2 ms·
I agree, but there's a difference between a human and an AI making that joke. A human who makes clever jokes poking fun at bigoted expressions will probably con
by flangola7 4y ago
I agree, but there's a difference between a human and an AI making that joke. A human who makes clever jokes poking fun at bigoted expressions will probably continue to do so. A language model is liable to forget to be clever 30 minutes later and say something bigoted without a punchline. Obviously Twitch would like to prevent this.
- Rebelgecko 4y agoI've been thinking about this a bit. Some people believe that depictions of x behavior are supporting x. Not a super common belief but I've seen more people say it with a straight face over time (Eg a movie with a character who says racist things is inherently racist. See the IASIP episode that was removed from streaming). Which is patently ridiculous... Schindler's List depicts some of the worst acts of antisemitism ever, but the movie itself is obviosly not antisemitic. So what's the difference between Schindler's List and an actually antisemitic movie, like Hebrews To Negroes? I think it comes down to the creator's intent. In Schindler's List, you're not supposed to agree with the antisemitic stuff that happens. It's bad and wrong. In Hebrews To Negroes, the antisemitisn is depicted as good and correct. When you have a large language model creating your art, how do you make that determination of "intent"?