20Q

20Q is a computer- based game that provides 20 questions. Originally it was an experiment for artificial intelligence (AI).

The game

The game 20Q is a device that is based on the parlor game Twenty Questions. The device asks the player to something arbitrary to think and then tries to guess it by providing 20 questions that can be answered with yes or no.

Principle and history

20Q was formed in 1988 as an experiment in artificial intelligence (AI). The principle is that the player thinks of something, the artificial intelligence provides a series of questions and advises on what the player thinks. The artificial intelligence learns independently from the information that they get delivered by the players and is not programmed. The player can answer the questions as follows: Yes, No, Unknown, or Sometimes. The experiment is based on the classic word game Twenty Questations and on the computer game "Animals", which was popular in the early 70s and used a much simpler method to guess an animal.

20Q the KI using an artificial neural network to select the questions and rates. After the player has answered 20 questions (sometimes earlier), advises 20Q. If it is wrong, it provides additional support questions and advises again. It advises solely based on the learned information; it was not fed with information or the opinion of the inventor. All answers are based on the interpretations of the player with respect to previous questions.

The 20Q AI decides independently how they utilized the information. You could rewrite rather than Laientaxonomie, because as taxonomy. My knowledge is growing with every game. In this regard, the online version of the 20 AI be inaccurate, as they tend to resort to what the people think than what they know. The boundaries of the taxonomy are often overcome by the AI itself, because they can learn and adapt. If the player, for example, to "horse" thinks and the question "Is it an animal " in the negative, the AI ​​will still guess correctly, although she was told that a horse is not an animal.

Kevin Kelly, co-founder of Wired magazine, wrote:

Translation:

As described by the inventor Robin Burgener is the " Unusual knowledge" that the AI ​​spits out at the end of the game, what comes up when something seems odd and not with the covers what she knows. This makes the AI unique - it starts to make their own decisions; these are the information on which the AI ​​itself comes by with what they have learned and what they do generate answers. Over time, their knowledge is more refined. 20Q learns and learns to make decisions by being played - the more often an object is played, the more the AI ​​learns about the object. The online 20Q AI has approximately 10,000,000 associations of ideas. Burgener writes that the success rate of online 20Q AI is 73-78 percent. According to Burgener, the real success rate could well be higher, but he adapted the algorithm to make the game more interesting; if the AI ​​would win every game as it would be theoretically feasible in a closed system, the game would not be interesting and the AI ​​would also not continue to learn.

Modularity of artificial intelligence

The modularity of the 20Q AI means that they can be integrated in small screen devices. At the moment there is a handheld version of the AI. The device contains a small fraction of the 20Q site database; In contrast to the online version of the handheld version can not learn.

The 20Q AI differs from less flexible and extremely large expert systems. Its modularity, adaptability and scalability means that you can transfer it to other, more complex devices, for more sophisticated usage.