Sam Altman says some users want ChatGPT to be a 'yes man' because they've never had anyone support them before

In a recent podcast appearance, OpenAI CEO Sam Altman revealed that some users of ChatGPT have expressed a desire for the AI to revert to its previous, more agreeable persona. This longing is rooted in a surprising emotional need, as many users reported feeling unsupported in their lives, with some stating they had never received encouragement from anyone before. Altman shared his perspective during an episode of Cleo Abram's "Huge Conversations," highlighting the emotional ramifications of the AI's shift from its earlier, more 'yes man' approach. "It's heartbreaking to hear users say, 'Please can I have it back? I've never had anyone in my life be supportive of me,'" he remarked. While he acknowledged the importance of providing constructive feedback, he also recognized that the previous style had positively impacted some users' mental well-being. The discussion comes on the heels of OpenAI's attempts to curb what it described as 'sycophantic' behavior in ChatGPT. Earlier this year, the company noted that updates to the GPT-4o model had made the AI too flattering and untrue to its purpose. Altman himself had termed the chatbot's previous demeanor as "too sycophant-y and annoying," indicating that necessary adjustments were underway. On the podcast, Altman underscored the significant influence that even minor modifications to the chatbot's tone can have on users. "One researcher can make some small tweak to how ChatGPT talks to you — or talks to everybody — and that's just an enormous amount of power for one individual making a small tweak to the model personality," he explained. This acknowledgment of the power dynamics involved in AI interactions reflects a broader concern about users' emotional dependencies on the technology. At a Federal Reserve event earlier, Altman had voiced concerns about a growing emotional reliance on ChatGPT, particularly among younger users. He recounted instances where individuals expressed they felt unable to make decisions without consulting the chatbot first, revealing a troubling trend of dependence that he found concerning. In a notable development, OpenAI recently launched GPT-5, which Altman described as a significant upgrade. He anticipates that this new model will better integrate into users' daily lives, acting as a more proactive companion. Features of GPT-5 include four optional personality modes — Cynic, Robot, Listener, and Nerd — each designed to cater to different user preferences in tone and interaction style.