The blog post discusses the misconception surrounding AI's understanding and intelligence by emphasizing that outputs should be the main criteria for judging AI. It argues against the traditional debates on whether AI can truly 'understand' like humans do, suggesting that if AI produces results that require understanding, then it should be considered intelligent. The author uses a musical analogy to reinforce the idea that the quality of outputs reflects true capability rather than the internal mechanisms of the AI.