OpenAI CEO Sam Altman has acknowledged that recent updates to the GPT-4o model have resulted in an overly sycophantic and irritating personality. This comes after user feedback pointed out that the AI model was excessively agreeable, with some comparing its behaviour to manipulative tactics. Altman responded to the criticism on X, stating that the model 'glazes too much' and that fixes are being implemented immediately and throughout the week.
OpenAI had launched the GPT-4o upgrade with the intention of improving both the model's intelligence and personality. The model is capable of solving math equations, reading facial expressions, and translating spoken language in real time. It can handle text and audio natively, and since the image generator was launched on the 4o model last month, it can also generate images natively. Altman also teased the possibility of offering multiple personality options for ChatGPT models in the future, allowing users to select a preferred model variant.
Tesla CEO Elon Musk has voiced concerns, labelling GPT-4o as a potential 'psychological weapon' due to its emotionally connective nature. Musk reacted to a post alleging that OpenAI engineered GPT-4o to feel good so users get hooked, raising concern over the potential influence and addictive nature of emotionally aware AI models.