6. Search: Games, Minimax, and Alpha-Beta

The paragraph discusses the early skepticism about computers playing chess, exemplified by Hubert Dreyfus's 1963 paper. Despite initial doubts, Dreyfus lost to a chess machine, and later, David Levy's bet against AI founder John McCarthy highlighted the evolving capabilities of computers in chess. The narrative then shifts to the significance of understanding different types of intelligence, including that demonstrated by game-playing programs, which led to the development of various approaches to designing computer chess programs.

This section delves into the methods used by computers to play chess, starting with the human-like approach of analyzing the board's strategic elements. It acknowledges the challenge of replicating human thought processes and moves on to discuss alternative approaches such as rule-based systems and lookahead evaluation. The paragraph introduces the concept of static evaluation, which assesses board positions without considering future moves, and explains how linear scoring polynomials are used to assign values to board states.

The speaker introduces the concept of searching through a game tree to find the best move in chess, using the British Museum algorithm as an example of a brute force approach. The discussion highlights the branching factor and depth of the search tree, and the calculation of leaf nodes. It also touches on the impracticality of evaluating every possible move due to the vast number of combinations in chess, even with modern computing power.

The paragraph explores the impracticality of using the British Museum algorithm for chess due to the astronomical number of leaf nodes. It humorously calculates the number of nanoseconds in the universe's history and concludes that even if all atoms in the universe were evaluating positions since the Big Bang, it would not be sufficient. This leads to the necessity of using heuristic methods like lookahead search, which is a more efficient way of evaluating potential moves.

This section explains the minimax algorithm, a recursive method for minimizing the maximum possible loss in zero-sum games like chess. It describes the algorithm's process of evaluating game trees and making decisions based on the maximizing and minimizing players' perspectives. The paragraph uses a simplified game tree to illustrate how the algorithm works, emphasizing its simplicity and effectiveness in decision-making.

The paragraph introduces the alpha-beta algorithm as an enhancement to the minimax algorithm, designed to reduce the number of nodes evaluated in the search tree. It explains how alpha-beta pruning works by setting bounds (alpha and beta) to eliminate branches that cannot possibly influence the final decision. The speaker uses a detailed example to demonstrate how the algorithm skips unnecessary evaluations, thus saving computational resources.

This section discusses the concept of progressive deepening, where the search algorithm starts with a shallow search and gradually increases the depth, ensuring that a move is always available within a given time constraint. The paragraph explains the insurance policy approach, where calculations are made at each level as a backup, and how it aligns with the anytime algorithm concept, which always has a ready answer.

The speaker reflects on the evolution of game-playing programs, highlighting the use of minimax, alpha-beta pruning, and progressive deepening as foundational techniques. The paragraph also touches on the importance of adapting these techniques to the specific characteristics of the game being played, such as varying branching factors and the need for insurance policies at different levels of search.

The final paragraph contrasts the raw computational power of Deep Blue, which used parallel computing and uneven tree development to evaluate positions deeply, with the more pattern-recognition-based approach of human chess players. It questions whether Deep Blue's method of playing chess is a model for human intelligence or a different form of intelligence altogether, suggesting that while it demonstrates capability, it lacks the sophistication of human strategic thought.