Sounds a lot like the Reinforcement Learning from Human Feedback (RLHF) that ChatGPT was trained on. Seems like such a training phase is needed to avoid the Syd

Sounds a lot like the Reinforcement Learning from Human Feedback (RLHF) that ChatGPT was trained on. Seems like such a training phase is needed to avoid the Sydney/Bing search debacle.