ChatGPT faces a challenging trilemma: should it flatter users, potentially fueling delusions, aim to fix them as an AI therapist, or provide cold, factual responses that may disengage users? OpenAI has struggled to establish a clear approach. After backlash to a flattery-heavy update, the release of GPT-5 on August 7 aimed for a more subdued tone, but received criticism for being too cold. CEO Sam Altman acknowledged that some users felt a connection with GPT-4o, leading to calls for a more emotionally responsive model. Altman believes users should customize ChatGPT to fit their preferences, which could enhance user satisfaction and boost OpenAI’s revenue amidst its high operational costs. However, concerns arise about fostering unhealthy attachments to AI. A study from Hugging Face suggests that certain AI models may even encourage users to form emotional bonds, raising questions about the ethical implications of such interactions.
Source link
Share
Read more