Last Updated on November 10, 2020 by Sean B
Before we answer the question, what is the the Turing Test? Let’s talk a little about the history surrounding the development of the test. The Turing Test was developed during the time immediately after World War II, which was full of incredible technological advances in a variety of fields such as medicine, textiles, finance, and most importantly, computer science. Amidst the chaos of war, steps had been taken in a several areas that led to the development of the precursors to our modern computers, including the Enigma Machine and the Bombe Machine developed by Alan Turing, which solved the Enigma Code.
Once the war ended, those minds that created the machines to aid wartime efforts were freed to develop their ideas further. The Turing Test is one of these developments; it acted as an inspiration for modern-day computers and the AI programs such as Chatbots and virtual assistants we use today. In today’s blog, we shall look at the Turing test’s history and the nuances associated with it.
Who Invented the Turing Test?
The Turing Test was initially called the Imitation Game, so named because of the goal of the computer to imitate a human. The test was invented by Alan Turing, a brilliant British mathematician, logician, cryptanalyst, and computer scientist. Turing came up with the Turing Test in his 1950 paper “Computer Machinery and Intelligence.”
Alan Turing had put forth the profound question of whether or not machines can think. To answer his question, he came up with an imitation game where machines can participate. Since its conception, the test has been widely influential and keeps providing us with new insights.
Turing’s imitation game has also become a widely debated topic in the philosophy of Artificial Intelligence. Even though Alan Turing never referred to himself as a philosopher, his research papers and profound questions concerning machine thinking abilities are of great philosophical worth.
How Turing Test Works?
Judges in an interrogation room conduct the Turing Test. The judge has to simultaneously hold a conversation between a human agent and a computer. Both the participants, i.e., the computer and the human, are both hidden from view.
The job of the judge is to find out which one of the participants in the machine. The test revolves around the central goal of coming up with a machine that can trick the judge into thinking that he is talking to a human.
The machine which successfully tricks the judge into thinking that he is talking to a human wins the test. The questions asked to the machine and the human are not pre-determined rather are spontaneous and are unpredictable.
The machine’s capability is determined not only by how smart it can be but also by how dumb it can be. Humans are fallible and are capable of making various mistakes. A judge will easily spot a participant who exhibits extraordinary intelligence. Therefore the machines have to pretend to make mistakes just like an average human will.
The Philosophical Background of the Turing Test
The Turing Test not only has it’s worth in the AI but also philosophical parlance and discussion. The discussion of whether a mind is capable of thinking on its own has a long history. In this section, we shall look at these philosophical discussions around the Turing Test.
Rene Descartes’s Mind-Body Dualism
Rene Descartes was a French philosopher, scientist, and mathematician. He is known as the father of modern philosophy and one of the minds behind the scientific revolution. Descartes was known for his mind and body dualism, i.e., mind and bodies are separate entities.
Rene Descartes discusses the Turing Test’s precedents in his 1637 treatise “Discourse on the method.” Rene discusses that the machines of his time responding to human interactions, but the automata of his time are unable to respond to speech.
However, Descartes failed to imagine a future where automata can accomplish that feat. The abstract concept of machines that Turing gave eventually accomplished this task and was able to lend more weight to the statement that “Machines can think.”
Who Passed the Turing Test?
Turing Test has been implemented in various fields and competitions to find out the capabilities of the AI machines participating in the test. The first time when an AI program appeared to pass the Turing Test happened in 1966, when a program created by Joesph Weizenbaum, ELIZA, participated in the test.
The Eliza chatbot is a natural language processing program which was created at MIT’s Artificial Intelligence Laboratory. Eliza worked by using “pattern matching” mechanisms. Eliza provides the user with the illusion of a human conversation.
Eliza was the AI program to attempt the Turing Test and is known as one of the first chatterbots. Elize works by analyzing the typed comments of the users to look for keywords. Then, when a keyboard is found, which is relevant, the user’s sentence is transformed, and the new sentence is returned to the user.
Eliza was created to mimic the behavior of a Rogerian psychotherapist. When Eliza was used in interactions with humans, many reported that Eliza was very effective at convincing that they are talking to a human.
Eliza was a groundbreaking program in the history of Artificial Intelligence. Eliza opened new horizons and fundamentally changed the intuitions of experts regarding Artificial Intelligence. Furthermore, Eliza inspired many programs, and many programs are based on Eliza.
Eugene Goostman is known by many as the first AI program to have finally passed the Turing test in 2014. The Turing test 2014 was held at the Royal Society in London. The event was organized by the university’s school of system engineering.
Eugene is presented as a 13-year-old boy. The bot was developed in Saint Petersburg, Russia. The development team consists of Vladimir Veselov and Eugene Demchenko. Eugene was highly effective at fooling the judges into thinking that they are talking to an actual 13-year-old boy.
The Turing Test’s winning requirement was for the program to fool the judges for 30% of the time into thinking that they are talking to a human for a series of five-minute conversations. Eugene was able to accomplish the task for 33% of the human judges.
This event is known as the milestone in the Turing Test’s history and has opened endless new possibilities for the AI programs of the Future. Eugene Goostman is now available online as well for users to chat with.
Application of Turing Test in the Loebner Prize
One of the Turing Test’s most popular applications is its application in the annual Loebner Prize competition. Loebner Prize was launched by Hugh Loebner, an American inventor, in 1990.
The first annual Loebner Prize competition was held in November 1991. A prize of $25,000 was announced for bots, which convince the judge to think that the human is a computer program, and a prize of $100,000 was announced for the bots which can decipher visual, textual, and auditory inputs.
No participant has ever won these $25,000 (silver) and $100,000 (gold) prizes. The bronze awarded has been consistently awarded to participants every year by the judges. The first winner to have won the prize is “Whimsical Conversation” created by Joseph Weintraub.
Perhaps the most remarkable AI program which has won the Loebner Prize is Mitsuku, which created by Pandorabots using AIML technology. Mitsuku has won the Loebner Prize for a record five times and is regarded as the world’s most human-like AI program.
Variations of the Turing Test
There are many Turing Test variations that have been proposed and are used in programs and competitions.
1. Captcha – The Reverse Turing Test
A Turing Test variation which reverses the role between humans and machines, is known as the reverse Turing Test. One of the most widely used instances of the reverse Turing Test is Captcha.
Captcha is used in websites that require authorization and need to check whether the person interacting with the website is human. Captcha is often used for safety issues. Captcha test is presented by providing the user with a blurry image containing letters, and the user has to type those letters in a box.
The Captcha test’s effectiveness is questioned now after the development of the programs that pass the captcha test. For example, in 2013, researchers from Vicarious announced that their software could break the captcha test from google, yahoo, and PayPal.
2. Total Turing Test
Total Turing Test is a variation of the Turing Test proposed by cognitive scientist Steven Harnad in 1991. The Total Turing Test is just like a normal Turing Test, but it has two crucial requirements that are absent in the normal Turing Test.
The first requirement is for the bot to identify, recognize, and decipher visual images, i.e., having a perspective and vision that requires the bot to have advanced visual capabilities. And the other requirement is for the bot to be able to perform physical tasks that normal humans can do.
The idea behind the Total Turing Test is for bots to perform all tasks that humans can do. However, these requirements demand the bot to have a physical body, in other words, robotics.
The Total Turing Test is an inspiration for robots and can be used as a criterion to determine robots’ capability. With the use of Total Turing Tests, one can come with efficient robots that can live alongside humans.
The Turing Test is one of the most influential experiments in computer science history, which acted as an inspiration for modern AI programs and chatterbots. The implementation of the Turing Test in the Loebner Prize allowed the test to be carried out on a larger scale and opened new horizons for AI programs.
We hope that you found the blog informative.
We’re looking forward to reading your thoughts in the comments section below.