Knowing our World: An Artificial Intelligence Perspective considers the methodologies of science, computation, and artificial intelligence to explore how we humans come to understand and operate in our world. While humankind’s history of articulating ideas and building machines that can replicate the activity of the human brain is impressive, Professor Luger focuses on understanding the skills that enable these goals.
Based on insights afforded by the challenges of AI design and program building, Knowing our World proposes a foundation for the science of epistemology. Taking an interdisciplinary perspective, the book demonstrates that AI technology offers many representational structures and reasoning strategies that support clarification of these epistemic foundations.
This monograph is organized in three Parts; the first three chapters introduce the reader to the foundations of computing and the philosophical background that supports the AI tradition. These three chapters describe the origins of AI, programming as iterative refinement, and the representations and very high-level language tools that support AI application building.
The book’s second Part introduces three of the four paradigms that represent research and development in AI over the past seventy years: the symbol-based, connectionist, and complex adaptive systems. Luger presents several introductory programs in each area and demonstrates their use.
The final three chapters present the primary theme of the book: bringing together the rationalist, empiricist, and pragmatist philosophical traditions in the context of a Bayesian world view. Luger describes Bayes' theorem with a simple proof to demonstrate epistemic insights. He describes research in model building and refinement and several philosophical issues that constrain the future growth of AI. The book concludes with his proposal of the epistemic stance of an active, pragmatic, model-revising realism.
Author(s): George F. Luger
Publisher: Springer
Year: 2021
Language: English
Pages: 274
City: Cham
Preface
Why Write This Book?
The Story
Acknowledgments
Contents
Part I: In the Beginning…
Chapter 1: Creating Computer Programs: An Epistemic Commitment
1.1 Introduction and Focus of Our Story
1.2 The Foundation for Computation
1.2.1 The Turing Machine
1.2.2 The Post Production System and Unary Subtraction
1.3 Computer Languages, Representations, and Search
1.4 In Summary
Chapter 2: Historical Foundations
2.1 Mary Shelley, Frankenstein, and Prometheus
2.2 Early Greek Thought
2.3 The Later Greeks: Plato, Euclid, and Aristotle
2.4 Post-medieval or Modern Philosophy
2.5 The British Empiricists: Hobbes, Locke, and Hume
2.6 Bridging the Empiricist/Rationalist Chasm: Baruch Spinoza
2.7 Bridging the Empiricist/Rationalist Chasm: Immanuel Kant
2.8 American Pragmatism: Peirce, James, and Dewey
2.9 The Mathematical Foundations for Computation
2.10 The Turing Test and the Birth of AI
2.11 In Summary
Chapter 3: Modern AI and How We Got Here
3.1 Three Success Stories in Recent AI
3.1.1 Deep Blue at IBM (Hsu 2002; Levy and Newborn 1991; url 3.2)
3.1.2 IBM’s Watson (Baker 2011, Ferrucci et al. 2010, 2013, url 3.3)
3.1.3 Google and AlphaGo (Silver et al. 2017, url 3.4)
3.2 Very Early AI and the 1956 Dartmouth Summer Research Project
3.2.1 The Logic Theorist (Newell and Simon 1956; Newell et al. 1958)
3.2.2 Geometry Theorem Proving (Gelernter 1959; Gelernter and Rochester 1958)
3.2.3 A Program that Plays Checkers (Samuel 1959)
3.2.4 The Dartmouth Summer Workshop in 1956
3.3 Artificial Intelligence: Attempted Definitions
3.4 AI: Early Years
3.4.1 The Neats and Scruffies
3.4.2 AI: Based on “Emulating Humans” or “Just Good Engineering?”
3.5 The Birth of Cognitive Science
3.6 General Themes in AI Practice: The Symbolic, Connectionist, Genetic/Emergent, and Stochastic
3.7 In Summary
Part II: Modern AI: Structures and Strategies for Complex Problem-Solving
Chapter 4: Symbol-Based AI and Its Rationalist Presuppositions
4.1 The Rationalist Worldview: State-Space Search
4.1.1 Graph Theory: The Origins of the State Space
4.1.2 Searching the State Space
4.1.3 An Example of State-Space Search: The Expert System
4.2 Symbol-Based AI: Continuing Important Contributions
4.2.1 Machine Learning: Data Mining
4.2.2 Modeling the Physical Environment
4.2.3 Expertise: Wherever It Is Needed
4.3 Strengths and Limitations of the Symbol System Perspective
4.3.1 Symbol-Based Models and Abstraction
4.3.2 The Generalization Problem and Overlearning
4.3.3 Why Are There No Truly Intelligent Symbol-Based Systems?
4.4 In Summary
Chapter 5: Association and Connectionist Approaches to AI
5.1 The Behaviorist Tradition and Implementation of Semantic Graphs
5.1.1 Foundations for Graphical Representations of Meaning
5.1.2 Semantic Networks
5.1.3 More Modern Uses of Association-Based Semantic Networks
5.2 Neural or Connectionist Networks
5.2.1 Early Research: McCulloch, Pitts, and Hebb
5.2.2 Backpropagation Networks
5.3 Neural Networks and Deep Learning
5.3.1 AlphaGo Zero and Alpha Zero
5.3.2 Robot Navigation: PRM-RL
5.3.3 Deep Learning and Video Games
5.3.4 Deep Learning and Natural Language Processing
5.4 Epistemic Issues and Association-Based Representations
5.4.1 Inductive Bias, Transparency, and Generalization
5.4.2 Neural Networks and Symbol Systems
5.4.3 Why Have We Not Built a Brain?
5.5 In Summary
Chapter 6: Evolutionary Computation and Intelligence
6.1 Introduction to Evolutionary Computation
6.2 The Genetic Algorithm and Examples
6.2.1 The Traveling Salesperson Problem
6.2.2 Genetic Programming
6.2.3 An Example: Kepler’s Third Law of Planetary Motion
6.3 Artificial Life: The Emergence of Complexity
6.3.1 Artificial Life
6.3.2 Contemporary Approaches to A-Life
Synthetic Biological Models of Evolution
Artificial Chemistry
Other Abstract Machines and Evolutionary Computation
Psychological and Sociological Foundations for Life and Intelligence
6.4 Evolutionary Computation and Intelligence: Epistemic Issues
6.5 Some Summary Thoughts on Part II: Chaps. 4, 5, and 6
6.5.1 Inductive Bias: The Rationalist’s a priori
6.5.2 The Empiricist’s Dilemma
6.6 In Summary
Part III: Toward an Active, Pragmatic, Model-Revising Realism
Chapter 7: A Constructivist Rapprochement and an Epistemic Stance
7.1 A Response to Empiricist, Rationalist, and Pragmatist AI
7.2 The Constructivist Rapprochement
7.3 Five Assumptions: A Foundation for an Epistemic Stance
7.4 A Foundation for a Modern Epistemology
7.5 In Summary
Chapter 8: Bayesian-Based Constructivist Computational Models
8.1 The Derivation of a Bayesian Stance
8.2 Bayesian Belief Networks, Perception, and Diagnosis
8.3 Bayesian-Based Speech and Conversation Modeling
8.4 Diagnostic Reasoning in Complex Environments
8.5 In Summary
Chapter 9: Toward an Active, Pragmatic, Model-Revising Realism
9.1 A Summary of the Project
9.2 Model Building Through Exploration
9.3 Model Revision and Adaptation
9.4 What Is the Project of the AI Practitioner?
9.5 Meaning, Truth, and a Foundation for a Modern Epistemology
9.5.1 Neopragmatism, Kuhn, Rorty, and the Scientific Method
9.5.2 A Category Error
9.5.3 The Cognitive Neurosciences: Insights on Human Processing
9.5.4 On Being Human: A Modern Epistemic Stance
Bibliography
URL References (All url references checked 16 June 2021)
Index
