"This book provides a framework for thinking about foundational philosophical questions surrounding the use of deep artificial neural networks (""deep learning"") to achieve artificial intelligence. Specifically, it links recent breakthroughs to classic works in empiricist philosophy of mind. In recent assessments of deep learning's potential, scientists have cited historical figures from the philosophical debate between nativism and empiricism, which concerns the origins of abstract knowledge. These empiricists were faculty psychologists; that is, they argued that the extraction of abstract knowledge from experience involves the active engagement of psychological faculties such as perception, memory, imagination, attention, and empathy. This book explains how recent deep learning breakthroughs realized some of the most ambitious ideas about these faculties from philosophers such as Aristotle, Ibn Sina (Avicenna), John Locke, David Hume, William James, and Sophie de Grouchy. It illustrates the utility of this interdisciplinary connection by showing how it can provide benefits to both philosophy and computer science: computer scientists can continue to mine the history of philosophy for ideas and aspirational targets to hit, and philosophers can see how some of the historical empiricists' most ambitious speculations can now be realized in specific computational systems."
Acknowledgments Preface Note on Abbreviated Citations to Historical Works 1 Moderate Empiricism and Machine Learning 1.1 Playing with fire? Nature vs. nurture for computer science 1.2 How to simmer things down: From Forms and slates to styles of learning 1.3 From dichotomy to continuum 1.4 Of faculties and fairness: Introducing the new empiricist DoGMA 1.5 Of models and minds 1.6 Other dimensions of the rationalist-empiricist debate 1.7 The DoGMA in relation to other recent revivals of empiricism 1.8 Basic strategy of the book: Understanding deep learning through empiricist faculty psychology 2 What is Deep Learning, and How Should We Evaluate Its Potential? 2.1 Intuitive inference as deep learning's distinctive strength 2.2 Deep learning: Other marquee achievements 2.3 Deep learning: Questions and concerns 2.4 Can we (fairly) measure success? Artificial intelligence vs. artificial rationality 2.5 Avoiding comparative biases: Lessons from comparative psychology for the science of machine behavior 2.6 Summary 3 Perception 3.1 The importance of perceptual abstraction in empiricist accounts of reasoning 3.2 Four approaches to abstraction from the historical empiricists 3.3 Transformational abstraction: Conceptual foundations 3.4 Deep convolutional neural networks: Basic features 3.5 Transformational abstraction in DCNNs 3.6 Challenges for DCNNs as models of transformational abstraction 3.7 Summary 4 Memory 4.1 The trouble with quantifying human perceptual experience 4.2 Generalization and catastrophic interference 4.3 Empiricists on the role of memory in abstraction 4.4 Artificial neural network models of memory consolidation 4.5 Deep reinforcement learning 4.6 Deep-Q Learning and Episodic Control 4.7 Remaining questions about modeling memory 4.8 Summary 5 Imagination 5.1 Imagination: The mind's laboratory 5.2 Fodor's challenges, and Hume's imaginative answers 5.3 Imagination's role in synthesizing ideas: Autoencoders and Generative Adversarial Networks 5.4 Imagination's role in synthesizing novel composite ideas: vector interpolation, variational autoencoders, and transformers 5.5 Imagination's role in creativity: Creative Adversarial Networks 5.6 Imagination's role in simulating experience: Imagination-Augmented Agents 5.7 Biological plausibility and the road ahead 5.8 Summary 6 Attention 6.1 Introduction: Bootstrapping control 6.2 Contemporary theories of attention in philosophy and psychology 6.3 James on attention as ideational preparation 6.4 Attention-like mechanisms in DNN architectures 6.5 Language models, self-attention, and transformers 6.6 Interest and innateness 6.7 Attention, inner speech, consciousness, and control 6.8 Summary 7 Social and Moral Cognition 7.1 From individual to social cognition 7.2 Social cognition as Machiavellian struggle 7.3 Smith and De Grouchy's sentimentalist approach to social cognition 7.4 A Grouchean developmentalist framework for modeling social cognition in artificial agents 7.5 Summary Epilogue References Index
Cameron J. Buckner is an Associate Professor in the Department of Philosophy at the University of Houston. He received an Alexander von Humboldt Postdoctoral Fellowship at Ruhr-University Bochum from 2011 to 2013 and has been a visiting fellow at the University of Cambridge.
Reviews for From Deep Learning to Rational Machines: What the History of Philosophy Can Teach Us about the Future of Artificial Intelligence
"AI, in the form of in deep learning, is emerging as one of the most transformative technologies of our time. Cameron Buckner provides an extremely useful framework for assessing its contributions and pitfalls. He frames debates over deep learning in terms of the history of philosophical debates between empiricism and rationalism and develops and defends a moderate empiricism that offers an illuminating perspective from which to understand and evaluate the claims and counterclaims about AI's prospects. Non-AI researchers will acquire valuable tools for engaging AI while AI researchers will find insightful suggestions for advancing their endeavor. * William Bechtel, University of California, San Diego * This terrific book is packed full of insights. Based on a deep understanding of deep neural networks, it showcases a variety of ways in which these computational models illuminate aspects of the human mind. Buckner's conclusions will be of interest to researchers from across the cognitive sciences. His accessible treatment will also be useful to philosophers more broadly * in any fieldwho want to understand what's important about this revolutionary new technology.Nicholas Shea, Institute of Philosophy, University of London * It is both exciting and alarming that areas once considered the exclusive zone of human rationality are rapidly being conquered by new forms of artificial intelligence. The astonishing success of deep neural networks, in everything from strategic gaming to natural language, raises questions about whether these new machines are rational, and whether we are at some level similar machines ourselves. For anyone searching for answers to such questions, Cameron Buckner is your best possible guide: he delivers a clear explanation of the crucial technical features of contemporary AI, together with a profound philosophical analysis of the relationship between innate structure and experience. This book marks a new stage in the human understanding of rationality."" -Jennifer Nagel, University of Toronto"
