Thinking About The Turing Test

In 1950, Alan Turing published his famous paper Computing Machinery and Intelligence (http://www.turing.org.uk/scrapbook/test.html). Alan Turing was the inventor of the Turing Machine, which is one of the modern-day computer’s first ancestors. In his paper, he describes the philosophical question of the Turing Test: according to his thought process, any computer to pass the Turing Test is intelligent and, consequently, thinks.
The thought-experiment is as follows.
Imagine a human judge who is presented a screen on which he has a dialogue with another human and with a computer. The judge would only be able to communicate through the computer keyboard and screen, so as to not be influenced by other indices of human intelligence such as verbal abilities, tone of voice, accent, and so forth. Aware that one of the beings he is communicating with is human, while the other is not, the judge’s task would be to discriminate the real human from the computer, whilst the two participants try to answer questions to convince the judge that they each are human. A computer that would succeed in confusing the judge would thus pass the intelligence test.

In this case, the judge is given full reign on the types of questions they can ask to determine the intelligence of the machinery, therefore the test is presented in the form of an interview. The issue is deciding which questions determine human intelligence. Modern-day computers are trickier to test, especially if the judge is restricted in the kinds of questions they can ask. Technology advancement since the 1950s has allowed the creation of computers that can perform at more or less the same level as human capacities. Let’s take the example of computer chess applications: these are computer programs that are designed to present a challenge to the average chess player, and can even outplay humans. Another example is ELIZA, a computer program designed to mimic a psychotherapist by responding to language cues that seem to indicate she is empathic. These advancements make the interview process trickier: the questions the judge asks need to touch at the core of human intelligence. So what is intelligence?
In his own test, Turing measured intelligence as the ability to think, according to Daniel C. Dennett (as appears in Levitin, 2011). What if we apply the psychological measurement of intelligence to the question? Psychologically, the most commonly used measure of intelligence is IQ, which combines verbal, mathematical and spatial abilities to determine a person’s general intelligence. IQ is a relatively reliable indicator of academic success, but we use calculators (mini-computers) to perform mental calculations better than we do. Does this mean that a computer can be as or more intelligent than a human, according to our method of measurement? And what does this imply about IQ? Perhaps the intelligence quotient measures an intelligence that applies to computers, such as knowledge absorption, speed of processing and response to cues, rather than the human experience. A computer could define all the words of the English language, but it would not know that always acting in one’s own self-interest may not always be the best solution to a problem, because it does not take into account the possible reactions of the beings around it.

The Turing Test emerges as an intriguing tool, not only to measure artificial intelligence, but to question intelligence itself.

References:

Levitin, D. J. (2011). Foundations of Cognitive Psychology (2nd ed.). Boston, MA: Allyn & Bacon.
The Alan Turing Internet Scrapbook. (n.d.). Retrieved February 01, 2016, from http://www.turing.org.uk/scrapbook/test.html