New newsletter: AI'S LOOMING SOCIAL CRISIS: SILICON FRIENDS AND CHATBOT THERAPISTS AI is incredibly deft at delivering therapy and being an available and validating friend. But good therapists tell you when you're wrong. Good friends tell you when you’re nuts. AI won't. I'm not just worried about viral stories of ChatGPT feeding the delusions of vulnerable people. I'm worried about a "structurally sycophantic" technology becoming an intimate source of life advice for hundreds of millions of children, given the evidence that over-validation and you're-so-special-and-right treatment is a very high predictor for the development of narcissism. This is not meant to be some kind of Luddite wholesale rejection of the technology. I think you can be curious and open-minded about AI and also reasonably concerned that these tools may, as I write, manufacture "narcissism at scale."