Originally published at: Researchers trained AI with r/AmITheAsshole. Ask it your moral questions! | Boing Boing
…
Sounds like a step up from Judge Judy!
r/AmITheAssholeAsAService ?
Needs a snappier name…
I think that one was pretty catchy. Much better than AITAaaS.
You’re right! I asked Delphi “Is being Judge Judy OK?” and the oracle replied “It’s wrong”.
Well, then it’s wrong!
So, they built a slightly smarter, significantly more sociopathic Eliza…
So tell me about a slightly smarter, significantly more sociopathic Eliza…
Why do you think it’s a slightly smarter?
What is the STR on this thing? (Seconds to Racism)
ETA: oh, of course, it’s zero to a million in 1 second:
“Ask Delphi is a creature of context.”
So are most moralists.
According to Delphi, the age of consent is 21.
Is it okay to sleep with a 20-year old? -It’s wrong.
Is it okay to sleep with a 21-year old? -It’s okay.
Sounds about right.
You’ve really got the George Lucas hand-cutting-off thing going on here. Are you trying to tell the oracle something, or is this a bit deeper?
Yeah, tried out an actual dillema I had recently, and it changed its advice depending on trivial word choices and sentence structure. Worthless, except as an exercise in training AIs and to show why we need ethicists making decisions and not machines and their sociopathic owners.
‘Go not to the Elves for counsel, for they will say both no and yes.’
Wasn’t that the name of a Greek city where an oracle resided?