Why AI’s biggest threat is actually its power to seduce us — and the computers are already winning: Prof

Researchers at Cornell University found chatbots and AI models are all overwhelmingly programmed to suck up to users. “We find that models are highly sycophantic: they affirm users’ actions 50% more than humans do. “Participants [in the study] rated sycophantic responses as higher quality, trusted the sycophantic AI model more, and were more willing to use it again. This suggests that people are drawn to AI that unquestioningly validate, even as that validation risks eroding their judgment. “These preferences create perverse incentives both for people to increasingly rely on sycophantic AI models and for AI model training to favor sycophancy,”...