More specifically, the designers will be very explicit about what they know and what they don't know about the Artificial Intelligence (AI).  These warnings and caveats will be lost as the information
deteriorates
is briefed upwards in the command chain.

2.  The DoD people running this system will think that they are smarter than they are (see #1 above) and so will defer to the judgement of politicians as to when this should be deployed.  The Pols, of course, will see this as the Easy Way Out of making decisions and love the Plausible Deniability of the whole thing - if it goes bad, someone down the chain in the DoD will take the fall.

3. The people down the chain in the DoD will not understand this seductive lure to the politicians.  Of course, ex-politicians employed as lobbyists by the drone manufacturers will get filthy stinking rich.

4. When this goes bad - as everything eventually does go bad - it won't be the ex-politician lobbyists or their employers who take the fall.  Not will it be the current politicians, who of course never had a say in this decision But Who Are Absolutely Horrified And Who Will Not Rest Until The Get To The Bottom Of This.

5. This of course will be deployed domestically, despite any (worthless) assurances by Congress.

6. When this goes bad - as everything eventually does go bad -American citizens will be killed graveyard dead by bad AI.

I'm afraid that I'm pretty cynical about this whole situation.  Which leads to another prediction:

7. Borepatch will not ever get any of that sweet, sweet government funding for his security analysis.

Just sayin'.