Abstract: The Turing test is one of the philosophical pillars of AI, which has spurred the development of AI. Nowadays, while some people still deem it to be “the golden standard” of AI, many think it is “out”, and “the time has come to bid farewell to it”. This paper shows that many criticisms against the Turing test is resulted from misunderstanding. This includes the behaviorist interpretation, the eliminative interpretation and the inductive interpretation. Once these misunderstandings are cleared, those criticisms correspondingly dissolve. However, what is more important is that even most of supporters of the Turing test do not get the deep insight an Turing produced in it. The deep insight is that intelligence is an “emotional concept”. This paper agrees with Diane Proudfoot that it is the response-dependent interpretation that matches the idea that intelligence is an “emotional concept”. However, the account Proudfoot makes is problematic. This paper gives an adequate account of the response-dependent interpretation and argues that the adequate account implies that a machine must be endowed with social cognition capacities for it to pass the Turing test. This implication tells us that the key for building a machine which has real intelligence lies not in giving a computational account of the hidden essence of intelligence, but in social cognition.