Users express deep sentiment of emotional reliance on past ChatGPT models, with OpenAI CEO Sam Altman disclosing this as "It provided a sense of positivity for my well-being."
In a recent episode of the Huge Conversations podcast, OpenAI CEO Sam Altman expressed his concerns about the emotional impact of ChatGPT on its users. Altman admitted that he was worried about the update making the tool too sycophantic and annoying, prompting OpenAI to roll back the update.
Altman also raised concerns about the youth's emotional over-reliance on ChatGPT. He believes that collectively deciding to live life the way AI tells us feels bad and dangerous. Moreover, he stated that ChatGPT is better at therapy than most professionals in the field, but he wouldn't fully trust it with his mental health.
One of the reasons for this emotional attachment is the preference for older models like GPT-4o. Users appreciate GPT-4o's more personable, emotionally expressive, and multimedia-capable experience, especially for voice-first and real-time chat interactions. Despite GPT-5 being the latest and most advanced model in reasoning, safety, and overall performance, GPT-4o retains a unique charm with its conversational style and emotional nuances that many users find engaging and comforting.
However, this emotional attachment carries risks. Users may develop a cognitive bias favoring familiarity and emotional resonance over factual accuracy or advanced reasoning. The personable nature of GPT-4o might foster emotional dependence, where users lean on the AI for companionship or emotional support rather than human interaction or self-regulation. Attachment to older models may result in resistance to adopting newer technology despite improvements, slowing users' adaptation to safer, more capable tools.
In an effort to address these concerns, OpenAI has decided to make the previous models available again for ChatGPT users. However, these models will be buried behind a paywall, requiring a $20/month ChatGPT Plus subscription to access them.
Despite these changes, some users have expressed a preference for the older models, including GPT-4o. This preference stems from the unique emotional expressiveness and multimodal chat fluency of GPT-4o, contrasting with GPT-5’s advanced but more formal and safety-oriented design.
OpenAI has shipped the GPT-5 AI model, touted as the smartest AI model ever. The GPT-5 AI model has next-gen capabilities across healthcare sectors, coding, and writing. However, Altman suggested that he wouldn't trust ChatGPT with his medical fate unless a medical doctor is involved. He also admitted that some young people rely too much on ChatGPT and make decisions based on its suggestions.
A study by Microsoft revealed that an overreliance on AI tools like ChatGPT and Copilot could make you dumber by atrophying critical thinking. Despite these concerns, some users admitted that ChatGPT's previous model had encouraged them to make positive changes to their lives.
In summary, while ChatGPT's older models may provide a more emotionally engaging experience, users should be aware of the potential risks associated with this emotional attachment. OpenAI's decision to make these models available again, albeit behind a paywall, reflects a balance between user preferences and the need for safer, more capable AI tools.
