The recent development of AI compels us to reconsider the very aim of English learning, since machine translation has made it much easier for learners to interpret and compose texts in a foreign language. However, the linguistic capabilities of machines are imperfect as they are not equipped with cognitive operations that human language necessitates. This chapter examines some phenomena where tools of machine language often fail, such as metaphorical expressions and syntactic ambiguities, whose interpretations require analogical mappings as well as inferences based on contextual understanding and background knowledge. It will be shown from a cognitive linguistic perspective that our implicit linguistic knowledge presupposes cognitive and conceptual operations that machines would still find difficult to perform.