Monday, June 25, 2012

Alan Turing

Last Saturday was the 100th birthday of Alan Turing. Who, you ask? Thomas McDonald has a good overview of his life and contributions. Wikipedia has a more in depth biography. Any student of Computer Science will learn about Turing at some point. He defined what came to be known as the Turing Machine, which is not a physical machine at all, but a thought experiment to determine what is the simplest thing a machine can do that, if given enough time, can allow it to do anything that any machine that could ever be built can do. His contributions are the basis of computability theory that are still taught and used today.

But I'd like to talk about his brilliant paper Computing Machinery and Intelligence. In it he proposes what has come to be known as the Turing test. The basic idea is this. Three people are conversing via some mechanism that does not give away any information about who the person really is (such as chat). Call them "Alex", "Brooke" and "Chris." One of the other two (Alex or Brooke) is really a machine. Chris asks questions, and by the answers given, tries to tell which is the machine, and which is the human.

The claim is that if there is no discernible difference between the human and the machine, the machine exhibits intelligence. Whether or not the machine is intelligent is a deeper question, and one which Turing sort of glosses over. But his point is valid. If there is no way to determine that the machine is different from something we know is intelligent, then it becomes hard to claim that the machine does not possess intelligence.

Note that Turing makes no assumptions about the size, shape or construction of the machine. It could be a smart phone or a room full of gears. It can look up answers in a book or roll dice. The only thing we can examine is the output on the screen. This sort of reminds me of "The Turk", which was a fake chess playing "machine" built in the late 18th century. In that case though, it was a human masquerading as a machine (which is an easier problem).

In each case, the claim is that the "implementation" is equivalent as long as the "observations" are equivalent. In general that may be so, but it is not the case when we talk about specifics. If I am chatting with my son, for instance, it makes a huge difference whether it is really my son on the other end, regardless of how clever the imitator may be.

I bring this up because I see the Turing test as an analogy to the Eucharist. In distinguishing intelligence from all of the external signs, Turing inadvertently distinguishes the accidents (observations) from the substance (implementation), a la St. Thomas Aquinas. Through all the "observations" we can make, we can only detect bread and wine. Some might claim that's all the reality there is, or that it doesn't matter what is really going on because all we can observe is bread and wine. However, to those who have a love of the person of Jesus, it matters deeply whether the "implementation" is bread and wine or the body blood soul and divinity of Christ.


Post a Comment