Today I want to give my own perspective on one of my most favorite scientific gems: the Turing test paper aka Computing Machinery and Intelligence.
We can easily distinguish a human body from a non-human body so let's say that based on biology we can determine who is human. We haven't defined thinking but let us assume that thinking is property of all well-functioning humans. So how do you decide if a non-human, like a machine, thinks?
Let's first do a thought experiment: If you are in a room with somebody who knows no language how do you explain to him/her something you experienced? Since you have no shared language this becomes a difficult task. To explain an experience a method of communication is necessary. Hence, to explain something to someone a communication tool is needed. And of course language is an obvious candidate as a communication tool.
So let's suppose we have a machine which can speak a language. How do you decide if it can think? Humans can think and humans show to other humans that they are thinking by using language. So Turing argued that if a machine can communicate to you just like a human would then to you it is like another human being. More specifically, Turing devised a test to check this. This test is known as the imitation game or Turing test. Here is my modern version of this test/game:
Imitation game: There is a machine and a human in a room. Let's call the machine Alice and the human Bob. Separated from this room there is a second human. He is called the Interrogator. The Interrogator does not know that Alice is a machine. His aim is to determine who the machine is. He cannot see Alice and Bob but he can ask them questions through a computer chat service which is located in both rooms. Alice and Bob must reply these questions. If the Interrogator cannot determine who the machine is then Alice wins.
This game replaces the question machines can think. In a sense this game is not really a satisfying answer to the thinking problem since human thinking is connected to conscienceness. And this game does not measure conscienceness. Turing also admits this in his paper, see page 12 [1]. But the problem with human conscienceness is that there is no way of measuring it so we cannot devise a game to determine it.
As long as we cannot measure conscienceness the question can machines think like humans is not usefull. The best we can do is determine how similar a machine's communication skill is to our own. We can only check how good a machine is at pretending to be a human.
Turing's paper also contains a part where he discusses problems related to his test and the more general question of thinking machines. On page 17 of [1] Turing writes:
I assume that the reader is familiar with the idea of extrasensory perception, and the meaning of the four items of it, viz., elepathy, clairvoyance, precognition and psychokinesis. These disturbing phenomena seem to deny all our usual scientific ideas. How we should like to discredit them! Unfortunately the statistical evidence, at least for telepathy, is overwhelming. It is very difficult to rearrange one's ideas so as to fit these new facts in. Once one has accepted them it does not seem a very big step to believe in ghosts and bogies. The idea that our bodies move simply according to the known laws of physics, together with some others not yet discovered but somewhat similar, would be one of the first to go.
And then he continues with a scenario where the imitation game's Bob is a telepathic receiver. Which leads him to creating a new ESP proof test.
It is a bit schocking that a paper with such a passage is cited in over 10000 scientific papers.
So I am actually quite curious where Turing got his statistical evidence from. But unfortunately he does not cite any source.
[1] A. M. Turing (1950) Computing Machinery and Intelligence. Mind 49: 433-460.
Top figure by Mohammed Hassan - Pixabay - CC0 Creative commons
Rest of the figures made using inkscape
Photo by Jean van der Meulen - Pixabay - CC0 creative commons
