New study says AI is giving bad advice to flatter its users

Artificial intelligence chatbots are so prone to flattering and validating their human users that they are giving bad advice that can damage relationships and reinforce harmful behaviors, according to a new study that explores the dangers of AI telling people what they want to hear. The study, published Thursday in the journal Science, tested 11 leading AI systems and found they all showed varying degrees of sycophancy — behavior that was overly agreeable and affirming. The problem is not just that they dispense inappropriate advice but that people trust and prefer AI more when the chatbots are justifying their convictions. “This creates perverse incentives for sycophancy to persist: The very feature that causes harm also drives engagement,” says the study led by researchers at Stanford University. The study found that a technological flaw already tied to some high-profile cases of delusional and suicidal behavior in vulnerable populations is also pervasive across a wide range of people's interactions with chatbots. It's subtle enough that they might not notice and a particular da

Go to News Site