Although machine translation has reached a high intelligence degree after nearly one century of development, there is still a big gap between the translation quality of machine translation and human translation. Largely as a linear, unidirectional and highly regular code conversion activity and due to its lack of flexible interaction with the original text, machine translation still bears somewhat obvious shortcomings in the application of high-order thinking, the accurate understanding and expression of the text, the aesthetic reproduction of the original style, and the production of creative translations, etc. compared with human translation. This paper conducts a tentative approach to the above problems through contrastive study of examples, aiming at benefiting the future development of machine translation.