Designing Autonomous AI

Spis treści książki

• Foreword
• Preface
  • What Is Autonomous AI?
  • Who Should Read This Book?
    • Process Experts
    • Data Scientists and Software Engineers
    • Innovation Leaders
    • Teachers
    • Problem Solvers
  • What Can You Expect to Learn from This Book?
  • Conventions Used in This Book
  • OReilly Online Learning
  • How to Contact Us
  • Acknowledgments
• Introduction: The Right Brain in the Right Place (Why We Need Autonomous AI)
  • The Changing World Requires Adapting Skills
    • Problems Need Solutions, Not AI
    • What Can AI Do for Me in Real Life?
    • AI Decision-Making Is Becoming More Autonomous
    • Beware of Data Science Colonialism
  • The Changing Workforce Demands Transferred Skills
    • Expertise Is Hard to Acquire
    • Expertise Is Hard to Maintain
    • Expertise Is Simple to Teach, but Requires Practice
  • Pressing Problems Demand Completely New Skills
    • AI Is a Tool; Use It for Good
• I. When Automation Doesnt Work
• 1. Sometimes Machines Make Bad Decisions
  • Math, Menus, and Manuals: How Machines Make Automated Decisions
    • Control Theory Uses Math to Calculate Decisions
    • Optimization Algorithms Use Menus of Options to Evaluate Decisions
      • Solutions Are like Points on a Map
      • Solving the Game of Checkers
      • Reconnaissance
      • What About Uncertainty?
    • Expert Systems Recall Stored Expertise
• 2. The Quest for More Human-Like Decision-Making
  • Augmenting Human Intelligence
  • How Humans Make Decisions and Acquire Skills
    • Humans Act on What They Perceive
    • Humans Build Complex Correlations into Their Intuition with Practice
    • Humans Abstract to Strategy for Complex Tasks
      • Beginner
      • Advanced beginner
      • Competent
      • Proficient
      • Expert
    • Theres a New Kind of AI in Town
      • Reinforcement learning acquires skill through practice
      • Neural networks can correlate any relationship between variables
  • The Superpowers of Autonomous AI
    • Autonomous AI Makes More Human-Like Decisions
    • Autonomous AI Perceives, Then Acts
    • The Difference Between Perception and Action in AI
    • Autonomous AI Learns and Adapts When Things Change
    • Autonomous AI Can Spot Patterns
    • Autonomous AI Infers from Experience
    • Autonomous AI Improvises and Strategizes
    • Autonomous AI Can Plan for the Long-Term Future
    • Autonomous AI Brings Together the Best of All Decision-Making Technologies
  • When Should You Use Autonomous AI?
  • Autonomous AI Is like a Brilliant, Curious Toddler That Needs to Be Taught
• II. What Is Machine Teaching?
• 3. How Brains Learn Best: Teaching Humans and AI
  • Learning Multiple Skills Simultaneously Is Hard for Humans and AI
  • Teaching Skills and Strategies Explicitly
  • Teaching Allows Us to Trust AI
  • The Mindset of a Machine Teacher
    • Teacher More Than Programmer
    • Learner More Than Expert
  • What Is a Brain Design?
    • How Decision-Making Works
      • Exploring without a map
      • Learning systems navigate better with skills as landmarks than rote instructions
      • So, what is a skill?
      • Why cant you just give me a map?!
      • Skills let your learner learn
    • Acquiring Skill Is like Learning to Navigate by Exploring
      • Zones of proximal development
      • Scaffolding
    • A Brain Design Is a Mental Map That Guides Exploration with Landmarks
• 4. Building Blocks for Machine Teaching
  • Case Study: Learning to Walk Is Hard to Evolve, Easier to Teach
    • So, Why Do We Walk?
    • Strategy Versus Evolution
    • Teaching Walking as Three Skills
      • Defining skills
      • Setting goals for each skill
      • Organizing the skills
  • Concepts Capture Knowledge
  • Skills Are Specialized Concepts
  • Brains Are Built from Skills
    • Building Skills
    • Expert Rules Inflate into Skills
      • Teach expert rules, and let the learner inflate the concepts through practice
    • Perceptive Concepts Discern or Recognize
      • See and hear
      • Predict
      • Detect
      • Classify
      • Filter
    • Directive Concepts Decide and Act
    • Selective Concepts Supervise and Assign
      • Programmed concepts
      • Learned concepts
  • Brains Are Organized by Functions and Strategies
    • Sequences or Parallel Execution for Functional Skills
      • Sequences live in the selector
      • Fixed order sequences
      • Parallel execution of functional skills
      • Variable order sequences
    • Hierarchies for Strategies
      • Discovering strategies
      • Strategies wax and wane in effectiveness over time
      • Strategies capture trade-offs
      • Selective concepts navigate strategy hierarchies
  • Visual Language of Brain Design
• III. How Do You Teach a Machine?
  • Understanding the Process
  • Meet with Experts
  • Ask the Right Questions
  • Case Study: Lets Design a Smart Thermostat
• 5. Teaching Your AI Brain What to Do
  • Determining Which Actions the Brain Will Take
    • Perception Is Required, but Its Not All We Need
    • Sequential Decisions
  • Triggering the Action in Your AI Brain
  • Setting the Decision Frequency
  • Handling Delayed Consequences for Brain Actions
  • Actions for Smart Thermostat
• 6. Setting Goals for Your AI Brain
  • Theres Always a Trade-off
    • Throughput Versus Efficiency
    • Supervisors Have Different Goals Than Crews Do
    • Dont Prioritize Goals; Balance Them Instead
    • Watch Out for Expert Rules Disguised as Goals
    • Ideal Versus Available
  • Setting Goals
    • Step 1: Identify Scenarios
    • Step 2: Match Goals to Scenarios
    • Step 3: Teach Strategies for Each Scenario
  • Goal Objectives
    • Maximize
    • Minimize
    • Reach, like the Finish Line for a Race
    • Drive, like the Temperature for a Thermostat
    • Avoid, like Dangerous Conditions
    • Standardize, like the Heat in an Oven
    • Smooth, like a Line
  • Expanding Task Algebra to Include Goal Objectives
  • Setting Goals for a Smart Thermostat
• 7. Teaching Skills to Your AI Brain
  • Teaching Focuses and Guides Practice (Exploration)
  • Skills Can Evolve and Transform
  • Skills Adapt to the Scenario
  • Levels of Teaching Sophistication
    • The Introductory Teacher Conveys the Facts and Goals
    • The Coach Sequences Skills to Practice
    • The Mentor Teaches Strategy
    • The Maestro Democratizes New Paradigms
  • How Maestros Democratize Technology
  • Levels of Autonomous AI Architecture
    • Machine Learning Adds Perception
    • Monolithic Brains Are Advanced Beginners
    • Concept Networks Are Competent Learners
    • Massive Concept Networks Are Proficient Learners
  • Pursuing Expert Skill Acquisition in Autonomous AI
    • Brains That Come with Hardwired Skills
    • Brains That Define Skills as They Learn
    • Brains That Assemble Themselves
    • Brains with Skills That Coordinate
  • Steps to Architect an AI Brain
    • Step 1: Identify the Skills That You Want to Teach
    • Step 2: Orchestrate How the Skills Work Together
    • Step 3: Select Which Technology Should Perform Each Skill
  • Pitfalls to Avoid When Teaching Skills
    • Pitfall 1: Confusing the solution for the problem
    • Pitfall 2: Losing the forest for the trees
  • Example of Teaching Skills to an AI Brain: Rubber Factory
  • Brain Design for Our Smart Thermostat
• 8. Giving Your AI Brain the Information It Needs to Learn and Decide
  • Sensors: The Five Senses for Your AI Brain
    • Variables
    • Proxy Variables
    • Trends
  • Simulators: A Gym for Your Autonomous AI to Practice In
    • Simulating Reality Using Physics and Chemistry
    • Simulating Reality Using Statistics and Events
    • Simulating Reality Using Machine Learning
    • Simulating Reality Using Expert Rules
  • Sensor Variables for Smart Thermostat
• IV. Tools for the Machine Teacher
• 9. Designing AI Brains That Someone Can Actually Build
  • Designers and Builders Working Together in Harmony (Mostly)
    • The Autonomous AI Design Fallacy Designs but Wont Iterate
    • The Autonomous AI Implementation Fallacy Skips Design Altogether
  • Specification for Documenting AI Brain Designs
  • Platform for Machine Teaching
  • Platform for Wiring Multiple Skills Together as Modules
    • Platform for mixing math, menus, and manuals with AI
  • What Difference Will You Make with Machine Teaching?
• Glossary
• Index

pokaż cały spis treści

ukryj recenzje