Why I take silly facebook quizzes and how it relates to artificial intelligence
I noticed a facebook friend took "What kind of wife you would make?", and based on her comment of "Gross" (she got the result of "Loving"), I couldn't resist.
The facebook quiz is based on a graded answer scale. Each answer has an assigned value towards a certain result, and at the end of the quiz, the values are added together and whichever result total is higher, that's your answer.
Anyone who has taken one of these quizzes knows its limitations. The worst question is "what's your favorite kind of music?" The five or six choices rarely encompass most people's music tastes ("Why can't I choose Norwegian Black Metal?")
These quizzes reveal the travails of any programmer, specifically for games. A programmer anticipates the most likely outcomes and prepares appropriate responses, covered by all-inclusive "ELSE" statement.
However, the QA repetition testing done by your typical twelve year-old will reveal any programming bugs, like the Galaga cheat that stops the descending insectoids from firing. That was not an intended response by the Galaga programmers.
Computer games today are typically created by large gaming houses, with an enormous cast of game designers, coders, artists, and testers. Now the easter eggs and the cheats are a planned part of the game.
But the limitations of these games still evident. Take Grand Theft Auto: it is a free flowing world, but if you go past the borders of the game, you are either blocked, forced to turn back, looped around, or you die. Obviously, storage space and game flow is the real roadblock.
Personally, I like using objects that test the game physics, specifically grenades. Bouncing it off a wall will exercise several game calculators, like the code for object reflection and area effect damage. And I could bounce that grenade back at my feet, an unintended response by the game player!
The ultimate test of inclusive game design, the goal of having as many of the possible responses covered, is Deep Blue, the IBM chess computer which was the first machine to defeat Garry Kasparov, the world champion.
Deep Blue's program included 700,000 grandmaster games, and it still required tuning by another grandmaster during its series of matches against Kasparov in order to defeat him.
Gaming is probably the most common use of artificial intelligence (actually aspects of artificial intelligence include computer mice to Amazon recommendations, the term is falling out of usage), since games directly involve human interaction. Most programs, from http server software to operating systems, interact only with other computer programs. Games are built around the human player.
The term often used in game design is "illusion of intelligence". A program can include aspects of human deduction and learning, but are still limited to the architecture of the code, circuits, and memory.
Besides the IBM programmers stacking the deck to defeat Kasparov, like withholding game results and logs, the real reason why Deep Blue was able to beat him was its raw computing power in application to evaluating positions deep into the game.
Chess machines do an excellent job storing opening games and selecting appropriate moves. However as the game progresses, it was previously impossible to make the calculations to stay competitive with grandmaster-level players. There was simply too many possibilities for a computer to make the best decision.
So how are human grandmasters able to make these endgame calculations? They can't. They rely on intuition to decide their final moves.
Deep Blue was able to simulate human intuition by being powerful enough to run through enough scenarios to make winning moves against Kasparov, although the mid-match tweaks by IBM's human grandmaster could have a simple code designed specifically against Kasparov's style of player.
As computers continue to increase in computing power and storage, they will be able to simulate human intelligence well enough to fool a real person, especially as interaction devices like voice modulators become more human-like. These computers will be able to cover almost all interaction with an appropriate response.
And the answer for "What Kind Of Wife You Would Make", was "Humble/Supportive".
The facebook quiz is based on a graded answer scale. Each answer has an assigned value towards a certain result, and at the end of the quiz, the values are added together and whichever result total is higher, that's your answer.
Anyone who has taken one of these quizzes knows its limitations. The worst question is "what's your favorite kind of music?" The five or six choices rarely encompass most people's music tastes ("Why can't I choose Norwegian Black Metal?")
These quizzes reveal the travails of any programmer, specifically for games. A programmer anticipates the most likely outcomes and prepares appropriate responses, covered by all-inclusive "ELSE" statement.
However, the QA repetition testing done by your typical twelve year-old will reveal any programming bugs, like the Galaga cheat that stops the descending insectoids from firing. That was not an intended response by the Galaga programmers.
Computer games today are typically created by large gaming houses, with an enormous cast of game designers, coders, artists, and testers. Now the easter eggs and the cheats are a planned part of the game.
But the limitations of these games still evident. Take Grand Theft Auto: it is a free flowing world, but if you go past the borders of the game, you are either blocked, forced to turn back, looped around, or you die. Obviously, storage space and game flow is the real roadblock.
Personally, I like using objects that test the game physics, specifically grenades. Bouncing it off a wall will exercise several game calculators, like the code for object reflection and area effect damage. And I could bounce that grenade back at my feet, an unintended response by the game player!
The ultimate test of inclusive game design, the goal of having as many of the possible responses covered, is Deep Blue, the IBM chess computer which was the first machine to defeat Garry Kasparov, the world champion.
Deep Blue's program included 700,000 grandmaster games, and it still required tuning by another grandmaster during its series of matches against Kasparov in order to defeat him.
Gaming is probably the most common use of artificial intelligence (actually aspects of artificial intelligence include computer mice to Amazon recommendations, the term is falling out of usage), since games directly involve human interaction. Most programs, from http server software to operating systems, interact only with other computer programs. Games are built around the human player.
The term often used in game design is "illusion of intelligence". A program can include aspects of human deduction and learning, but are still limited to the architecture of the code, circuits, and memory.
Besides the IBM programmers stacking the deck to defeat Kasparov, like withholding game results and logs, the real reason why Deep Blue was able to beat him was its raw computing power in application to evaluating positions deep into the game.
Chess machines do an excellent job storing opening games and selecting appropriate moves. However as the game progresses, it was previously impossible to make the calculations to stay competitive with grandmaster-level players. There was simply too many possibilities for a computer to make the best decision.
So how are human grandmasters able to make these endgame calculations? They can't. They rely on intuition to decide their final moves.
Deep Blue was able to simulate human intuition by being powerful enough to run through enough scenarios to make winning moves against Kasparov, although the mid-match tweaks by IBM's human grandmaster could have a simple code designed specifically against Kasparov's style of player.
As computers continue to increase in computing power and storage, they will be able to simulate human intelligence well enough to fool a real person, especially as interaction devices like voice modulators become more human-like. These computers will be able to cover almost all interaction with an appropriate response.
And the answer for "What Kind Of Wife You Would Make", was "Humble/Supportive".
http://www.latimes.com/news/nationworld/nation/la-na-obama-facebook29-2009sep29,0,6007621.story
ReplyDeleteLooks like facebook quizzes are in the news...