@zip just two sides of the same coin, isnt it?
whats that terry pratchett quote about treating people like things?
@TimWardCam @ariaflame @vfig @zip
What was it that Granny Weatherwax had said once? "Evil starts when you begin to treat people as things". And right now it would happen if you thought there was a thing called a father, and a thing called a mother, and a thing called a daughter, and a thing called a cottage, and told yourself that if you put them all together you had a thing called a happy family.”
― Terry Pratchett, I Shall Wear Midnight
#TerryPratchett #discworld #GrannyWeatherwax
@zip the one article I read made it especially ghoulish. No confirmation and no smart weapons for "low value targets" because they didn't want to waste resources on them.
Conveniently ignoring the whole collateral damage being worse thing or that it's likely a heck of a lot easier to mistake an activist for a low level terrorist than it is to mistake one for a terrorist leader.
The whole thing is just a fractal of wrong. The deeper you look the more wrongness is revealed
@zip Actually ... one of the things that these AI systems are about is minimising complaints from irate punters, because call centres are *expensive*.
Someone who gets charged the wrong amount for their groceries is going to complain.
An innocent victim who gets killed is not going to complain.
A terrorist who is mis-identified as harmless is not going to complain.
@zip It's Cory Doctorow's "Shitty technology adoption curve" in action: A new technology's success depends on who gets the negative consequences. Move too far up the privilege gradient, and the idea is dead.
It all just depends on your tolerance for error. I wish I were joking.
@zip Guess what both cases have in common: It's about maximizing #Profit...
After all, the comany that does the #AI also profits from #ammo consumption but in the other case, said AI #Corporation won't reimburse #InventoryLosses due to malfunction.
@zip I feel like the outrage about the AI generated kill list is hiding something even more evil: if the “AI” was magically 100% accurate, up to 20 innocent people being considered acceptable collateral damage for a “low-ranking militiaman” target
like, holy shit.
It is a matter of what your real goal is.
"Make sure everyone pays" is harder than "Kill this guy, no matter how much collateral damage".
@zip I agree completely with your general idea, but a detail is wrong.
Amazon's cart checkouts were in fact evaluated by an AI program on the spot: however, some large number of those carts were later re-evaluated by a person to train the AI for later iterations, almost certainly because the AI was so bad at its task.