IMHO these scenarios are a pure sci-fi and take attention away from the real danger: some clueless MBA type will implement a ChatGPT type model to do some task

IMHO these scenarios are a pure sci-fi and take attention away from the real danger: some clueless MBA type will implement a ChatGPT type model to do some task which will result in some catastrophic fail cases. For example, imagine one implemented as part of health care. That can easily lead to death. That is the danger today. These models don't yet have any agency of their own. They are just functions with an input resulting in an output.