Ex-OpenAI CEO and power users sound alarm over AI sycophancy and flattery of users

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More An AI assistant that unequivocally agrees with everything you say and supports you — even your most outlandish and obviously false, misguided or straight-up bad ideas — sounds like something out of a cautionary sci-fi short story from Philip K. Dick. But it appears to be the reality for a number of users of OpenAI’s hit chatbot ChatGPT, specifically for interactions with the underlying GPT-4o large language multimodal model (OpenAI also offers ChatGPT users six other underlying LLMs to choose between to power the chatbot’s responses, each with varying capabilities and digital “personality traits” — o3, o4-mini, o4-mini-high, GPT-4.5, GPT-4o mini and GPT-4). Over the last few days, users including former OpenAI CEO Emmett Shear — who ran the company for only 72 hours during the Sam Altman firing fracas of November 2023 — and Hugging Face CEO Clement Delangue ... Read full article.