20Q
From Wikipedia, the free encyclopedia
20Q is an experiment in Artificial Intelligence (A.I.)
Contents |
[edit] Principle and history
The 20Q artificial intelligence was created in 1988 as an experiment in artificial intelligence (A.I.) The principle is that the player thinks of something and the 20Q artificial intelligence asks a series of questions before guessing what the player is thinking. This artificial intelligence learns on its own, and is not programmed with the information relayed back to the players who interact with it. The player can answer these questions with: Yes, No, Unknown, or Sometimes. The experiment is based on the classic word game of "Twenty Questions," and on the computer game "Animals," popular in the early 1970s, which used a somewhat simpler method to guess an animal. [1]
The 20Q A.I. uses a true artificial neural network to pick the questions and to guess. After the player has answered the twenty questions posed (sometimes fewer), 20Q makes a guess. If it is incorrect, it asks more questions, then guesses again. It makes guesses based on what it has learned; it is not “programmed” with information or what the inventor thinks. Answers to any question are based on players’ interpretations of the questions asked.
The 20Q A.I. can draw its own conclusions on how to interpret the information. It can be described as more of a folk taxonomy than a taxonomy. Its knowledge develops with every game played. In this regard, the online version of the 20Q A.I. can be inaccurate due to the fact that it gathers its answers from what people think rather than from what people know. Limitations of taxonomy are often overcome by the A.I. itself because it can learn and adapt. For example, if the player was thinking of a "Horse" and answered "No" to the question "Is it an animal?," the A.I. will, nevertheless, guess correctly, despite being told that a horse is not an animal.
Kevin Kelly, co-founder of Wired magazine, notes: “Burned into its 8-bit chip is a neural net that has been learning for 17 years. Inventor Robin Burgener programmed a simple neural net on a DOS machine 1988. He taught it 20 questions about a cat. He than passed the program around to friends on a floppy and had them challenge the neural net with their yes/no answers to the object they had in mind. The neural net learns only when it plays a game; no data is added except for the yes/no answers of visitors. So the more people who test it, the more they teach it. In 1995 Burgener put the now robust neural net onto the new web where anyone could play it (that is, train it) 24 hours a day. And they did. Burgener's genius was to turn the hard tedious work of training a neural net into a fun game for humans.
The 20Q A.I. is adaptable, scalable, modular and embeddable, and for this reason, it is possible for it to expand its knowledge and learn about more specific things. It is now learning in twenty-one languages, as well as everything it can about music, sports, movies and television.” [2]
As decribed by the inventor, Robin Burgener, the “Uncommon Knowledge” generated by the A.I. at the end of a game is what it comes up with when something seems odd and it can’t fit it in with what it knows. This makes this A.I. unique—it is beginning to make its own distinctions; it is information the A.I. ‘thinks up’ on its own, generating answers based on what it has learned and what it knows. Over time, its knowledge will become more refined. 20Q learns, and learns to make distinctions, through play—the more times an object is played, the more the artificial intelligence learns about that object. The online 20Q A.I. has about 10,000,000 synaptic connections.
The 20Q A.I. was invented in 1988, by Robin Burgener, of Ottawa, Canada. Having been on the internet since 1995, the A.I. learns on a broader level than it did in 1988 when it lived on a floppy disk that Mr. Burgener swapped amongst friends to help it learn. Originally, it knew one object and one question. After it moved to the internet, and more people started playing, its level of learning increased, as well as the connections it can make when guessing. Mr. Burgener notes that the success rate of the online A.I. is between 73 to 78 per cent. According to Mr. Burgener, the "real" success rate is higher, but he has adapted the algorithm in order to make it more interesting for the players; if the artificial intelligence won every game, all the time, as it is capable of doing, it would not be very interesting to play, and the A.I. would not continue to learn. [3]
[edit] Modularity of the artificial intelligence
The modular capability of the 20Q artificial intelligence means that it can be embedded in small screen devices. Currently, there is a handheld version of the A.I. The device contains a small portion of the original 20Q website knowledgebase; unlike the online versions of the game, the handheld version does not have the ability to learn.
The 20Q artificial intelligence is different from less flexible, and extremely large, expert systems. Its modularity, adaptability and scalability means that it can be applied to other, more complex devices, for more complex uses.
“Robin Burgener wants to turn child's play into rocket science. When he speaks to a room full of NASA scientists, programmers and technicians [in May of 2006], he'll explain how a simple parlour game he first adapted into a computer program 20 years ago might just be the answer to some of the agency's most pressing issues.” [4]
[edit] References
- ^ with information from: LiCalzi O'Connell, Pamela. "Vegetables And Minerals On The Radar" The New York Times. March 27, 2003; Burgener, Robin, computer architect, inventor.
- ^ Kelly, Kevin. “Cool Tools.” [1]. May, 10, 2005.
- ^ Burgener, Robin. Computer architect. All numbers are from data provided by the inventor.
- ^ Harvey, Ian. “Is it animal, vegetable or artificial intelligence.” The Globe and Mail, Toronto, Canada. May, 2, 2006.
