Turing Test in Artificial Intelligence
This article will see an outline of the Turing Test in AI, which was introduced by renowned mathematician Alan Turing in 1950. The test was designed to check whether a machine can exhibit Human-like Intelligence in its work or thought process. As per the research paper by Alan Turing, the test was called ‘The Imitation Game’. He believed that a computer/machine could be considered intelligent only if it can mimic human-like responses. The machine passes the test if the interrogator (Judge) cannot decide whether the response is coming from a machine or a human.
How is the Turing Test Performed?
Turing test is played with three players. It has one computer, one human responder, and one human interrogator. The computer and Human responder sits on two separate terminals and an Interrogator who is unaware of their identity asks a series of questions with both of them. Based on the dialogue with these two parties, the interrogator has to decide which one is a computer and which one is a human responder. If the interrogator is equally likely to pick either of the two or unable to distinguish between the two, we can say that the computer has successfully demonstrated human-like intelligence. As per Turing, the computer is deemed successful, if the interrogator decided incorrectly as often when the responder is the computer (the game is played with the computer) as when the game is played with a man and a woman.
Here is a sample conversation between Responder (Computer or Human) and an interrogator.
Interrogator: Are you a Computer?
Interrogator: Please write me one Shakespeare sonnet.
Responder: Sorry, I am not a big fan of poetry.
Interrogator: Add 20957 to 46707.
Responder: (After a brief pause of about 20 Seconds) 67664.
Interrogator: Do you like playing Chess?
Interrogator: I have only my King left, which is in K1 square. You have only King and Rook left, King is on the K6 square, and a Rook is on R1 square. What will be your move?
Responder: (After a brief pause) Rook to R8 and checkmate.
Based on this conversation, the interrogator has to decide whether the responder is human or a computer.
Modern Day Turing Test
The recent version of the game has just one responder, and the interrogator has to find out if the responder is a computer or human. To reduce human errors, the modern Turing Test has multiple human interrogators. The machine is considered successful if, after five minutes of communication, 30% of the judges consider the computer as humans. Another prominent form of the Turing Test is The Loebner Prize which is an annual Turing Test competition that was started in 1991. The competition had additional rules where computer programs need to have 25 minutes conversation with the four presiding judges. The winner (Need not pass the Turing Test) is one whose computer receives the most number of votes, irrespective of the fact whether 30% of the judges vote for it. Alan Turing made this prediction that by the year 2000, machines will pass the Turing Test. However, machines didn’t pass until the year 2014. In 2014, chat-bot named Eugene Goostman passed the Turing test when it received votes from 33% of the judges who believed that the chat-bot was human. The chat-bot mimicked the personality of a 13-year-old boy.
Requirements for Clearing the Turing Test in AI
The following are a few of the key features that the machine must possess to clear the Turing test.
1. Natural Language Processing: The machine must be able to understand and synthesize the natural/human language in which the interrogator is communicating.
2. Knowledge Representation: The machine must be able to store and retrieve information in real-time to communicate effectively with the interrogator.
3. Automated Reasoning: Machine should be able to draw reasonable, logical inferences based on previously stored information. This logical inference can trick the interrogator into believing that it is conversing with a human.
4. Machine Learning: Machine should learn from its past mistakes and should have the ability to recognize patterns in the conversation to respond in a human-like manner.
(The remaining two features are needed for clearing Total Turing Test)
5. Vision: Machine should have the ability to recognize Interrogator’s action and other surrounding objects.
6. Motor Control: Machine should be able to move and perform the action as requested.
Features of Turing Test in AI
- The test provides a key basis for evaluating Artificial Intelligence applications.
- Turing tests do not directly test intelligence; instead, it checks how similar a computer behaves like humans.
- The test gives an unbiased notion about the machine’s ability to behave or perform actions like humans.
- It eliminates any biased notion towards humans as human negotiators only focus on the content provided by the respondents.
Critics of Turing Test
Although the Turing Test has many takers in the Artificial Intelligence industry. There are many non-believers to this test. The most famous argument against the Turing Test is the ‘Chinese Room Argument’. It was presented by John Searle in the year 1980. He argued that by programming a computer we could make it understand the semantics of the language, but it can never understand the conscience that makes up a human mind, as true intelligence in humans is due to a combination of different complex nerve cells which is very hard to emulate for a machine. He further added that machines like PARRY and ELIZA have successfully passed the Turing test by manipulating the symbols of the language in which the interrogator asked the question. Although they had no understanding of the language in which the question was asked. Therefore passing the Turing test is not concrete proof of machine thinking like humans.
Despite a large number of critics and opposing thoughts Turing test has opened lots of new avenues for innovation and discovery in the Artificial Intelligence sphere. To this date, it remains a prominent test to find out if the machine can mimic and perform human-like action.
This is a guide to Turing Test in AI. Here we discuss the different requirements for clearing the Turing Test in AI along with its features. You can also go through our suggested articles to learn more –