Making AI Intelligible: Philosophical Foundations

✍ Scribed by Herman Cappelen, Josh Dever

• Publisher: Oxford University Press
• Year: 2021
• Tongue: English
• Leaves: 184
• Category: Library

✦ Synopsis

Can humans and artificial intelligences share concepts and communicate? Making AI Intelligible shows that philosophical work on the metaphysics of meaning can help answer these questions. Herman Cappelen and Josh Dever use the externalist tradition in philosophy to create models of how AIs and humans can understand each other. In doing so, they illustrate ways in which that philosophical tradition can be improved.

The questions addressed in the book are not only theoretically interesting, but the answers have pressing practical implications. Many important decisions about human life are now influenced by AI. In giving that power to AI, we presuppose that AIs can track features of the world that we care about (for example, creditworthiness, recidivism, cancer, and combatants). If AIs can share our concepts, that will go some way towards justifying this reliance on AI. This ground-breaking study offers insight into how to take some first steps towards achieving Interpretable AI.

✦ Table of Contents

Cover
Making AI Intelligible Philosophical Foundations: Philosophical Foundations
Copyright
Contents
Part I: Introudction and Overview
Chapter 2: Alfred (the Dismissive Sceptic): Philosophers, Go Away!
A Dialogue with Alfred (the Dismissive Sceptic)
Part II: A Proposal for how to Attribute Content to AI
Chapter 3: Terminology: Aboutness, Representation, and
Metasemantics
Loose Talk, Hyperbole, or ‘Derived Intentionality’?
Aboutness and Representation
AI, Metasemantics, and the Philosophy of Mind
Chapter 4: Our Theory: De-Anthropocentrized
Externalism
First Claim: Content for AI Systems Should Be Explained Externalistically
Second Claim: Existing Externalist Accounts of Content Are Anthropocentric
Third Claim: We Need Meta-Metasemantic Guidance
A Meta-Metasemantic Suggestion: Interpreter-centric Knowledge-Maximization
Chapter 5: Application: The Predicate ‘High Risk’
The Background Theory: Kripke-Style
Externalism
Starting Thought: SmartCredit Expresses High Risk Contents Because of its Causal History
Anthropocentric Abstraction of ‘Anchoring’
Schematic AI-Suitable Kripke-Style Metasemantics
Complications and Choice Points
Taking Stock
Appendix to Chapter 5: More on Reference Preservation in ML Systems
Chapter 6: Application: Names and the Mental Files Framework
Does SmartCredit Use Names?
The Mental Files Framework to the Rescue?
Epistemically Rewarding Relations for Neural Networks?
Case Studies, Complications, and Reference Shifts
Taking Stock
Chapter 7: Application: Predication and Commitment
Predication: Brief Introduction to the Act Theoretic View
Turning to AI and Disentangling Three Different Questions
The Metasemantics of Predication: A Teleofunctionalist Hypothesis
Some Background: Teleosemantics and Teleofunctional Role
Predication in AI
AI Predication and Kinds of Teleology
Why Teleofunctionalism and Not Kripke or Evans?
Teleofunctional Role and Commitment (or Assertion)
Theories of Assertion and Commitment for Humans and AI
Part III: Conclusion
Chapter 8: Four Concluding Thoughts
Dynamic Goals
A Story of Neural Networks Taking Over in Ways We Cannot Understand
Why This Story is Disturbing and Relevant
Taking Stock and General Lessons
The Extended Mind and AI Concept Possession
Background: The Extended Mind and Active Externalism
The Extended Mind and Conceptual Competency
From Experts Determining Meaning to Artificial Intelligences Determining Meaning
Some New Distinctions: Extended Mind Internalist versus Extended Mind Externalists
Kripke, Putnam, and Burge as Extended Mind Internalists
Concept Possession, Functionalism, and Ways of Life
Implications for the View Defended in This Book
An Objection Revisited
Reply to the Objection
What Makes it a Stop Sign Detector?
Adversarial Perturbations
Explainable AI and Metasemantics
Bibliography
Index