What if your AI looked at an image, answered, then doubted itself? | Dark Hacker News