Can machines think? This question was posed by the British mathematician Alan Turing in his fundamental paper “Computing Machinery and Intelligence” published in 1950. Essentially, Turing suggested that the question “Can machines think?” be replaced by a simpler question: “Can machines do what we (as thinking entities) can do?” The classical interpretation of the Turing test sounds as follows:
“A person interacts with one computer and one person. Based on answers to questions, the person is supposed to determine who he is speaking to: a person or a computer programme. The computer programme’s objective is to mislead the person into making the wrong choice.”
Those interested can read about this in a rather detailed article in Wikipedia. You can also watch an excellent (and short) documentary film made in the Soviet period.
Since then, many researchers have studied and developed programmes that simulate human interaction. Suffice it to recall the ELIZA programme that was created at MIT in the 1960s. It simulated a psychotherapist so well that some people really believed that a real psychotherapist was sitting on the other side of the wall.
Surely, the Turing Test is the subject of many science fiction stories and jokes.
At present, the Turing test is used to conduct various competitions between teams who create chatterbot programmes. For example, the Loebner Prize provides an annual platform for such competitions. Last year, Great Britain hosted many events dedicated to the 100th anniversary of the birth of Alan Turing. Held in Bletchley Park (where the British decrypted German messages during the Second World War), a large-scale Turing test contest was won by the Russian machine developer Vladimir Veselov.
At present, conversational artificial intelligence is developed for both research and commercial purposes. In Russia, chatterbot programmes have been commercially used for some years now. This area of activities is rapidly developing. Some examples of chatterbots are quite successful. Some of them can serve as mobile personal assistants: Siri (iOS) or Speaktoit (Skolkovo). Chatterbots can also serve as consultants for companies. This was implemented at Nanosemantics (Skolkovo) and Avatarisation Laboratory (Business avatars). Such creatures are even developed as new information life forms and live digital heritage (for example, Spiritual avatars.) And, surely, chatterbots are used for entertainment (for example, Russian chatterbot) There is even a specialised innovation-related chatterbot that knows everything about innovations and start-ups. If you do not believe me, check this out!
There are different areas in the field of artificial intelligence. Not all people say that chatterbots have “true intelligence.” This is arguable, because, to our ancestors born in the early 20th century, Siri or Speaktoit, for example, may sound like real people carrying on a meaningful conversation.
At Skolkovo, we would not like to discuss the place of chatterbots and intelligent assistants. Life will put everything in its place, anyway. However, to promote research in the field of semantics and commercial application of artificial intelligence, we have decided to hold a small chatterbot contest as part of the Open Innovations exhibition forum.
The contest partner is Abi InfoPoisk, a Skolkovo participant.
Below, you can find the general diagram for the contest
Based on the standard Turing test, the contest rules are as follows:
The terms of the Russian chatterbot contest based on the standard Turing test.
Participating chatterbots and hidden people are supposed to interact with experts via text messages using a standardised interface (from the point of view of an expert) 1. Any visitor of the forum Open Innovation forum may serve as an expert on a first-come-first-served basis.
2. Each chatterbot and each hidden person receives a unique random number.
3. Each expert is supposed to enter into correspondence with three interlocutors (consecutively). An expert is provided with interlocutors on a random basis. In other words, there may be any situations -– 3 machines, 3 persons, 1 person and 2 machines, etc.
4. Each expert is given no more than 5 minutes to communicate with each interlocutor, but the expert may terminate the conversation at any time.
5. The chatterbots and hidden people must continue chatting for 5 minutes, if the expert does not stop their conversation earlier.
6. The chatterbots must send messages to the expert after some time required for their typing at an average text typing speed.
7. Any communication themes, styles and tones may be used. The hidden people and the chatterbots may make typing errors deliberately or inadvertently. The hidden people and the chatterbots may not use emoticons and punctuation that signifies emoticons.
8. The hidden people are supposed to be warned about the nature of the event they participate in. It is not their objective to confuse the expert. In other words, they are supposed to communicate on behalf of a person. The chatterbots are also supposed to communicate on behalf of a person.
9. After his/her communication, the expert must fill out a form that consists of two questions: “Do you think that your partner in this session was a human or a machine?” and “If you think that your partner in this session was a machine, evaluate how humanoid its communication was on the scale from 0 to 100, where 0 is bad, 50 is good, and 100 is humanoid communication.” An answer to the first question is mandatory. An answer to the second question is optional.
Any team interested in participating in the contest is welcome to contact Albert Yefimov at aefimov@sk.ru. Deadline for registration: October 9, 2013.