AI safety teams that are primarily concerned with stuff like racial fairness make me feel like I would if skyscraper structural safety teams were concerned primarily with racial fairness - really fucking unsettled
98
145
1,647
Conversation
This Tweet was deleted by the Tweet author. Learn more
She's afraid AI is going to turn the whole universe into paperclips, and you're worrying because it's better at recognising some people than others.
12
74
One is a hypothetical thought experiment, and the other is an actually demonstrated problem in the real world 🤷♂️
4
31
I'd be down if there was a separate skyscraper team that was focused on making sure there was no racial bias in idk who can enter the building, but I wouldn't want it to be the primary focus of a the single skyscraper safety team
The harm of an AI injecting racial bias (or any other bias) into potential use cases like credit/loans, criminal justice, medical recommendations, etc. seems a lot more tangible/likely to me than paperclips or basilisks.
16
OK, see, the skyscraper thing is a hypothetical you made up.