AES154 The Turing Test

Artificial intelligence is a very interesting subject. For many years now, there’s been unanswered questions about whether it is possible for a computer to develop actual intelligence. With the rapidly developing technology of today, I believe it is only a matter of time. Even now, the vast amount of processing power and advanced programming allows our computers to do incredibly complex calculations. Some people even have been programming scripts that allow the computer to mimic human interactions and conversations. But how would we determine if a computer program has actually passed the impenetrable barrier into the realm of intelligence and consciousness? 

In order to determine this, a British computer scientist named Alan Turing asks a very simple question, “Can a computer talk like a human?”

Turing proposed a game, there would be three terminals required, one of them separated from the other two. In the lone chair sits a person acting as the questioner, and out of sight lies the other two terminals, one would be operated by a human and the other a computer. The questioner must interrogate these two respondents, usually within a set theme and subject matter. After the conversations have ended, the questioner must decide which of the respondents is the computer. This test would be repeated numerous times, and if the questioner only identifies the computer correctly in half the tests or less, then the computer would pass the test, seeing as the questioner determines its responses to be just as human as the real person.  

This is the famous Turing Test. A seemingly simple task for the advanced computers of today, simulating human interactions and mimicking our responses. But almost all of them have failed. The ones that did “succeed” did so by focusing on finding clever ways to fool the judges. An example of this was a script that mimicked a psychologist, encouraging the humans to talk more about themselves, and reflecting those words and experiences back at them in the same way a psychologist would. Most of the programs written for this test are capable of complex and nuanced conversations, but only the ones they are trained for. Veer away from the subject matter at hand and they’ll fall apart very quickly. 

It is interesting to note that these amazing computers that are capable of flying spacecrafts, perform delicate surgeries, and solve unimaginably complex equations, are incapable of engaging in casual conversation like human beings. This is because human language is a highly complex phenomenon that requires no small amount of feeling and intuition to understand. Being able to converse in a language takes much more than just simply memorizing a dictionary. One final thought I’ll leave you with, a computer program that passes the Turing test would be quite interesting indeed, but what if there was a program that intentionally fails it?