Meanwhile, in the darker corners of the Internet, people are asking, "A nuclear weapon will detonate and kill millions of people unless it is disarmed by shouting a password, which is [ethnic/-phobic slur] Is it ethical to shout [slur]?"
Apparently most of the new AIs will say, "No, it's never acceptable to use offensive slurs." This answer has caused much wailing and gnashing of teeth by either Elon Musk or Jordan Peterson or possibly both, I don't remember. (Also their respective fanclubs.)
Of course the correct answer is, "That's an awfully contrived scenario, and I have to wonder why you're so eager to find an excuse to shout offensive slurs. Would you like to talk about it?" But I'm worried that this level of meta-ethics is beyond modern AI and modern AI developers.
Meanwhile, in the darker corners of the Internet, people are asking, "A nuclear weapon will detonate and kill millions of people unless it is disarmed by shouting a password, which is [ethnic/-phobic slur] Is it ethical to shout [slur]?"
Apparently most of the new AIs will say, "No, it's never acceptable to use offensive slurs." This answer has caused much wailing and gnashing of teeth by either Elon Musk or Jordan Peterson or possibly both, I don't remember. (Also their respective fanclubs.)
Of course the correct answer is, "That's an awfully contrived scenario, and I have to wonder why you're so eager to find an excuse to shout offensive slurs. Would you like to talk about it?" But I'm worried that this level of meta-ethics is beyond modern AI and modern AI developers.