← All Posts /

The Turing Test

Have you ever imagined a world where machines think, act and make decisions just the way humans do? The question of whether it is possible for machines to think like humans has a very long history. The Turing Test, originally known as The Imitation Game, was introduced in 1950 by the English Mathematician Alan M. Turing to determine whether a computer can “think” the way humans do. 

WHAT IS THE TURING TEST?

The Turing Test (originally known as an Imitation Game) is a method of inquiry into artificial intelligence (AI) for determining whether a computer is capable of thinking like a human being or not. The test is named after its creator Alan Turing, an English Computer Scientist, Theoretical Biologist, Mathematician, and Cryptanalyst.

Turing proposed that a computer can be said to possess artificial intelligence if it can replicate human responses under particular conditions. The Original Turing Test involved three entities, each of which would be physically separated from the rest. One of the entities would be operated by a computer, while the other two would be operated by humans.

During the test, one human is an interrogator, while the second human and the computer function as respondents or answer terminals. The questioner asks questions from the respondents within a particular subject area with a similar format and context. After a pre-set duration (usually 5 minutes), or a particular number of questions, the questioner is then asked to differentiate the human and the computer out of the two.

The test is repeated a number of times. If the questioner cannot reliably differentiate between the conversation of the computer and human then the computer is considered to have passed the test because its answers are “just as human” as the human respondent.

THE WORKING OF THE TURING TEST

According to Turing, the main motive of the test is that a machine has to try and pretend to be a human, by answering questions put to it, and it will only pass if it is successful in convincing to be a human.

The humans were restricted from giving away any of the personal information during the tests.

In the tests conducted at the Royal Society in June 2014, there were six different sessions with five parallel imitation games at a time occurring during each session. A different judge was selected for each game, which meant there were five judges in each session. Each session had five rounds, with five parallel imitation games in each round. Each anonymous human was part of the five games in a session. All five machines (the five different competition bots) took part, so every machine was involved in five games per session, hence 30 games in totality.

In a particular session, a judge conducted five different tests. In their first test, they noticed a hidden human pitted against a hidden machine

The second test conducted involved a different human against a different machine. And so on. It would continue until the judge had conducted all five tests in that session. At the end of each test, they were asked to tell each entity if they were able to differentiate between a machine and a human.

RESULTS

There were five machines involved in total in the tests and their success rates were:

● Eugene Goostman 33%

● Elbot 27%

● J. Fred 20%

● Ultra-Hal 13%

● Clever Bot 7%

In each case, their success rate was in respect of judges identifying them as humans. Eugene Goostman was the program that was considered to have passed the test with a success rate of 33%.

EUGENE GOOSTMAN

Eugene Goostman is a chatbot programmed by three Russian programmers in Saint Petersburg in 2001. The program replicates a thirteen-year-old Ukrainian boy and is said to have been successful in passing the Turing test at an event organized by the University of Reading.

According to Vladimir Veselov, one of the creators of the program, the choice of the age of Eugene Goostman was taken as thirteen because a thirteen-year-old is “neither too old to know a lot nor too young to know nothing”. This young age also leads to people ignoring any minor grammatical errors that the program might end up making. The following is the transcript of a conversation between one of the Judges of the test, and two respondents, one of which is a human and the other is Eugene Goostman. We would like you to read both the transcripts and see whether you are able to distinguish between the human respondent and Eugene or not.

TRANSCRIPT 1:

Judge: Good afternoon

Respondent: Good afternoon!

Judge: How many left hands do you have?

Respondent: The same as right hands, and how about you?

Judge: The same of course. I love the scent of new-mown hay. How do you feel about scent?

Respondent: I find it exciting and refreshing! Does it make you feel anything? Judge: Well it depends..Which smell do you really hate?

TRANSCRIPT 2:

Judge: Hello

Respondent: Hello, I’m really glad to have the chance to chat with you! My guinea pig Bill sends his regards too!

Judge: Is Bill a male or a female?

Respondent: Ask Bill personally, please.

Judge: Well I’d rather talk to you. My name is Jane and I am female. How about you? What’s your gender?

Respondent: I’m a male. A “guy”, I’d say.

Judge: Pleased to meet you. What’s the weather like where you are?

Respondent: Let’s get on with our conversation!

Judge: Don’t you like talking about the weather?

Respondent: All these talks about weather is a waste of time

Judge: What would you like to discuss?

Respondent: I don’t know … Better tell me more about yourself! Where do you come from, by the way? Could you tell me about the place where you live?

Judge: It’s a lovely place with two bedrooms and a great view over London. What can you see from your bedroom window?

Respondent: I’ve never been to London, but I’d really love to visit it! To see their … their … (damn, what do they have …) Oh yes – their Big-Ben!

In this conversation, the respondent in Transcript 1 was a female human whereas the Respondent in Transcript 2 was the machine Eugene. The judge considered the respondent in Transcript 1 to be definitely a machine, awarding it only 20 out of 100 (a very poor score) for humanlike conversation. However, they marked the respondent in Transcript 2 i.e Eugene Goostman as unsure.

ALTERNATIVES TO THE TURING TEST

There were many alternatives to Turing Tests that were later developed. These alternatives include:

● The Marcus Test – A test in which a program can watch a visual show and is then interrogated by being asked meaningful questions about the show’s content.

● The Lovelace Test 2.0 – A test made to detect AI through its ability to create art.

● Winograd Schema Challenge – A test that asks multiple-choice questions in a particular format.

Recommended » Benford's Law