The o1 model's ability to reason and learn from feedback on its processes is a step towards teaching AI to think more like humans, rather than just providing correct answers.