So we never learned if the extermination of humans was a bug or a feature?

So we never learned if the extermination of humans was a bug or a feature?

Depends on who the target user is, surely?

Man, I wish I'd known about Roombas early last year when I was spending like wildfire on the new place. I might actually have presentable floors today.

Its wrong for me to want a Scooba in an apartment as small as mine. I realize this.

But the white linoleum is a PITA to keep clean. If I had a little robot going every day...

How To Get A Human Being On The Phone

Cheat sheet guide to many common automated phone menus (banks, stores, utilities, etc).

Handy and on-topic!

Fay's reaction to the news about Sharon reminded me that the new Star Wars novel is titled: Star Wars: Dark Lord.

DARTH VOLDEMORT.

Personally (and no offence) I think our current best chances for a rogue AI will either come out of google projects mating in the dark, or all the Linux distros banding together.

I think that some benign AI program will never go rogue. If we get rogue AI it'll be a result of an espionage program going out of control, or some sort of cyber-warfare system running amok during a cyber-war.

('cyber-war' is so dated. What's the current term for computer warefare (hacking into networks as a military operation)?)

I think that some benign AI program will never go rogue

Well, maybe not rogue in the moustache-twirling sort of way (and are you calling Google and Linux benign? Hmmm) but there's no reason for our priorities to be machine priorities. Worried about pollution? Kill most of the humans.

An AI with independent learning capabilities could easily learn a value system that's not actively hostile to humans, but still problematic. Like, TrafficAI decides one day that red lights don't really do anyone any good, and it's not going to use them anymore. Or GoogleAI and YahooAI get into a long drawn-out argument, and nobody can look anything up online for hours. And suchlike.

but there's no reason for our priorities to be machine priorities. Worried about pollution? Kill most of the humans.

OK, I'm venturing further out into conjecture-land, but I think that by the time we get to autonomous commercial AI programs (for example, one that processes insurance claims) that these sort of straightforward AI applications will be well-understood, with clearly-defined limits. For example, an AI that's allowed to access the internet would be forced to abide by correct internet protocols, and would be prohibited from hacking into other computers. And I think we'd be able to understand such programs well enough that we could be sure it was impossible for such a thing to even occur to the AI.

OK, maybe that's rather optimistic of me. But like I said before, an AI that's designed to infiltrate, attack, strategize, take over systems, etc. - I'd be afraid of those. Plus they'd tend to be top-secret, with perhaps little oversight....