Alan Turing was a mathematician and logician, and one of the first computer scientists. In the second world war, he played an important role in helping British intelligence crack the German codes that were used to encrypt military communications.
For AI, Turing is still hugely influential. He is primarily associated with two distinct concepts:
- The Turing Machine.
- The Turing Test.
The two should not be confused. They are totally different things that happen to be both associated with the same man who discovered them. (In the same way like the phonograph and the electric lightbulb have both been invented or perfected by Edison; but they are still totally different things).
The Turing Machine is a theoretical (non-existent) machine. It is a thought experiment that aims to clarify what we mean by computation, and which functions can be considered computable. We will talk about that in another post.
The Turing Test (also discussed in a later post) is a test for intelligence that is based on conversation with a computer. It was not really called that by Turing himself. He had a different kind of test, which he called the Imitation Game; but the Imitation Game was very confusingly described by Turing himself, and it has no relevance to modern AI research and discussion. For all practical purposes, we can safely substitute a modern description of the Turing Test for Turing’s original Imitation Game.
The Turing Test
The Turing Test goes like this:
- You have two rooms. Behind closed doors, in one room is a computer. In the other room is a human being. You don’t know who (or what) is in each particular room.
- Outside the rooms is a human judge. He has to find out which room contains the computer and which room has a human in it.
- The only way he has to communicate with the rooms is by using a chat interface, like WhatsApp or SMS. He can type a message to a room, and get a typed reply from that room. Nothing else. All communication has to be through text only.
- The idea is that if the judge cannot distinguish which room contains the human and which room contains the machine, then we are justified in calling the machine “intelligent.”
Since this was first proposed by Turing (in a paper called “Computing Machinery and Intelligence,” 1950) researchers have repeatedly tried to make computers pass the Turing Test. To date, none has succeeded.
We discuss the challenges and whether the Turing Test is a good test for intelligence in detail in a separate post.