By David B. Fogel
Blondie24 tells the tale of a working laptop or computer that taught itself to play checkers much better than its creators ever may possibly through the use of a software that emulated the elemental rules of Darwinian evolution--random version and average selection-- to find by itself easy methods to excel on the video game. in contrast to Deep Blue, the distinguished chess computer that beat Garry Kasparov, the previous global champion chess participant, this evolutionary application did not have entry to techniques hired via human grand masters, or to databases of strikes for the endgame strikes, or to different human services in regards to the video game of chekers. With simply the main rudimentary details programmed into its "brain," Blondie24 (the program's web username) created its personal technique of comparing the advanced, altering styles of items that make up a checkers online game by way of evolving synthetic neural networks---mathematical types that loosely describe how a mind works.It's becoming that Blondie24 should still look in 2001, the 12 months once we take note Arthur C. Clarke's prediction that someday we'd reach making a considering computer. during this compelling narrative, David Fogel, writer and co-creator of Blondie24, describes in convincing aspect how evolutionary computation will help to convey us toward Clarke's imaginative and prescient of HAL. alongside the way in which, he provides readers an inside of inspect the attention-grabbing historical past of AI and poses provocative questions on its destiny. * Brings probably the most fascinating parts of AI study to existence through following the tale of Blondie24's improvement within the lab via her evolution into an expert-rated checkers participant, in accordance with her amazing luck in web competition.* Explains the rules of evolutionary computation, easily and clearly.* offers advanced fabric in an interesting kind for readers without heritage in laptop technology or man made intelligence.* Examines foundational matters surrounding the production of a pondering machine.* Debates even if the recognized Turing try rather checks for intelligence.* demanding situations deeply entrenched myths in regards to the successes and implication of a few famous AI experiments * indicates Blondie's strikes with checkerboard diagrams that readers can simply keep on with.
Read or Download Blondie24: Playing at the Edge of AI (The Morgan Kaufmann Series in Artificial Intelligence) PDF
Best intelligence & semantics books
During this first version ebook, tools are mentioned for doing inference in Bayesian networks and inference diagrams. enormous quantities of examples and difficulties let readers to understand the knowledge. a few of the issues mentioned contain Pearl's message passing set of rules, Parameter studying: 2 possible choices, Parameter studying r choices, Bayesian constitution studying, and Constraint-Based studying.
This hole. In 16 survey articles an important theoretical effects, algorithms and software program tools of computing device algebra are lined, including systematic references to literature. furthermore, a few new effects are offered. therefore the quantity could be a precious resource for acquiring a primary impact of machine algebra, in addition to for getting ready a working laptop or computer algebra direction or for complementary interpreting.
Freeman and Skapura offer a realistic advent to man made neural platforms (ANS). The authors survey the most typical neural-network architectures and convey how neural networks can be utilized to resolve genuine medical and engineering difficulties and describe methodologies for simulating neural-network architectures on conventional electronic computing platforms
- Graphical Models for Machine Learning and Digital Communication
- Information Modelling and Knowledge Bases XVII
- An Information-Theoretic Approach to Neural Computing
- Recent Advances on Hybrid Approaches for Designing Intelligent Systems
- Inside Versus Outside: Endo- and Exo-Concepts of Observation and Knowledge in Physics, Philosophy and Cognitive Science
- Computer Chess and Search
Additional info for Blondie24: Playing at the Edge of AI (The Morgan Kaufmann Series in Artificial Intelligence)
One choice is to continue emulating specific manifestations of intelligent behavior as we observe them in ourselves (or even in ants, dogs, or other living creatures). This approach seeks to model the end products of a long process of evolution without asking why or how those products emerged. Details become important for their own sake. This "bottom-up" approach poses a great potential for reversing cause and effect. For example, suppose we wanted to design a flying machine. We might look to nature for inspiration and see a vast array of feathered birds flapping their wings.
Yet the basic design for chess programs has remained almost constant. It's fitting, even revealing, to return to the earliest efforts ZO S E T T I N G T H E STAGE at designing chess programs and to understand their essential elements. In 195o Claude Shannon, one of the founders of the branch of mathematics known as information theory, was also one of the first researchers to propose an automatic method of playing chess (although he didn't actually implement a chess program). 5 Shannon suggested that a mechanical algorithm could play the game if that algorithm involved two facets: an evaluation f u n c t i o n ~ a mathematical formula that assessed alternative features about different board positions and a rationale, which he called "minimax," that sought to minimize the maximum damage that the opponent could do in any situation.
The formulas reward you for winning and penalize you for losing, but they do so in proportion to the presumed likelihood that you were going to win. Your rating will simultaneously fall about three points. But if you're fortunate enough to win the match, your rating will increase twenty-nine points, and 24 S E T T I N G T H E STAGE your opponent's rating will plummet by the same amount. Figure 3 shows how many points you could expect to earn if you had a 1,5oo rating and defeated someone with a rating varying from I,Ioo to 1,9oo.