> Analyzing "emotion" in the model is completely anthropocentric. Yeah, asking a text generator designed to sound as-human-as-possible about its "welfare" then

> Analyzing "emotion" in the model is completely anthropocentric. Yeah, asking a text generator designed to sound as-human-as-possible about its "welfare" then actually giving credence to the output is a category error. It's like asking a ceramic mug with "Best Dad!" written on the side if I'm the best dad, then uncritically just believing the words painted there. :( :( :(