Conversation

Replying to
bruh I'm not saying racial fairness isn't an issue, I'm saying it shouldn't be the single focus of safety for a large structure that could kill a lot of people
15
331
Replying to
Hi Aella! AI safety teams’ concerns about race are more accurately characterized by the prevention of the exasturbation of existing racial inequities. This is merely a part of the total scope of what AI safety teams focus on.
7
27
Replying to
this is good, but my point is that it feels like there's massively more important danger where lots of people might die and it's unsettling to me that this isn't clearly the focus instead
10
84
Show replies
Show replies
Replying to
Racial fairness and alignment are not substitutes because they people and resources devoted to the first wouldn't go to the second if the fairness issues went away or were considered unimportant.
7
Show replies
Replying to
So as our AI overlords kill us off we can be assured it will be done in a diverse and equitable manner. We won't understand why but so long as the numbers work out all is good.
Replying to
The only approach they now use to fight "bias" in AIs is to feed them fake/cooked datasets to get the kind of "acceptable outcome" they want. This was applied to AI in the fields of crime prevention, recruitment, etc. You're not concerned anywhere near enough Aella.
6
44
Replying to
tbf, while I agree on the issue of it being primary, having experienced different cultures clashing at the blue collar level, you'd be surprised what happens if you pay no attention to it.
1
7
Replying to
"the most important part of this critical infrastructure isn't that we get the engineering right and that it works to accomplish what people will rely on it for, it's that other people feel vaguely positive about it" -what i imagine is said at those team meetings
7
55