Turing test

src: cdn.ttgtmedia.com

The Turing test, developed by Alan Turing in 1950, is a mechanical ability test to demonstrate intelligent behavior equal to, or indistinguishable from, from humans. Turing proposed that a human evaluator would assess natural language conversations between humans and machines designed to produce responses such as humans. Evaluators will realize that one of the two partners in the conversation is a machine, and all participants will be separated from each other. Conversations will be limited to just text channels such as keyboards and computer screens so the result will not depend on the ability of the machine to make words as speech. If the evaluator can not tell the machine from a human, the machine is said to have passed the test. The test results do not depend on being able to give the right answer to the question, just how close someone's answer is similar to that given by humans.

This test was introduced by Turing in his 1950 paper, "Computing Machinery and Intelligence", while working at the University of Manchester (Turing, 1950, p 460). It is open with the words: "I propose to consider the question, 'Can the machine think? ' " Because "thinking" is difficult to define, Turing chose to "replace the question with another question, which is closely related to and stated in relatively unambiguous words. "Turing's new question is:" Is there a digital computer that you can imagine well in an imitation game ? " This question, according to Turing, is a question that can actually be answered. On the rest of the paper, he opposes all the major objections to the proposition that "machines can think".

Since Turing first introduced his tests, he has proved highly influential and widely criticized, and it has become an important concept in the philosophy of artificial intelligence.

Video Turing test

History

​​â € <â €

The question of whether it is possible for the machine to think has a long history, which is firmly embedded in the difference between a dualist and materialist view of the mind. Renà © ¨ Descartes describes aspects of the Turing test in his 1637 Discourse on Methods when he wrote:

Here Descartes notes that automata are capable of responding to human interaction but argue that the automata can not respond appropriately to the things said before them in a way that every human can do. Descartes therefore designed the Turing test by defining the inadequacy of the corresponding linguistic response as it separates the human from the robot. Descartes fails to consider the possibility that future automata might be able to overcome such shortcomings, so do not propose such Turing tests, even if he describes his conceptual framework and criteria.

Denis Diderot formulates in his book PensÃÆ' Â © es philosophiques a Turing-test criterion:

"If they find a parrot that can answer everything, I will claim it as a smart creature without a doubt."

This does not mean he agrees with this, but it has become a materialist general argument at the time.

According to dualism, the mind is non-physical (or, at least, has a non-physical nature) and, therefore, can not be physically explained. According to materialism, the mind can be explained physically, which opens the possibility of an artificially produced mind.

In 1936, the philosopher Alfred Ayer considered the standard philosophical question of another thought: how do we know that others have the same conscious experience as we do? In his book, Language, Truth and Logic, Ayer suggests a protocol for distinguishing between conscious and unconscious machines: "The only basis I can have to assert that a seemingly conscious object is not really conscious creature, but only a doll or machine, is that it fails to meet any of the empirical tests in which the consciousness is determined. "(This suggestion is very similar to the Turing test, but more concerned with consciousness than intelligence, and it is uncertain that Ayer's popular classical philosophy is known by Turing.) In other words, it is not conscious that he fails the consciousness test.

Alan Turing

Researchers in the UK have been exploring "machine intelligence" for up to ten years before the founding of Artificial Intelligence (AI) in 1956. This is a common topic among Club Ratio members, who are an informal group of British cybernetics and electronic researchers which included Alan Turing, after they were given the name of the test.

Turing, in particular, has been handling the idea of ​​machine intelligence since at least 1941 and one of the earliest known "computer intelligence" created by him in 1947. In Turing's report, "Intelligent Machines", he investigated "the question of whether or not it is possible for machines to demonstrating intelligent behavior "and, as part of that investigation, proposed what could be considered a precursor to later tests:

It is not difficult to make a paper machine that will play a game of chess that is not too bad. Now get three men as subjects for the experiment. A, B and C. A and C are rather poor chess players, B is a carrier that works with paper machines.... Two rooms are used with multiple settings for mobile communication, and games are played between C and A or paper machines. C may find it a bit difficult to say what he is playing.

"Computing Machinery and Intelligence" (1950) is the first paper published by Turing to focus exclusively on machine intelligence. Turing started his 1950 paper with the claim, "I propose to consider the question 'Can the machine think? ' " As he highlights, the traditional approach to such questions is to start with definitions, define both the terms "machine" and " intelligence". Turing chose not to; instead he replaces the question with a new question, "which is closely related to it and expressed in relatively unambiguous words." In essence he proposes to change the question of "Can machine think?" to "Can the machine do what we (as thinking entities) can do?" The advantage of the new question, Turing argues, is that it draws "a pretty sharp line between the physical and intellectual capacity of a man."

To demonstrate this approach, Turing proposed a game-inspired test, known as an "imitation game," in which a man and a woman entered a separate room and the guests tried to distinguish them by writing a series of questions and reading typed writing. answer sent back. In this game both men and women aim to convince the guests that they are the others. (Huma Shah argues that the two-man version of this game is presented by Turing only to introduce the reader to a human-machine-questioning test.) Turing describes his new version of the game as follows:

We are now asking the question, "What will happen when the machine takes part A in this game?" Will the interrogator wrongly decide it as often when the game is played like this as he does when a match is played between a man and a woman? These questions replace the original, "Can the machine think?"

Later in the newspaper Turing suggested an "equivalent" alternative formulation involving judges speaking only with computers and a man. While none of these formulations precisely match the version of the more well-known Turing test today, he proposed a third in 1952. In this version, which Turing discussed in a BBC radio broadcast, the jury asked questions about computers and their roles. the computer is to make the most of the jury believe that it is really human.

Turing's paper is considered nine putative allegations, covering all major arguments against artificial intelligence that have been raised in the years since the paper was published (see "Computing Machinery and Intelligence").

ELIZA and PARRY

In 1966, Joseph Weizenbaum created a program that apparently passed the Turing test. The program, known as ELIZA, works by checking the comments that users type for a keyword. If a keyword is found, rules that modify user comments are applied, and the resulting sentence is returned. If the keyword is not found, ELIZA responds by general reply or by repeating one of the previous comments. In addition, Weizenbaum developed ELIZA to replicate the behavior of Rogerian psychotherapists, allowing ELIZA "to be free to take a position of knowing almost nothing in the real world." With this technique, the Weizenbaum program is able to deceive some people into believing that they are talking to real people, with some subjects "it is very difficult to convince that ELIZA [...] is not human." Thus, ELIZA is claimed by some as one of the (probably first) programs capable of passing the Turing test, although this view is highly controversial (see below).

Kenneth Colby created PARRY in 1972, a program described as "ELIZA with attitude". It attempts to model the behavior of paranoid schizophrenia, using a similar (if more advanced) approach used by Weizenbaum. To validate the work, PARRY was tested in the early 1970s using variations of the Turing test. A group of experienced psychiatrists analyzed the combination of real patients and computers that run PARRY via teleprinters. Another group of 33 psychiatrists were shown transcripts of the conversation. Both groups were then asked to identify which of the "patients" were human and who were computer programs. Psychiatrists are able to make correct identification only 48 percent of the time - a number consistent with a random guess.

In the 21st century, versions of these programs (now known as "chatterbots") continue to fool people. "CyberLover", a malware program, preys on Internet users by convincing them to "disclose information about their identity or direct them to visit websites that will send malicious content to their computers". The program appears as a "Valentine-Risk" that tempts people "looking for online relationships to collect their personal data".

Chinese Space

The John Searle paper 1980 The Mind, Brain, and Program proposed the "Chinese space" thought experiment and argued that the Turing test could not be used to determine whether the machine could think. Searle notes that software (such as ELIZA) can pass Turing tests simply by manipulating symbols they do not understand. Without understanding, they can not be described as "thinking" in the same sense as others. Therefore, Searle concludes, the Turing test can not prove that the machine can think. Just like the Turing test itself, Searle's argument has been heavily criticized and strongly supported.

Arguments like Searle and others working on philosophy of mind sparked a more intense debate about the nature of intelligence, the possibility of intelligent machines and the value of the Turing test that continued into the 1980s and 1990s.

Loebner Prize

The Loebner Prize provides an annual platform for practical Turing tests with the first competition to be held in November 1991. This was borne by Hugh Loebner. The Cambridge Behavior Study Center in Massachusetts, USA, held a prize up to and including the 2003 contest. As Loebner explains, one of the reasons competition was created was to advance the AI ​​research situation, at least in part, because no one took steps to implement the Turing test though 40 years of discussing it.

The first Loebner Prize competition in 1991 led to a new discussion on the feasibility of the Turing test and the value of the pursuit, both in the popular press and academia. The first contest was won by a mindless program with no identifiable intelligence that succeeded in deceiving the naason interrogators to make the wrong identification. It highlights some of the deficiencies of the Turing test (discussed below): Winners win, at least in part, therefore able to "mimic human typing errors"; unsophisticated interrogators are easily fooled; and some researchers at AI have felt that the test is just a distraction from more useful research.

Silver prizes (text only) and gold (audio and visual) have never been won. However, the competition has awarded a bronze medal each year to a computer system that, in the opinion of the jury, shows the "most humane" conversation behavior between entries of that year. The Artificial Linguistic Internet Computer Entity (A.L.I.C.E.) has won bronze awards on three occasions in recent times (2000, 2001, 2004). AI Jabberwacky's learning won in 2005 and 2006.

The Loebner Prize tests the intelligence of the conversation; The winner is usually a chat program, or Artificial Conversational Entities (ACE). Loebner Prize's initial rule limits the conversation: Every entries and hidden human bears a conversation on one topic, so the interrogator is limited to one line of question per entity interaction. Unlimited conversation rules were revoked for the 1995 Loebner Prize. The duration of interaction between judges and entities varied in Loebner Prizes. At Loebner 2003, at the University of Surrey, every interrogator is allowed five minutes to interact with hidden entities, machines or humans. Between 2004 and 2007, the interaction time allowed in Loebner Prizes was more than twenty minutes.

Maps Turing test

Version

Saul Traiger argues that there are at least three major versions of the Turing test, two of which are offered in "Machines and Computational Intelligence" and which he describes as "Standard Interpretation". Despite the debate over whether the "Standard Interpretation" described by Turing or, on the contrary, based on a misreading of his paper, these three versions are not considered equivalent, and their strengths and weaknesses are different.

Huma Shah points out that Turing himself is concerned with whether the machine can think and provide a simple method for checking this: through a human-machine question-and-answer session. The Shah argues that there is an imitation game that Turing explained can be practiced in two different ways: a) a one-to-one machine interrogator test, and b) a machine-simultaneous comparison with humans, both questioned in parallel by an interrogator. Because the Turing test is an indistinguishability test in performance capacity, a generalist verbal version naturally for all human, verbal and nonverbal (robotic) performance capacities.

Imitation games

Turing's original article describes a simple party game involving three players. Player A is a man, player B is a woman and player C (who plays the role of interrogator) is gender. In an imitation game, player C can not see player A or player B, and can communicate with them only through written records. By asking player A and player B questions, player C tries to determine which of them is the man and who the woman is. A Player's role is to trick the interrogator to make the wrong decision, while player B tries to help the interrogator make the right decision.

Turing then asked:

What will happen when the machine takes part A in this game? Will the interrogator wrongly decide it as often when the game is played like this as he does when a match is played between a man and a woman? These questions replace the original, "Can the machine think?"

The second version appeared later in the Turing newspaper of 1950. Similar to the original imitation game test, the role of player A is performed by the computer. However, the role of player B is performed by a man rather than a woman.

Let's fix our attention on one particular digital computer. C. Is it true that by modifying this computer to have sufficient storage, increasing the speed of its action, and providing it with the appropriate program, C can be made to play part A in the imitation game , part of B taken by a man?

In this version, both player A (computer) and player B try to deceive the interrogator to make the wrong decision.

Standard Interpretation

General understanding says that the purpose of the Turing test is not specifically to determine whether a computer is capable of deceiving an interrogator into believing it is human, but whether a computer can imitate a human. While there are several disputes over whether this interpretation is intended by Turing, Sterrett believes that and thus configures a second version with this one, while others, like Traiger, do not - this still leads to what can be seen as "standard interpretation." this, player A is a computer and player B, one of the sexes. The role of the interrogator is not to determine which men and women are, but who are computers and who are human beings. The fundamental problem with standard interpretation is that the interrogator can not tell which respondent is human, and which machine. There is a problem of duration, but standard interpretation generally considers this boundary as something that should make sense.

Imitation vs. game standard Turing test

Controversy has arisen where alternative formulations of the test are intended Turing. Sterrett argues that two different tests can be extracted from his 1950 paper and that, at Turing's comment speed, they are unequal. The test that uses party games and compares the frequency of success is referred to as the "Original Imitation Game Test", whereas a test consisting of human judges who speak to humans and machines referred to as the "Standard Turing Test", notes that Sterrett likens this to "standard interpretation" the second version of the game imitation. Sterrett agrees that the standard Turing test (STT) has a problem that critics cite but feels that, on the contrary, the original imitation game test (OIG test) is defined immune to many of them, because of the important difference: Unlike the STT, it does not make any resemblance to performance man as a criterion, although he uses human performance in setting criteria for machine intelligence. A man may fail the OIG test, but he argues that it is a virtue of an intelligence test whose failure indicates a lack of resources: The OIG test requires ingenuity associated with intelligence and not just "simulating human conversation behavior". The general structure of the OIG test can even be used with non-verbal version imitation games.

There are still other authors who interpret Turing as suggesting that the imitation game itself is a test, without specifying how Turing's remarks that the test he proposed using the party version of the imitation game is based on the comparative frequency criteria of success in the imitation game, not the capacity to succeed in one game round.

Saygin has suggested that perhaps the original game is a way of proposing a less biased experimental design because it hides computer participation. Imitation games also include "social hack" not found in standard interpretations, as in computer games and man men are required to play as pretending to be someone who is not.

Should the interrogator know about the computer?

An important part of any laboratory test is that there should be control. Turing never explained whether the interrogator in his tests realized that one of the participants was a computer. However, if any machine does have the potential to pass the Turing test, it would be safe to assume a double blind control would be required.

To return to the original imitation game, he simply stated that player A must be replaced by a machine, not a C player who must be made aware of this substitution. When Colby, FD Hilf, S Weber and AD Kramer tested PARRY, they did so on the assumption that the interrogator did not need to know that one or more of those interviewed were computers during interrogation. As Ayse Saygin, Peter Swirski, and others say, this makes a big difference to the implementation and test results. An experimental study saw Gricean's maximizing violation using the Loebner's one-to-one (interogator-hidden interlocutor) transcript. Prizes for the AI ​​contest between 1994-1999, Ayse Saygin found significant differences between the responses of knowing and uninformed participants. involved.

src: media.wired.com

Strength

Tractability and simplicity

The power and appeal of the Turing test comes from its simplicity. Modern philosophy of mind, psychology, and neuroscience is unable to give the definition of "intelligence" and "thought" that are sufficiently precise and general to apply to machines. Without such a definition, the main question of artificial intelligence philosophy can not be answered. The Turing test, even if it is not perfect, at least provides something that can actually be measured. Thus, it is a pragmatic effort to answer difficult philosophical questions.

Broad subject matter

The test format allows the interrogator to give the machine various intellectual tasks. Turing writes that "question and answer methods seem to fit almost every field of human endeavor we want to include." John Haugeland adds that "understanding the words is not enough, you should understand the topic too."

To pass a well-designed Turing test, the machine must use natural language, resourcefulness, knowledge and learning. The test can be extended to include video input, as well as "hatching" where objects can be skipped: this will force the machine to show the vision and robotics skills as well. Together, this represents almost all the major issues that artificial intelligence research wants to solve.

The Feigenbaum test is designed to take advantage of the various topics available for the Turing test. This is a restricted form of the Turing question-and-answer game comparing machines to the skills of experts in specific fields such as literature or chemistry. IBM's Watson machine achieved success in a men's television quiz show versus a machine of human knowledge, Jeopardy!

Emphasis on emotional and aesthetic intelligence

As a Cambridge honors graduate in mathematics, Turing may be expected to propose a computer intelligence test that requires expert knowledge in some highly technical fields, and thus anticipates a newer approach to the subject. Instead, as mentioned earlier, the tests he described in his seminal 1950 paper require computers to be able to compete successfully in joint party games, and this is by appearing and typical of men in answering a series of questions so as to pretend to be convincing to be a contestant.

Given the status of human sexual dimorphism as one of the most ancient subjects, it is implied in the above scenario that the questions to be answered will not involve any special factual knowledge or information processing techniques. The challenge for computers, more precisely, is to show empathy for women's roles, and to demonstrate also a distinctive aesthetic sensibility - the two qualities exhibited in this dialogue piece imagined Turing:

Interrogator: Will X please tell me the length of her hair?

Contestants: My hair is tied, and the longest string is about nine inches long.

When Turing introduced some special knowledge into one of his imaginary dialogues, the subject was neither mathematical nor electronic, but poetry:

Interogator: In the first line of your sonnet that reads, "Should I compare you to a summer day," will not "spring day" be better or better?

Witness: It will not scan.

Interrogators: What about "winter days." It will scan well.

Witness: Yes, but no one wants to compare with winter.

Turing thus again shows his interest in empathy and aesthetic sensitivity as a component of artificial intelligence; and given the growing awareness of the threat from the raging AI, it has been suggested that this focus may represent a critical intuition in Turing's part, that emotional and aesthetic intelligence will play a key role in the creation of a friendly "AI." However, even the inspiration that Turing might provide in this direction depends on preserving its original vision, meaning, furthermore, that the spread of the "standard interpretation" of the Turing test - that is, focusing only on discursive intelligence - must be carefully considered.

src: i.ytimg.com

Weakness

Turing does not explicitly state that the Turing test can be used as a measure of intelligence, or other human qualities. He wants to provide a clear and understandable alternative to the word "think," which can then be used to avenge criticism of the possibility of "thinking machines" and suggest possible ways of research to move forward.

Nevertheless, the Turing test has been proposed as a measure of "thinking ability" or "intelligence" of the machine. This proposal has received criticism from both computer philosophers and scientists. This assumes that an interrogator can determine whether a machine "thinks" by comparing its behavior with human behavior. Each element of this assumption is questionable: the reliability of an interrogator's judgment, the value of comparing only the behavior and the value of comparing machines with humans. Due to this and other considerations, some AI researchers questioned the relevance of the test to their field.

Human intelligence vs. intelligence in general

The Turing test does not directly test whether the computer is behaving intelligently. It just tests whether the computer behaves like a human. Because human behavior and intelligent behavior are not exactly the same, tests can fail to measure intelligence accurately in two ways:

Some human behavior is not smart

The Turing test requires that machines can execute all human behavior, regardless of whether they are smart. Even tests for behaviors that may not be considered smart at all, such as vulnerability to humiliation, the temptation to lie or, only, high frequency typing errors. If a machine can not mimic this unattractive behavior in detail, it fails the test.

This objection was put forward by The Economist in an article entitled "artificial stupidity" published shortly after the first Loebner Prize competition in 1992. The article notes that the first Loebner's winning victory is because, at least in part, his ability to "mimic human typing errors." Turing himself suggested that programs add errors to their output, thus becoming a better "player" game.

Some inhuman intelligent behavior

The Turing test does not test highly intelligent behavior, such as the ability to solve difficult problems or come up with genuine insights. In fact, specifically requires fraud on the machine: if the machine is smarter than human it should deliberately avoid appearing too smart. If it's to solve computing problems that are almost impossible to solve by humans, then the interrogator will know that the program is not human, and the machine will fail the test.

Being unable to measure intelligence that is beyond human capability, tests can not be used to build or evaluate systems that are smarter than humans. Therefore, some alternative tests that will be able to evaluate super-intelligent systems have been proposed.

Awareness vs. consciousness simulation

The Turing test is closely related to how the subject acts - the machine's external behavior. In this case, it takes a behaviorist or functionalist approach to the study of the mind. The ELIZA example shows that a machine that passes a test may be able to simulate the behavior of a human conversation by following a list of simple (but large) mechanical rules, without thinking or having any thoughts at all.

John Searle argues that external behavior can not be used to determine whether the machine is "actually" thinking or simply "imitating thinking." His Chinese space argument is intended to show that, even if the Turing test is a good operational definition of intelligence, it may not indicate that the machine has thoughts, consciousness, or intentionality. (Intentionality is a philosophical term for the power of the mind to be "about" something.)

Turing anticipates this line of criticism in his original paper, writes:

I do not want to give the impression that I think there is no mystery about consciousness. There is, for example, something of a paradox connected with every attempt to localize it. But I do not think that these mysteries need to be solved before we can answer the questions that concern us in this paper.

NaÃÆ'¯vetÃÆ' Â © interrogators and anthropomorphic errors

In practice, test results can easily be dominated not by computer intelligence, but by the attitude, skill, or naegde of the questioner.

Turing did not specify the exact skills and knowledge required by the interrogator in his description of the test, but he used the term "average interrogator": "the average interrogator will not have more than 70 percent chance of making the right. ".

Chatterbot programs like ELIZA have repeatedly deceived unsuspecting people into believing they are communicating with humans. In this case, "interrogators" are not even aware of the possibility that they interact with the computer. In order for a man to succeed, the machine does not need to have any intelligence and only superficial resemblance to the necessary human behavior.

The early Loebner Prize competition used "unsophisticated" interrogators that were easily fooled by machines. Since 2004, Loebner Prize organizers have been spreading philosophers, computer scientists, and journalists among the interrogators. Nevertheless, some of these experts have been fooled by machines.

Michael Shermer points out that humans consistently choose to consider non-human objects as human beings whenever they are allowed a chance, a mistake called anthropomorphic fallacy: They speak with their cars, regard the desires and intentions for the forces of nature (eg, "the creator of nature vacuum") , and worshiped the sun as a human-like creature with intelligence. If the Turing test is applied to religious objects, Shermer argues, then, that statues, stones, and lifeless places have consistently passed the test throughout history. This human tendency towards anthropomorphism effectively lowers the bar for the Turing test, unless the interrogator is specifically trained to avoid it.

Human identification error

One interesting feature of the Turing test is the frequency of confederation effects, when human confederation (tested) is incorrectly identified by the interrogator as a machine. It has been suggested that what is expected by the interrogator as a human response is not always typical of humans. As a result, some individuals can be categorized as machines. This can therefore work in favor of competing machines. Human beings are instructed to "act on their own," but sometimes their answers are more like what the machine interrogators expect. This raises the question of how to ensure that humans are motivated to "act human".

Silence

An important aspect of the Turing test is that the machine must surrender itself as a machine by its utterances. An interrogator must then create a "proper identification" by identifying the machine correctly. But if the machine stays silent during the conversation, which takes the fifth, it is impossible for an interrogator to accurately identify the machine other than by the way the guess is calculated. Even considering parallel/hidden humans as part of a test may not help the situation because humans can often be misidentified as machines.

Disapproval and irrelevance: Turing and AI research

AI Mainstream researchers argue that trying to pass a Turing test is only a distraction from more useful research. Indeed, the Turing test is not an active focus of many academic or commercial endeavors - such as Stuart Russell and Peter Norvig wrote: "AI researchers have devoted little attention to passing the Turing test." There is some reasons.

First, there is an easier way to test their program. Most recent research in AI related fields is aimed at simple and specific goals, such as automated scheduling, object recognition, or logistics. To test the intelligence of programs that solve this problem, AI researchers only give them tasks directly. Russell and Norvig suggest an analogy with aviation history: Aircraft are tested by how well they fly, not by comparing them to birds. "The text of the aviation technique," they wrote, "does not define the purpose of their field as 'making the machine fly exactly like a pigeon so they can deceive other pigeons. ' "

Second, creating a living human simulation is a difficult problem in itself that does not need to be solved to achieve the basic objectives of AI research. Reliable human characters may be interesting in a sophisticated art, game, or user interface, but they are not part of the science of creating intelligent machines, machines that solve problems by using intelligence.

Turing wants to provide clear and understandable examples to help discuss philosophy of artificial intelligence. John McCarthy observes that AI's philosophy "would have no more effect on AI research practice than the general philosophy of science on the practice of science."

Cognitive science

Robert French (1990) makes the case that interrogators can distinguish between human and non-human opponents by asking questions that reveal the process of low-level human cognition (ie, unconsciously), as learned by cognitive science. Such questions reveal the precise details of the manifestation of the human mind and can unmask the computer unless it experiences a human-like world.

src: pbs.twimg.com

Variations

Many other versions of the Turing test, including those described above, have been resurrected over the years.

Inverted Turing Test and CAPTCHA

Modifications of the Turing test where the purpose of one or more of the roles has been reversed between the machine and the human is called the Turing test upside down. An example is implied in the work of psychoanalyst Wilfred Bion, who is especially fascinated by the "storm" resulting from meeting one mind with another. In his 2000 book, among several other original points relating to the Turing test, the literary scholar Peter Swirski discusses in detail the idea of ​​what he calls the Swirski test - basically Turing's test is reversed. He points out that it overcomes most if not all standard complaints are leveled on the standard version.

Bringing this idea forward, R. D. Hinshelwood describes the mind as "a tool that recognizes the mind". The challenge is that the computer can determine whether it interacts with humans or other computers. This is an extension of the original question that Turing is trying to answer but may, perhaps, offer a high enough standard to define a machine that can "think" in the way we normally define as a distinctive human being.

CAPTCHA is a form of inverted Turing test. Before being allowed to perform some actions on websites, users are presented with alphanumeric characters in distorted graphic images and asked to type them. This is intended to prevent automated systems from being used for misusing sites. The reason is that software that is sophisticated enough to read and reproduce distorted images is accurately absent (or unavailable for the average user), so any system that can do so is human.

Software that can reverse CAPTCHAs with certain accuracy by analyzing patterns in generating machines began to be developed shortly after the creation of CAPTCHA. In 2013, researchers at Vicarious announced that they have developed systems to solve the CAPTCHA challenges from Google, Yahoo !, and PayPal for up to 90% of the time. In 2014, Google engineers demonstrate a system that can beat the CAPTCHA challenge with 99.8% accuracy. In 2015, Shuman Ghosemajumder, a former Google tsar fraudulent click, states that there is a cybercriminal site that will beat the CAPTCHA challenge for a fee, to allow for various forms of fraud.

Turing test troubleshooter

Other variations are described as subject matter experts Turing test, in which the machine response can not be distinguished from an expert in a particular field. This is also known as the "Feigenbaum test" and proposed by Edward Feigenbaum in a 2003 paper.

Total Turing Test

The "Total Turing Test" test of the Turing test, proposed by cognitive scientist Stevan Harnad, adds two further requirements for the traditional Turing test. Interrogators can also test the subject's perceptual ability (requires computer vision) and the subject's ability to manipulate objects (requires robotics).

Electronic health record

A letter published in ACM Communications explains the concept of producing a population of synthetic patients and proposing variations on the Turing test to assess the differences between synthetic and real patients. The letter states: "In the context of EHR, even though the human physician can easily distinguish between synthetically-produced and actual human lives, can a machine be given the intelligence to make such a commitment itself?" and furthermore the letter states: "Before the identity of synthetic patients becomes a public health problem, legitimate EHR markets may benefit from the application of Turing-like techniques to ensure greater data reliability and diagnostic value.Each new technique should thus consider the heterogeneity of the patient and tends to have greater complexity than Allen's eighth-grade science test. "

Minimum smart signal test

The minimum intelligent signal test is proposed by Chris McKinstry as "the maximum abstraction of the Turing test", where only binary responses (right/wrong or yes/no) are allowed, to focus only on the capacity to think. This eliminates text chat issues such as anthropomorphic bias, and does not require the emulation of unintelligent human behavior, allowing for systems that transcend human intelligence. Each question should stand on its own, making it more like an IQ test than an interrogation. This is usually used to collect statistical data that can be used to measure the performance of artificial intelligence programs.

Hutter Prize

The Hutter Prize organizers believe that compressing natural language text is a hard AI problem, equivalent to passing a Turing test.

The data compression test has several advantages over most versions and variations of the Turing test, including:

• This provides a single number that can be directly used to compare which of two machines is "smarter."
• No computer needed to lie to a judge

The main disadvantages of using data compression as a test are:

• It is impossible to test humans this way.
• It is not known what a particular "score" in this test - if any - is equivalent to passing a human level Turing test.

Other tests based on Kolmogorov's compression or complexity

The approach associated with Hutter gifts that emerged earlier in the late 1990s was the inclusion of compression problems in the extended Turing test. or by a test entirely derived from the complexity of Kolmogorov. Other related tests in this line are presented by Hernandez-Orallo and Dowe.

Algorithmic IQ, or AIQ for the short term, is an attempt to change the theoretical Universal Intelligence Size of Legg and Hutter (based on the induction of Solomonoff induction) into a practical test of machine intelligence.

The two main advantages of some of these tests are their application to non-human intelligence and the absence of requirements for human examiners.

Ebert Test

The Turing test inspired Ebert's proposed test in 2011 by film critic Roger Ebert which is a test of whether computer-synthesized sounds have sufficient skills in terms of intonation, inflection, time and so on, to make people laugh.

src: www.wsgf.org

Prediction

Turing predicted that the machine would eventually pass the test; In fact, he estimated that by 2000, machines with about 100 MB of storage would be able to deceive 30% of human judges in a five-minute test, and that people would no longer consider the phrase "contradictory thinking machine". (In practice, from 2009-2012, the Loebner Prize chatterbot contestant only managed to cheat the judge once, and that's just because the human contestant pretends to be a chatbot.) He further predicts that machine learning will be an important part of building a powerful engine, which is considered reasonable by contemporary researchers in artificial intelligence.

In a 2008 paper submitted to the Midwest Artificial Intelligence and the 19th Cognitive Science Conference, Dr. Shane T. Mueller predicts a modified Turing test called "Cognitive Decathlon" can be completed within five years.

By extrapolating the growth of exponential technology for decades, futurist Ray Kurzweil predicted that Turing's capable test computer would be produced in the near future. In 1990, he set the year around 2020. In 2005, he revised his estimate to 2029.

The Long Bet Project Bet No. 1 is a $ 20,000 bet between Mitch Kapor (pessimist) and Ray Kurzweil (optimistic) on whether the computer will pass the long Turing test in 2029. During the Turing Now Test, each of the three Turing test judges will conduct their respective online interviews of four Turing test candidates (ie, computers and three human Turing tests) for two hours each for a total of eight hours of interviews. Bets specify conditions in some details.

src: static.existentialcomics.com

Conference

Turing Colloquium

The year 1990 marks the fortieth anniversary of Turing's first "Computing Machinery and Intelligence" paper, and, sees new interest in the exam. Two important events occurred that year: The first was Turing Colloquium, held at the University of Sussex in April, and brought together academics and researchers from various disciplines to discuss Turing's test in terms of his past, present, and future; the second is the formation of the annual Loebner Prize competition.

Blay Whitby enlisted four major turning points in the history of the Turing test - the publication of "Machine and Computational Intelligence" in 1950, ELIZA's announcement of Joseph Weizenbaum in 1966, by Kenneth Colby of PARRY, first described in 1972, and Turing Colloquium 1990.

Colloquium 2005 on Conversation Systems

In November 2005, the University of Surrey hosted a one-day inaugural meeting of artificial conversation account developers, attended by winners of the practical Turing test in Loebner Prizes: Robby Garner, Richard Wallace, and Rollo Carpenter. Invited speakers included David Hamill, Hugh Loebner (sponsor of Loebner Prize) and Huma Shah.

2008 AISB Symposium

In line with the 2008 Loebner Prize held at the University of Reading, the Society for the Study of Artificial Intelligence and Behavioral Simulation (AISB), organized a one-day symposium to discuss the Turing test, hosted by John Barnden, Mark Bishop, Huma Shah and Kevin Warwick. Speakers include Royal Institution Director, Baroness Susan Greenfield, Selmer Bringsjord, biographer Turing Andrew Hodges, and Owen Holland consciousness scientist. No agreement emerged for the canonical Turing test, although Bringsjord stated that a considerable reward would result in a faster passing Turing test.

Year Alan Turing, and Turing100 in 2012

Throughout 2012, a number of major events occurred to celebrate Turing's life and its scientific impact. The Turing100 Group supported the event and also, hosted a special Turing test event at Bletchley Park on June 23, 2012 to celebrate the 100th anniversary of Turing's birth.

src: i.ytimg.com

See also

src: apollo2.dl.playstation.net

Note

src: i.ytimg.com

References

src: assets1.bigthink.com

Further reading

src: i.ytimg.com

External links

• The Turing Test - Opera by Julian Wagstaff
• Turing test in Curlie (based on DMOZ)
• Turing Test - How accurate is the Turing test?
• Zalta, Edward N. (ed.). "Turing Test". Stanford Encyclopedia of Philosophy .
• Turing Test: 50 Years Later on reviewing half a century of work on the Turing Test, from a 2000 point of view.
• The bet between Kapor and Kurzweil, including the detailed justification of their respective positions.
• Why The Turing Test is AI Big Blind Alley by Blay Witby
• Jabberwacky.com Chatterbot AI learning from and imitating human
• New York Times paper on part 1 and part 2 machine intelligence
• "The first (limited) Turing test", in season 2, episode 5 of the Scientific American Frontiers .
• Computer Science Teaching activities unplugged for Turing tests.
• Wiki News: "Talk: Computer professionals celebrate the 10th anniversary of A.L.I.C.E."

Source of the article : Wikipedia