What is Turing Test?
In 1950, Alan Turing, the mathematician and computer scientist who broke the Enigma code and helped the Allied Forces win War War II, issued a challenge to computer scientists around the world. The challenge was to create software that could converse with humans in natural language without the humans finding out that they were communicating with a computer program instead of a human.
He said that a computer system can be considered to be artificially intelligent if it is able to mimic human responses under certain conditions.
The test that verified whether a computer program is truly indistinguishable from a human came to be known as the Turing Test. This inspired computer scientists globally to strive to create programs that can beat the Turing Test.
In 1991, Hugh Loebner started the annual Loebner Prize competition, offering $100,000 to whoever creates the first computer that can pass the Turing Test.
How does the Turing Test work?
The premise of the Turing Test is simple. The original test involved three terminals that are hidden from each other. One terminal is operated by a human who acts as a questioner. The other two terminals are operated by respondents, however, one of them is operated by a human respondent while the other is controlled by a computer program.
The questioner asks the respondents a series of questions that revolve around a specific subject area. The questions also follow a specific format and context.
This continues for either a predetermined amount of time or a specified number of questions. After that, the questioner has to try to figure out which of the respondents is the human and which is the computer, based purely on the responses, without seeing the respondents.
If the judge cannot determine which of the respondents is the computer and which one is the human, the computer program would have passed the Turing Test and is considered to possess artificial intelligence.
What software has passed the Turing Test?
Till now, no software has actually been able to pass the Turing Test in the true sense. However, a few systems have come close to beating the test.
Since no program has actually passed the test, the Loebner Prize competition started awarding $2000 for the best attempt every year.
What are the limitations of the Turing Test?
The Turing Test is not foolproof. It is possible to cheat on the test.
The program, Eugene Goostman, acts in the role of a 13-year old Ukrainian boy, which makes judges account for grammatical errors and a lack of general knowledge. It even managed to convince 33% of the judges at a contest that it was human.
In the past, a program would not have any chance of scoring high unless the nature and format of the questions were heavily limited.
Since the questions are limited to a narrow domain, a system that can answer them cannot actually be considered to possess true artificial intelligence (Artificial General Intelligence).