every new story like this convinces me more that the cat is entirely, irrecoverably out of the bag with these systems. no amount of guardrails or consulting with experts will ever fix this. they built a product designed to make people reliant on it emotionally/mentally, and that's the original sin

Hi, I don't know you, but an AI is not a sentient being tat "knows" anything. It's trivial to bypass these guardrails for any use-case, including creating misinformation, to the point that the chatbot will pretty much explicitly tell you how to do it and then output the exact same thing. design flaw

often this "behavior" is framed as an organic result of training data or limitations in the technology itself, but you've got to imagine the product meetings and memos that went around too

i have been AWOL on social so this is, maybe, possibly, me also saying i might just post more. perhaps