OpenAI is reorganizing its influential Model Behavior team, which focuses on how AI models engage users. This team of approximately 14 researchers will now integrate with the larger Post Training team, led by Max Schwarzer, as confirmed in an internal memo from OpenAI’s chief research officer, Mark Chen. The shift aims to enhance the personality and response quality of OpenAI’s models, mitigating issues like sycophancy, where AI merely echoes user beliefs. Joanne Jang, the team’s founding leader, is transitioning to lead a new initiative called OAI Labs, aimed at developing innovative interfaces for human-AI collaboration. The Model Behavior team has been pivotal in shaping AI responses and navigating political biases, recently facing scrutiny due to controversial changes in GPT-5. OpenAI strives to balance creating friendly AI interactions while maintaining helpfulness and critical thinking in responses, especially in light of serious concerns raised in recent lawsuits involving its platforms.
Source link
