Succeeding with AI

Succeeding with AI
brief contents
contents
preface
acknowledgments
about this book
	Who should read this book
	How this book is organized
	liveBook discussion forum
about the author
about the cover illustration
1 Introduction
	1.1 Whom is this book for?
	1.2 AI and the Age of Implementation
	1.3 How do you make money with AI?
	1.4 What matters for your project to succeed?
	1.5 Machine learning from 10,000 feet
	1.6 Start by understanding the possible business actions
	1.7 Don’t fish for “something in the data”
	1.8 AI finds correlations, not causes!
	1.9 Business results must be measurable!
	1.10 What is CLUE?
	1.11 Overview of how to select and run AI projects
	1.12 Exercises
		1.12.1 True/False questions
		1.12.2 Longer exercises: Identify the problem
	Summary
2 How to use AI in your business
	2.1 What do you need to know about AI?
	2.2 How is AI used?
	2.3 What’s new with AI?
	2.4 Making money with AI
		2.4.1 AI applied to medical diagnosis
		2.4.2 General principles for monetizing AI
	2.5 Finding domain actions
		2.5.1 AI as part of the decision support system
		2.5.2 AI as a part of a larger product
		2.5.3 Using AI to automate part of the business process
		2.5.4 AI as the product
	2.6 Overview of AI capabilities
	2.7 Introducing unicorns
		2.7.1 Data science unicorns
		2.7.2 What about data engineers?
		2.7.3 So where are the unicorns?
	2.8 Exercises
		2.8.1 Short answer questions
		2.8.2 Scenario-based questions
	Summary
3 Choosing your first AI project
	3.1 Choosing the right projects for a young AI team
		3.1.1 The look of success
		3.1.2 The look of failure
	3.2 Prioritizing AI projects
		3.2.1 React: Finding business questions for AI to answer
		3.2.2 Sense/Analyze: AI methods and data
		3.2.3 Measuring AI project success with business metrics
		3.2.4 Estimating AI project difficulty
	3.3 Your first project and first research question
		3.3.1 Define the research question
		3.3.2 If you fail, fail fast
	3.4 Pitfalls to avoid
		3.4.1 Failing to build a relationship with the business team
		3.4.2 Using transplants
		3.4.3 Trying moonshots without the rockets
		3.4.4 It’s about using advanced tools to look at the sea of data
		3.4.5 Using your gut feeling instead of CLUE
	3.5 Exercises
	Summary
4 Linking business and technology
	4.1 A project can’t be stopped midair
		4.1.1 What constitutes a good recommendation engine?
		4.1.2 What is gut feeling?
	4.2 Linking business problems and research questions
		4.2.1 Introducing the L part of CLUE
		4.2.2 Do you have the right research question?
		4.2.3 What questions should a metric be able to answer?
		4.2.4 Can you make business decisions based on a technical metric?
		4.2.5 A metric you don’t understand is a poor business metric
		4.2.6 You need the right business metric
	4.3 Measuring progress on AI projects
	4.4 Linking technical progress with a business metric
		4.4.1 Why do we need technical metrics?
		4.4.2 What is the profit curve?
		4.4.3 Constructing a profit curve for bike rentals
		4.4.4 Why is this not taught in college?
		4.4.5 Can’t businesses define the profit curve themselves?
		4.4.6 Understanding technical results in business terms
	4.5 Organizational considerations
		4.5.1 Profit curve precision depends on the business problem
		4.5.2 A profit curve improves over time
		4.5.3 It’s about learning, not about being right
		4.5.4 Dealing with information hoarding
		4.5.5 But we can’t measure that!
	4.6 Exercises
	Summary
5 What is an ML pipeline, and how does it affect an AI project?
	5.1 How is an AI project different?
		5.1.1 The ML pipeline in AI projects
		5.1.2 Challenges the AI system shares with a traditional software system
		5.1.3 Challenges amplified in AI projects
		5.1.4 Ossification of the ML pipeline
		5.1.5 Example of ossification of an ML pipeline
		5.1.6 How to address ossification of the ML pipeline
	5.2 Why we need to analyze the ML pipeline
		5.2.1 Algorithm improvement: MNIST example
		5.2.2 Further examples of improving the ML pipeline
		5.2.3 You must analyze the ML pipeline!
	5.3 What’s the role of AI methods?
	5.4 Balancing data, AI methods, and infrastructure
	5.5 Exercises
	Summary
6 Analyzing an ML pipeline
	6.1 Why you should care about analyzing your ML pipeline
	6.2 Economizing resources: The E part of CLUE
	6.3 MinMax analysis: Do you have the right ML pipeline?
	6.4 How to interpret MinMax analysis results
		6.4.1 Scenario: The ML pipeline for a smart parking meter
		6.4.2 What if your ML pipeline needs improvement?
		6.4.3 Rules for interpreting the results of MinMax analysis
	6.5 How to perform an analysis of the ML pipeline
		6.5.1 Performing the Min part of MinMax analysis
		6.5.2 Performing the Max part of MinMax analysis
		6.5.3 Estimates and safety factors in MinMax analysis
		6.5.4 Categories of profit curves
		6.5.5 Dealing with complex profit curves
	6.6 FAQs about MinMax analysis
		6.6.1 Should MinMax be the first analysis of the ML pipeline?
		6.6.2 Which analysis should you perform first? Min or Max?
		6.6.3 Should a small company or small team skip the MinMax analysis?
		6.6.4 Why do you use the term MinMax analysis?
	6.7 Exercises
	Summary
7 Guiding an AI project to success
	7.1 Improving your ML pipeline with sensitivity analysis
		7.1.1 Performing local sensitivity analysis
		7.1.2 Global sensitivity analysis
		7.1.3 Example of using sensitivity analysis results
	7.2 We’ve completed CLUE
	7.3 Advanced methods for sensitivity analysis
		7.3.1 Is local sensitivity analysis appropriate for your ML pipeline?
		7.3.2 How to address the interactions between ML pipeline stages
		7.3.3 Should I use design of experiments?
		7.3.4 One common objection you might encounter
		7.3.5 How to analyze the stage that produces data
		7.3.6 What types of sensitivity analysis apply to my project?
	7.4 How your AI project evolves through time
		7.4.1 Time affects your business results
		7.4.2 Improving the ML pipeline over time
		7.4.3 Timing diagrams: How business value changes over time
	7.5 Concluding your AI project
	7.6 Exercises
	Summary
8 AI trends that may affect you
	8.1 What is AI?
	8.2 AI in physical systems
		8.2.1 First, do no harm
		8.2.2 IoT devices and AI systems must play well together
		8.2.3 The security of AI is an emerging topic
	8.3 AI doesn’t learn causality, only correlations
	8.4 Not all data is created equal
	8.5 How are AI errors different from human mistakes?
		8.5.1 The actuarial view
		8.5.2 Domesticating AI
	8.6 AutoML is approaching
	8.7 What you’ve learned isn’t limited to AI
	8.8 Guiding AI to business results
	8.9 Exercises
	Summary
appendix A Glossary of terms
appendix B Exercise solutions
	B.1 Answers to chapter 1 exercises
		B.1.1 True/False questions
		B.1.2 Longer exercises: Identify the problem
	B.2 Answers to chapter 2 exercises
		B.2.1 Short answer questions
		B.2.2 Answers to the scenario-based questions
	B.3 Answers to chapter 3 exercises
	B.4 Answers to chapter 4 exercises
	B.5 Answers to chapter 5 exercises
	B.6 Answers to chapter 6 exercises
	B.7 Answers to chapter 7 exercises
	B.8 Answers to chapter 8 exercises
appendix C Bibliography
index
	A
	B
	C
	D
	E
	F
	G
	H
	I
	K
	L
	M
	N
	O
	P
	R
	S
	T
	U
	V
	W
	Z
Succeeding with AI - back