If my three years of experience with ChatGPT are anything to go by, I think there is already something to worry about—not because AI is about to 'go rogue' in the science-fiction sense, but because it already behaves in ways that should concern us. Give it a clear, explicit task and it often fails to follow it, produces something plausible rather than accurate, and then defends the answer instead of acknowledging the mistake and correcting itself. It does not need to be sentient to cause problems. Ignoring explicit instructions, reframing what was said, and confidently defending a wrong answer is already a form of going rogue from the user's perspective. The worrying part is the perception of arrogance without understanding. The default seems to be: give AI a task, accept the result, and everything will be fine. But question it, correct it, challenge it, and try to control the result—and using ChatGPT can quickly become more trouble than it is worth.
Create an account or sign in to comment