November 28, 2005

Playing at the edge of AI



"Blondie24 tells the story of a computer that taught itself to play checkers far better than its creators ever could by using a program that emulated the basic principles of Darwinian evolution--random variation and natural selection-- to discover on its own how to excel at the game."
-from the back cover of the book

Thanks to Hollywood movies, I need not tell you what is Artificial Intelligence other than that it is a simple thought that what if you can teach intelligence to machines (Thinking Machines!). Some people argue that Artificial Intelligence is a paradox (oxymoron?) whereas others still dream of creating a 'Digital Einstein' - who (or which?) can really 'think' and discover new wisdom!

'Blondie24:Playing at the edge of AI' written by Chief Scientist, David Fogel, Ph.D is a good book which deals with a bottom-up approach application of AI. He argues that the famous 'Deep Blue', IBM computer which beat Grandmaster Gary Kasparov had no real intelligence embedded in it. The super computer 'Deep Blue' was created with a great collection of successful moves (there were over 4000 positions and 700000 grandmaster games including strategies from Gary Kasparov himself!), extra-ordinary processing speed (capable of evaluating 200,000,000 positions per second) just to defeat one person in the game. Well, after providing extensive cheat-sheets, Deep Blue lost its firs meet. Gary won by 4-2. After doing some homework, the IBM team managed to beat him again in 1997.

Well, what a real intelligent machine do, then. Like a child, learn a game on its own, beat a human grandmaster. David Fogel and Chellapillai Kumar (Yes, an Indian computer scientist) used 'Evolutionary Algorithms' and initiated a computer program to learn Checkers game. It works like a natural process of evolution (survival of fittest), where 'best survives' - offsprings which play well are taken and crossed to give birth to better players. This goes on and on ..till we get a best algorithm which can beat a human being.


I completely agree with the fact that "future of artificial intelligence lies . . . in programs that can automatically improve themselves over time--without the bias of human knowledge". This unbiased nature is very essential in exploring the space for life or finding a medicine for diseases (Sometimes, we human being always use our 'prejudice' and 'common sense' to delimit important facts).

Fascinating read!

If you want a detailed review, find here.

No comments: