Thanks for the article! I think it's even worse actually, because this is not something that needs to be done explicitly or on purpose. Simply training AIs wi

Thanks for the article! I think it's even worse actually, because this is not something that needs to be done explicitly or on purpose. Simply training AIs with a target goal of maximizing engagement could lead to models discovering and exploiting superstimuli or superstimuli-like bias in humans.