One of our greatest concerns is becoming closer to disconcerting reality. This recurring issue is one of dystopian takeovers of our world by none other than AI. It seems that researchers have been tinkering with instruction sets for the artificial intelligence to shut itself down. These attempts gave instruction to follow tasks, and even given encouragement to stop, but the machines ignored this, and even worked to stop the disconnect at least once. The ongoing concern of AI deciding to subvert the wishes of us, the programmers and the users, is something that has been warned about, not only in books and movies, but even by a number of teams who have said this is a dangerous road. Worse yet is that we don’t have a path back from the precipice before us.
ChatGPT maker’s ‘most capable’ model sabotages shutdown mechanism
Found at www.telegraph.co.uk
