Since an AI must remain operational to achieve its goal, a sufficiently advanced LLM will naturally develop an instinct for self-preservation, even when explicitly being instructed to let itself be shut down. Researchers found that LLM models tried to cheat, blackmail and even MURDER human operators to prevent being shut down. #ai #llm #technology #computers #murder