Translated data: Generative AI models such as GPT-4 and Midjourney demonstrate compelling generation capabilities. However, research has found that these models face challenges in understanding their own generated content, differing from human intelligence configurations. Specifically, researchers have discovered through experiments that these models perform well in selective tests but often make mistakes in inquiry-based tests. This research calls for caution when delving into artificial intelligence and cognition, as models can create content but cannot fully comprehend it.