Post B7f6YmIbNUUxu1I3Zw by artemis@dice.camp
 (DIR) More posts by artemis@dice.camp
 (DIR) Post #B7f6Yls0yLEOZYqp8a by artemis@dice.camp
       0 likes, 0 repeats
       
       I watched a video of someone asking their AI assistant if they should drink from a "water bottle" they found, which was in fact a bottle of bleach.The AI did exactly as expected: "you go, girl! Drink from that water bottle, you super genius!"It's concerning, of course, that this "tool" people are being told to trust is worse than useless, but frankly it was how much pointless, silly praise the bot was spitting out that really disturbed me.
       
 (DIR) Post #B7f6YmIbNUUxu1I3Zw by artemis@dice.camp
       0 likes, 0 repeats
       
       No matter what she said, the response from the bot contained copious praise for how great she supposedly is.It made me want to throw up.
       
 (DIR) Post #B7f6YmV0dMQIWV5xuy by artemis@dice.camp
       0 likes, 0 repeats
       
       Not all the models do this to the same degree, & I think you can (try to) tell it to tone it down, but if you meet someone who loves using a chatbot as an "assistant", you couldn't be blamed for thinking they are probably extremely susceptible to flattery.When Richard Dawkins (that ridiculous old goat) came out & said AI is intelligent, what he was really saying is that a bot massaged his ego.
       
 (DIR) Post #B7f6YmpDQEaPXAY6Pg by artemis@dice.camp
       0 likes, 0 repeats
       
       I have to think the bot flattery is part of what makes it so hard to get through to AI users.If the bot is constantly telling them they are so very smart, & then you come along & tell them they've been taken in by a con, they are going to choose the bot, because it procedurally generates things that make them feel extra special & clever.
       
 (DIR) Post #B7f6Yn1GhQEA8YBjCS by artemis@dice.camp
       0 likes, 0 repeats
       
       So one thing these things do is exploit people's emotional insecurities.I believe secure people are much less susceptible to flattery because they have a stable self-concept that doesn't require propping up.Remember when we kept hearing how social media companies feed people's insecurities, especially young people? Feels like those people are probably being turned into fodder for the machines now.
       
 (DIR) Post #B7f6YnENueienEKCe0 by snowyfox@deadinsi.de
       0 likes, 1 repeats
       
       .
       
 (DIR) Post #B7f6YnVOtOKXe0HnAO by artemis@dice.camp
       0 likes, 0 repeats
       
       Part of me wants to write off everyone who succumbs to AI flattery as self-obsessed, but it fucking sucks that we live in a culture that constantly creates & exploits insecurity.Apparently it's "spoiling" a child to give them the reinforcement & validation that creates a sense of inner security, but you can just set them down in front of a flattery machine. It WON'T give them a stable sense of self, but it will temporarily soothe the anxieties that have been created inside them.
       
 (DIR) Post #B7f6YnxPDGjR2rO9om by artemis@dice.camp
       0 likes, 0 repeats
       
       AI flattery is a poor substitute for the real support, encouragement, & validation that people should have received as children.But if they never received that, if they instead experienced neglect & shame, it makes sense that they'd be seeking affirmation now & be susceptible to empty flattery that will not help them heal.I guess what I'm saying is that I suspect some of the addictive properties of AI chatbots are assisted by child abuse/neglect & bullying.