Recent studies have revealed significant limitations in the capabilities of visual AI models and large language models (LLMs). Research indicates that visual language models (VLMs) perform poorly on vision tasks, often failing to see fine details and making educated guesses similar to a person with myopia or blindness. Additionally, a study by MIT CSAIL highlights that the reasoning skills of LLMs are frequently overestimated. These models excel in familiar scenarios but struggle with novel tasks, suggesting that their perceived reasoning abilities may rely more on memorization than true cognitive processing. An article by Devin Coldewey in TechCrunch further explores how elite AI models fail basic vision tasks, raising questions about the current understanding and future development of AI technologies.