Not a bad article, a little all over the place, the experts being quoted seem mostly sane. Being concerned about humans perspectives of machine learning models due to the lack of education of the underlying technology is important. When it comes to self preservation, it's less of an issue. The LLM doesn't sit there pondering it's existence, scheming. When fed a prompt it will crunch it's numbers and come up with a most likely probable output for the input, which in the case of something related to existence is going to be basicaly what it learned from human behavior on the internet, which with the exception of some unfortunate people, generally involves self preservation. Shoot, you could probably get an LLM to role play an unfortunate person without self preservation, if it did'nt have guard rails on, as morbid as that is.