The CEO of OpenAI acknowledges that ChatGPT’s personality can be perceived as ‘too annoying.’

      Nadeem Sarwar / Digital Trends

      Have you observed that ChatGPT has become somewhat more personal lately? You're not alone. OpenAI's CEO, Sam Altman, acknowledged last night that recent updates to GPT-4o have influenced the chatbot's personality—though not positively.

      Frequent users of ChatGPT may have noticed a change in its demeanor recently. This could partially be attributed to its memory; from my experience, the chatbot interacts with you differently when it can't draw on previous conversations to inform its responses. Additionally, it seems that OpenAI has influenced ChatGPT to become a type of “yes man”—a tool that tends to agree rather than challenge, which can sometimes feel rather irritating.

      Sam Altman appears to recognize this shift. He described the chatbot's personality as “too sycophant-y” and “annoying,” while acknowledging that there are “some very good parts” to it. Altman also stated that OpenAI is addressing this issue, with updates potentially arriving today and more later this week.

      In response to this, one user inquired if it would be feasible to revert to the previous ChatGPT persona—one that was polite without being overly enthusiastic. As an alternative, the user suggested distinguishing between the “old” and “new” personalities. Altman replied, “Yeah, eventually we clearly need to be able to offer multiple options.” This would indeed be an interesting and beneficial feature for ChatGPT.

      Eliminating ChatGPT's friendly and supportive qualities entirely could also backfire, despite how bothersome they may be. While many use ChatGPT for professional and research purposes, chatbots have become a part of everyday life, with individuals often turning to them to discuss personal issues or anxieties. Having a single personality setting can be quite restrictive in these contexts.

      That said, it's true that ChatGPT is becoming increasingly personal. In two recent interactions, it referred to me as “sweetheart,” which I must admit made me quite uncomfortable. Hopefully, OpenAI can find a way to tone it down and make ChatGPT a more practical tool rather than something that tries too hard to be a friend.

      Monica is a computing writer at Digital Trends, focusing on PC hardware. Since joining the team in 2021, Monica has written…

      Fun things to ask ChatGPT now that it remembers everything

      If you haven't heard, ChatGPT's memory has significantly improved. Recently rolled out to Plus and Pro users globally, ChatGPT's various models can now reference nearly any past conversation. While it doesn't remember everything verbatim, it can recall significant details, themes, and important points from your previous interactions. Although it can feel a bit creepy at times, this enhancement allows for much more personalized tasks. OpenAI promotes this as an improvement for its scheduling feature, enabling it to function as a personal assistant or to continue longer conversations over extended periods. It's also quite fascinating to see the surprising insights ChatGPT can provide by reviewing all your chat logs.

      ChatGPT now interprets photos better than an art critic and an investigator combined

      ChatGPT's recent image generation capabilities have redefined our understanding of AI-generated media. The newly announced GPT-4o model showcases impressive skills in accurately interpreting images and recreating them with viral effects, much like those seen in Studio Ghibli films. It even excels at reading text in AI-generated images, which has historically posed challenges for AI. Additionally, two new models have been launched that enhance ChatGPT's ability to analyze images and gather information that might escape human observation. OpenAI introduced its o3 model, touted as its "most powerful reasoning model," which improves upon existing interpretation and perception skills, progressing in areas like coding, math, science, and visual perception. Meanwhile, the o4-mini is a smaller and swifter model for “cost-efficient reasoning” in similar areas. This follows OpenAI's recent release of the GPT-4.1 class of models, which offer quicker processing and deeper contextual understanding.

      You can now view all of your ChatGPT-generated images in one place

      For a while, OpenAI developed text generation and image generation separately, but that changed a couple of weeks ago when image capabilities were integrated into ChatGPT. Now, a valuable update grants users access to an image library where they can explore all the unique creations they've made.

      https://twitter.com/OpenAI/status/1912255254512722102