Turing changes the “Can machine think?” question into the imitation game, a game where an interrogator has to distinguish between a machine and a human through asking each questions. If the machine is indistinguishable from the human, the machine has successfully passed the Turing Test. If a machine passed the test, Turing argued it fulfills the only measurable standard of intelligence we can actually test.
Underlying question: Is the output of a system all that is needed to classify whether it “thinks”?
This is a Behavioral point of view. Turing focused only on the external behavior. If a machine acted intelligent, then it should be treated as intelligent. He pointed out that we cannot actually prove other humans are “thinking” or conscious. We only judge them by their behavior.
After initial reading, I thought that in the process of a machine being able to pass the Turing Test, it must be intelligent. I thought of it in a machine learning point of view. If you trained a model to pass The Turing Test, it would need to become intelligent in order to generalize onto any question the interrogator could ask. Therefore the machine would become intelligent in the process of optimizing for The Turing Test loss function.
See objections in Minds, Brains, and Programs and Psychologism and Behaviorism.