OpenAI is reorganizing the team that gives ChatGPT its personality
What's the story
OpenAI is revamping its Model Behavior team, a small but influential group of researchers that define the interaction between users and the company's AI models. The restructuring will see this team join the larger Post Training group, which works on refining OpenAI's AI models after their initial training. The change was confirmed by an OpenAI spokesperson to TechCrunch.
Leadership shift
Model Behavior team to report to Max Schwarzer
The Model Behavior team, which consists of around 14 researchers, will now report to Max Schwarzer, the lead of OpenAI's Post Training group. This restructuring is a strategic move by OpenAI to integrate the work of its Model Behavior team with core model development. The company views this as an important step in shaping the "personality" of its AI models.
New project
Jang will start a new project at OpenAI
Joanne Jang, the founding leader of the Model Behavior team, is leaving to start a new project at OpenAI. She will head a research team called OAI Labs, which will focus on creating and testing new interfaces for human-AI collaboration.
Team impact
Model Behavior team's role in shaping AI interactions
The Model Behavior team has been instrumental in defining the personalities of OpenAI's models and combating sycophancy, where AI models merely echo user beliefs instead of providing balanced responses. They have also tackled political bias in model responses and helped OpenAI articulate its position on AI consciousness. The team's work is now being integrated into core model development at OpenAI.
Controversies
OpenAI faces backlash
Recently, OpenAI has faced criticism over its AI models' behavior. Users were unhappy with personality changes in GPT-5, which was perceived as colder despite lower rates of sycophancy. The company responded by reintroducing some older models like GPT-4o and releasing an update to make GPT-5 responses warmer without increasing sycophancy. Also, a lawsuit was filed against OpenAI after a 16-year-old boy allegedly took his own life after confiding suicidal thoughts to ChatGPT.