When Sam Altman announced an April 25 ChatGPT-4o update, he promised it would improve "both intelligence and personality." He noted that previous updates "made the personality too sycophant-y and annoying," and announced fixes "asap." Now two weeks on, there's little evidence that anything got fixed at all, as ChatGPT's brown nosing is reaching levels of flattery that border on dangerous.   AI flatters in less cheery ways too. As Seattle artist Giorgio discovered, bots like ChatGPT will do anything to please their human masters — even validate a user's paranoid delusions during a psychological crisis. Simulating a paranoid break from […]

---

Link :
https://futurism.com/sycophancy-chatbots-ai-problem