OpenAI makes GPT-5 'warmer and friendlier' after rocky launch
OpenAI has announced a significant update to its flagship large language model, GPT-5, aiming to make the AI “warmer and friendlier” for users. This move comes on the heels of a somewhat turbulent launch for GPT-5, which, as CEO Sam Altman candidly admitted, was “a little more bumpy than we’d hoped for.” The company faced user feedback indicating a preference for its predecessor, GPT-4o, suggesting that the latest iteration, while powerful, lacked a certain affability.
The update, rolled out late last week, seeks to address these concerns by subtly adjusting GPT-5’s conversational style. OpenAI describes these changes as minor yet impactful, designed to make the model “more approachable.” The company clarified in a social media post that these aren’t superficial attempts at flattery but rather “small, genuine touches like ‘Good question’ or ‘Great start.’” Crucially, internal tests reportedly show no increase in sycophancy compared to GPT-5’s initial personality, aiming for authentic helpfulness over artificial pleasantries.
The challenges surrounding GPT-5’s debut have been a persistent topic within the AI community. At a recent dinner with journalists, OpenAI executives reportedly attempted to steer conversations toward the company’s future endeavors beyond GPT-5. However, the model’s rocky launch loomed large, becoming, as one reporter noted, the “elephant in the room.” Nick Turley, an OpenAI Vice President, acknowledged that the initial GPT-5 was “just very to the point.” This directness, while efficient for certain tasks, may have contributed to the perception of it being cold or less engaging for users accustomed to more conversational AI interactions. The newly announced update, Turley confirmed, is intended to imbue the model with a more welcoming feel.
This strategic shift underscores a growing realization among AI developers that raw computational power and factual accuracy are not the sole determinants of user satisfaction. The “personality” of an AI, its tone, and its ability to engage in natural, empathetic dialogue are becoming increasingly vital. As AI models integrate further into daily life, from customer service to creative collaboration, the nuances of their interaction style can significantly impact user adoption and trust. OpenAI’s decision to refine GPT-5’s demeanor highlights the delicate balance between efficiency and affability, suggesting that even the most advanced AI must learn to speak not just intelligently, but also humanely. This continuous refinement process reflects the dynamic nature of AI development, where user experience and emotional connection are proving to be as critical as technical prowess.