Autonomous large language models sometimes do things they absolutely should not, such as suddenly go off and mine ...