Artificial Intelligence and Scientific Method

by Donald Gillies

Overview

Artificial Intelligence and Scientific Method examines the remarkable advances made in the field of AI over the past twenty years, discussing their profound implications for philosophy. Taking a clear, non-technical approach, Donald Gillies focuses on two key topics within AI: machine learning in the Turing tradition and the development of logic programming and its connection with non-monotonic logic. Demonstrating how current views on scientific method are challenged by this recent research, he goes on to suggest a new framework for the study of logic. He draws on work by such seminal thinkers as Bacon, Gödel, Popper, Penrose, and Lucas to address the hotly contested question of whether computers might become intellectually superior to human beings. These topics will attract a wide readership from followers of advances in artificial intelligence, to students and scholars of the history and philosophy of science.

Product Details

ISBN-13:	9780198751595
Publisher:	Oxford University Press
Publication date:	09/26/1996
Edition description:	New Edition
Pages:	192
Product dimensions:	7.30(w) x 11.20(h) x 0.90(d)

About the Author

Donald Gillies is Professor of the Philosophy of Science and Mathematics at King's College, London. His books include An Objective Theory of Probability (1973), Revolutions in Mathematics (1992), and Philosophy of Science in the Twentieth Century (1993). He was the editor of the British Journal for the Philosophy of Science from 1982 to 1985.

Table of Contents

Chapter 1: The Inductivist Controversy, or Bacon vs. Popper1.1. Bacon's Inductivism1.2. Popper's Falsificationism1.3. Kepler's Discovery of the Laws of Planetary Motion1.4. The Discovery of the Sulphonamide DrugsChapter 2: Machine Learning in the Turing Tradition2.1. The Turing Tradition2.2. The Practical Problem: Expert Systems and Feigenbaum's Bottleneck2.3. Attribute-Based Learning, Decision Trees, and Quinlan's ID32.4. GOLEM as an example of Relational Learning2.5. Bratko's Summary of the Successes of Machine Learning in the Turing Tradition, 19922.6 GOLEM's Discovery of a Law of Nature. Chapter 3: How Advances in Machine Learning Affect the Inductivist Controversy3.1. Bacon's Example of Heat3.2. The Importance of Falsification3.3. Bacon's Method has only recently come to be used3.4. The Need for Background KnowledgeChapter 4: Logic and Programming and a New Framework for Logic4.1. The Development of PROLOG4.2. PROLOG as a Non-Monotonic Logic4.3. Two Examples of Translations from One Logical System to Another4.4. Logic = Inference + Control4.5. PROLOG Introduces Control into Deductive Logic4.6. PROLOG and Certainty: Is Logic a priori or empirical?Chapter 5: Can There Be an Inductive Logic?5.1. The Divergence Between Deductive and Inductive Logic (up to the early 70's)5.2. Inductive Logic as Inference + Control5.3. Confirmation Values as Control in a Deductive Logic5.4. The Empirical Testing of Rival LogicsChapter 6: Do G/D"odel's Incompleteness Theorems Place a Limit on Artificial Intelligence?6.1. Anxietites Caused by Advances in AI6.2. Informal Exposition of G/D"odel's Imcompleteness Theorems6.3. The Lucas Argument6.4. Objections to the Lucas Argument: i)Possible Limitations on Self-Knowledge6.5. Objections to the Lucas Argument ii) Possible Additions of Learning Systems6.6. Why Advances in Computing are More Likely to Stimulate Human Thinking than to Render it Superfluous

Editorial Reviews

"Although a machine can perform deduction based on available knowledge, the real question is whether a machine can genuinely perform induction based on observations alone. The author presents a number of practical developments in nontechnical terms to illustrate the arguments for and against the induction approach. . . A very easy-to-read book for those more inclined toward philosophical perspectives." —Choice

"In this book Donald Gillies, a practitioner of the philosophy of science who in recent years has become interested in the field of artificial intelligence, discusses some of the implications each of these two disciplines holds for the other. . . . The work is best regarded as an extended essay, touching on a number of topics of particular interest to its author, arguing in a nondogmatic fashion for a particular point of view. Thus conceived it is quite successful. Informative, clearly written, carefully organized, and free of jargon, it could well serve as the basis for an undergraduate or graduate seminar supplemented by papers from the original journal literature."—ISIS

From the Publisher