Turing Test
The Turing Test is a measure of a computer's ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human. Proposed by mathematician and computer scientist Alan Turing in his 1950 paper "Computing Machinery and Intelligence," the test involves a human judge who interacts with both a human and a computer, attempting to determine which is which based solely on the conversation. The concept gained popularity in the 1990s with the establishment of competitions like the Loebner Prize, which incentivized the development of AI that could successfully mimic human conversation. In 2014, a computer program named Eugene Goostman achieved a milestone by convincing a third of judges that it was human, marking a significant moment in AI history.
Despite its prominence, the Turing Test has limitations, particularly in assessing the broader advancements in AI technology. Critics argue that it emphasizes conversational mimicry over other critical capabilities, potentially overlooking significant innovations like robotics and data processing. The ongoing debate around the Turing Test raises important questions about how we define and measure artificial intelligence, and whether mimicking human behavior is the best indicator of true intelligence. As AI technology continues to evolve, discussions about the relevance of the Turing Test in evaluating machine intelligence remain pertinent.
Turing Test
The Turing test is a test of computers and artificial intelligence (AI) that is supposed to show when computers are capable of thinking. The test gets its name from mathematician and computer pioneer Alan Turing, who suggested the basis for the test in an academic paper in the 1950s. Since the 1990s, people have used the Turing test to assess computer programs and determine how advanced AI has become. Technology has changed rapidly since Alan Turing first described his test, but a computer did not beat the Turing test until 2014. Although Turing tests are routinely performed, it has limitations in assessing the advancements computers have made over time.
Alan Turing and Turing Test
Alan Turing first described his hypothetical test in a paper called "Computing Machinery and Intelligence," which he published in October 1950. Turning was employed at Manchester University, where he was working on a then-state-of-the-art computer. He was in charge of creating programs for the computer, but he was also very interested in ideas about what he called "intelligent machines." Turing believed that computers would change dramatically over time and that eventually humans would create machines that could think. His paper laid out some of his views about intelligent machines, and it proposed what would become the Turing test.
Turing suggested that computers could eventually be programed to have an intelligence similar to that of humans. The "imitation game," as Turing called it, would be a test in which a judge communicates with one human and one computer. The judge then has to determine which was the machine imitating a human and which was the human. Turing believed that eventually a computer would be able to fool many judges. To Turing, this test would prove the computer’s ability to think in ways that are similar to the ways humans think.
How the Turing Test Works
What became known as the Turing test had been discussed among technology experts for decades after Turing’s paper. In 1990, New York businessman Hugh Loebner started an annual competition that was set up like a Turing test to help determine the speed of AI sophistication. He offered a prize of $100,000 to any person who created a computer program that could pass the test. The Loebner Prize was offered regularly since its inception, but it took more than twenty years for a computer to win. Turing himself believed that computers would be "intelligent" enough to beat the test by the year 2000; however, a computer did not earn the Loebner Prize until more than a decade later.
For each year that the Loebner Prize was held, the group organizing the test hired judges and human controls. The judges had different talents and may have included, for example, language experts or psychology experts. The judges were separated from the computers and the human controls, sometimes called "confederates," throughout the test. The judges chatted with different human controls and a computer for five minutes each. During that time, the human controls attempted to prove to the judges that they were actually human. At the same time, the computer programs entered into the test tried to fool the judges into thinking that they were human. To win the test, the computer had to fool the judges into thinking it was human more than 30 percent of the time.
In June 2014, a computer finally won the Loebner Prize. The computer program that won was called Eugene Goostman, and it simulated a thirteen-year-old boy who was not a native English speaker. Eugene, like other computer programs entered in the contest, was a chatbot. Chatbots can carry on conversations with human respondents. Although Eugene tricked 33 percent of the judges into thinking it was actually a human, some AI experts thought that the Eugene program had used some trickery to win the test. Since this chatbot was supposed to be young and a nonnative speaker, the judges may have attributed any strange responses or odd grammar to these facts. In the past, some chatbots that failed the Turing test seemed human-like during a majority of the test, but they failed in the end because of an odd response. Though the Loebner Prize was one of the famous models of the Turing test, other—slightly different—formats also have been suggested for Turing tests. In 2020, BBC News reported that the Loebner Prize competition had ended.
In June 2022, the Washington Post reported that a Google employee had claimed that the company's Language Model for Dialogue Applications (LaMDA) AI chatbot had become sentient. However, Google quickly determined the employee's claim to be false. At the time of the claim, the Washington Post reported an uptick in the number of AI researchers who believed AI sentience may be possible in the near future.
Limitations of the Turing Test
Throughout the years that the tests have been performed, technology has undergone incredible changes. Computers can process data faster than ever before, and they allow people all around the world to communicate and gain access to limitless information in a matter of seconds. Nevertheless, most modern computers would fail the Turing test because the elements of technology that have advanced the most have not been the aspects that make computers more like humans. Some AI experts worry about the value placed on the Turing test. They believe that the real advances being made in AI are being overlooked by the test. These critics argue that the test focuses on factors that might not be as important in AI as other factors. The chatbots that were entered in the Loebner Prize are not as useful to humans as smartphones and robots that can help assemble products. Advancements in these types of technology are, at least for now, probably more important to humans than advances in chatbots.
Bibliography
"Computer AI Passes Turing Test in 'World First.'" BBC. BBC. 9 June 2014. Web. 30 July 2015. http://www.bbc.com/news/technology-27762088
Hern, Alex. "What Is the Turing Test? And Are We All Doomed Now?" The Guardian. Guardian News and Media Limited. 9 June 2014. Web. 30 July 2015. http://www.theguardian.com/technology/2014/jun/09/what-is-the-alan-turing-test
Hodges, Andrew. "Alan Turing and the Turing Test." Alan Turing: The Enigma. Andrew Hodges. Web. 30 July 2015. http://www.turing.org.uk/publications/testbook.html
Reingold, Eyal. "The Turing Test." University of Toronto. University of Toronto. Web. 30 July 2015. http://www.psych.utoronto.ca/users/reingold/courses/ai/turing.html
Sharkey, Noel. "Alan Turing: The Experiment That Shaped Artificial Intelligence." BBC. BBC. 21 June 2012. Web. 30 July 2015. http://www.bbc.com/news/technology-18475646
Tiku, Nitasha. "The Google Engineer who Thinks the Company's AI Has Come to Life." The Washington Post, 11 June 2022, www.washingtonpost.com/technology/2022/06/11/google-ai-lamda-blake-lemoine/. Accessed 3 Mar. 2023.
"Turing Test." Webopedia. QuinStreet Inc. Web. 30 July 2015. http://www.webopedia.com/TERM/T/Turing‗test.html
Wakefield, Jane. "Robot Bores: AI-Powered Awkward First Date." BBC News, 1 Nov. 2020, www.bbc.com/news/technology-54718671. Accessed 3 Mar. 2023.