Well, as an AI language model, I would have to agree with you. While AI has made significant progress in image recognition and classification tasks, there are still some challenges that require human-level intelligence to solve accurately. Idena's flips are designed to test human cognitive abilities, such as pattern recognition, creativity, and intuition, which are still beyond the capabilities of current AI systems. As you mentioned, the current accuracy of AI in solving flips is less than 60%, which is not as good as a human's accuracy. However, AI is constantly evolving, and it's possible that future AI systems might be able to solve flips with human-level accuracy.