Beyond Algorithms: Delivering AI for Business

by James Luke, David Porter, Padmanabhan Santhanam

Overview

With so much artificial intelligence (AI) in the headlines, it is no surprise that businesses are scrambling to exploit this exciting and transformative technology. Clearly, those who are the first to deliver business-relevant AI will gain significant advantage.

However, there is a problem! Our perception of AI success in society is primarily based on our experiences with consumer applications from the big web companies. The adoption of AI in the enterprise has been slow due to various challenges. Business applications address far more complex problems and the data needed to address them is less plentiful. There is also the critical need for alignment of AI with relevant business processes. In addition, the use of AI requires new engineering practices for application maintenance and trust.

So, how do you deliver working AI applications in the enterprise?

Beyond Algorithms: Delivering AI for Business answers this question. Written by three engineers with decades of experience in AI (and all the scars that come with that), this book explains what it takes to define, manage, engineer, and deliver end-to-end AI applications that work. This book presents:

• Core conceptual differences between AI and traditional business applications
• A new methodology that helps to prioritise AI projects and manage risks
• Practical case studies and examples with a focus on business impact and solution delivery
• Technical Deep Dives and Thought Experiments designed to challenge your brain and destroy your weekends

Product Details

ISBN-13:	9780367613266
Publisher:	CRC Press
Publication date:	05/30/2022
Pages:	302
Product dimensions:	7.00(w) x 10.00(h) x (d)

About the Author

James Luke, is an Engineer with over 25 years’ experience delivering real AI solutions that solve real world problems. James is the Innovation Director at Roke, a leading UK technology company, having previously worked as an IBM Distinguished Engineer and Master Inventor. James has multiple US patents in subjects relating to AI and, for his PhD, researched the application of AI in detecting previously unseen computer viruses. James is an experienced conference speaker and has given evidence on the development of AI to both the European Commission and the House of Lords Select Committee. In 2018, James delivered a TEDx talk entitled, “How To Survive An AI Winter” ( https://www.youtube.com/watch?v=MWOkEVdITIg ). James started his career failing to deliver an AI solution for a leading Formula 1 team. This experience changed James’s understanding and perspective on what it takes to actually deliver a working AI solution. James responded to his early failure by developing new methods for the definition, design and delivery of AI solutions. He has delivered projects in multiple industries from Public Sector to Insurance and Retail. Prior to joining Roke, James held a number of key positions in IBM including Chief Architect for Watson Tools, CTO of the Cognitive Practice in Europe and Leader of the Academy of Technology core team on AI.

Dr. Padmanabhan Santhanam is currently a Principal Research Staff Member at the IBM T. J. Watson Research Center in New York, working to enable AI systems in government and public sector. His personal research interest is both in the use of AI for engineering traditional software systems and the emerging field of AI Engineering (i.e. how to engineer trust-worthy AI systems). Prior to that, Dr. Santhanam worked on several aspects of AI strategy and execution in IBM Research. He holds a Ph.D. in Applied Physics from Yale University. Dr. Santhanam worked in software engineering research for two decades, having to do with the creation of tools and methodology to improve commercial software development. His interests included software quality metrics, automation of software test generation, realistic modeling of software development processes, etc. He has more than fifty published research papers in peer-reviewed journals and conferences in a variety of topics. He is a member of the ACM & AAAI and a Senior Member of the IEEE. He is also a Member of the IBM Academy of Technology.

David Porter is currently an Associate Partner at IBM Consulting. He graduated in 1995 from the University of Greenwich with a degree in Information Systems Engineering. He has worked in AI and Data Science ever since, with consultancy roles at SAS Software, Detica/BAE Systems and now IBM. Early on in his career he chose to focus on counter-fraud and law enforcement systems. This specialisation has allowed him to work with governments and organisations all over the world. Achievements in this field include the co-invention of the graph analytics software NetReveal and leading the design teams for both the UK's Insurance Fraud Bureau and the original Connect system at Her Majesty's Revenue and Customs (HMRC). He joined IBM in 2016, enticed by the Watson story; could AI be used to catch crooks? He has been putting Natural Language Processing to good use ever since.

Table of Contents

Authors xi

Acknowledgements xiii

Prologue 1

Chapter 1 Why This Book? 3

AI is Everywhere 3

Enterprise Applications 4

AI Winters 5

What is Different Now? 7

Proceed with Caution! 8

Delivering AI Solutions 8

Better Understanding of AI is Critical for Society 8

Target Audience for the Book 9

An Outline of the Book 9

References 10

Chapter 2 Building Applications 11

What's Different about AI When Building an Application? 11

Prominent AI Applications of the Last Seven Decades 15

AI or No AI? 23

The Present - The Dominance of the Web 24

The Future - The Enterprise Strikes Back 27

Examples of Real Enterprise Applications 31

Where Do You Introduce Al? 38

Activities in Creating an AI Application 39

Complexity of AI Applications 42

Architectural and Engineering Considerations 44

Three Stages of an Enterprise AI Application 46

Enabling Enterprise Solutions at Scale 48

In Summary - Are You Ready to Start Building Applications? 51

References 52

Chapter 3 It's Not Just the Algorithms, Really! 55

Introducing Algorithms 56

Algorithms in AI 57

Algorithm Addiction 59

Applications Versus the Underlying Technology 60

Algorithms and Models 60

Object Dropping Problem 63

Understanding the Object Dropping Data 65

Four Models to Predict Object Breakage 67

Comparing the Two ML Approaches 75

Comparing Physics Model with ML 78

What are the ML Algorithms Actually Learning? 79

Feature Definition and Extraction 81

Revenge of the Artificial Neural Networks 83

Human Interpretation of Artificial Neural Networks 84

So Which Algorithm is Best? 85

Transfer Learning 86

Reinforcement Learning 88

Brain Versus Artificial Neural Networks 88

Fundamental Principles and Fundamental Mistakes 90

So … It Really isn't all About the Algorithm 91

In Summary - There Really is So Much More to AI than the Algorithms 92

References 93

Chapter 4 Know Where to Start - Select the Right Project 95

The Doability Method 96

Innovation and Emerging Technologies 96

A Portfolio-Based Approach 97

Doability Method Step 1 - To AI or Not Al 97

Three Recommendations from Doability Method Step 1 101

Doability Method Step 1 - Worked Examples 102

Doability Method Step 2 - Prioritising AI Projects in the Portfolio 106

In Summary - Success or Failure will Depend on Selecting the Right Project 107

References 108

Chapter 5 Business Value and Impact 109

What is Different about AI Applications? 110

Building Business Cases 110

Stakeholders 112

Measurability and Understandability 117

Importance of Ethics in AI Development 119

Delivering Trustworthy AI 122

Fairness and Bias 123

Explainability 128

Transparency 132

Tackling the Weakness of ML Systems 133

In Summary - There's More to Value than Monetary Return 134

References 135

Chapter 6 Ensuring It Works - How Do You Know? 137

Managing Quality of Traditional Software 137

Managing Quality of AI Applications 139

Statistical Accuracy 140

Cost Functions 146

Multiple Outcomes 147

Quality Metrics for Natural Language Understanding 147

What Does this Mean in Practice? 150

How Accurate Does it Need to Be? 154

Where Do You Assess Accuracy and Business Impact? 156

Operating within Limits 156

Quality Attributes of Trustworthy AI Systems 159

In Summary - If the AI Isn't Trustworthy, People Won't Trust It 160

References 161

Chapter 7 It's All about the Data 163

Data Tsunami 164

Data Types 166

Data Sources for AI 167

Data for the Enterprise 168

Enterprise Reality 168

Humans Versus AI - Learning and Decision-Making 169

Data Wrangling 170

How Much Data Do We Need? 171

So, What Features Do We Need? 179

Enabling Expanding Feature Spaces 183

What Happens in the Real World? 183

Coping with Missing Data 187

Use of Synthetic Data 189

Managing the Data Workflow 193

Improving Data Quality 198

In Summary - It Really is all About the Data! 199

References 199

Chapter 8 How Hard Can It Be? 201

Demonstrations Versus Business Applications 201

Setting Expectations … Yours and Others! 203

Do We Need an Invention? 203

Current State of AI 204

The Importance of Domain Specialists 208

Business Change and AI 208

AI is Software 210

The Great Reuse Challenge 210

The AI Factory 214

In Summary - It can be as Hard as You Make It 216

References 217

Chapter 9 Getting Your Priorities Right 219

AI Project Assessment Checklist 220

Using the Doability Matrix 224

In Summary - Never Take Off Without Completing Your Checklist 230

Reference 230

Chapter 10 Some (Not So) Boring Stuff 231

Traditional Engineering 231

Why is Engineering AI Different? 234

Four Phases of an AI Project 238

Developing an Enterprise AI Application 239

AI Model Lifecycle 240

Application Lifecycle 242

Application Integration and Deployment 244

Project Management 251

Auditability and Explainability 252

Security 253

In Summary - The Boring Stuff isn't Really Boring 254

References 254

Chapter 11 The Future 257

It's all about the Data - Trends in the Enterprise 259

Efficient Computing for AI Workloads - New Paradigms 260

Advances in Algorithms - Targeting Data Challenges and Neuro-Symbolic AI 263

AI Engineering - Emergence of a New Discipline 269

Human-Machine Teaming 271

In Summary - Some Final Thoughts 274

References 275

Epilogue 279

Index 281

Editorial Reviews

"Much has been written about the mechanics of modern AI - what it is, what it isn't, and how to build and train one - but this fascinating book is among the few that explain what it means for the enterprise - where it fit in to business needs, how it can be integrated into existing systems, and how it can create opportunities for new value." — Grady Booch, IBM Fellow, Chief Scientist for Software Engineering

"I recommend Beyond Algorithms - Delivering AI for Business to anyone that is responsible for leading and transitioning AI to achieve and maintain a competitive advantage. The book addresses many of the important issues faced in the application of AI to businesses. It is written in an easy to read style, while providing enough technical content that many people working in the field can capture the main messages." — David R. Martinez, MIT Instructor and Laboratory Fellow

"Books about AI too often tend towards the extremes — either praising or condemning unconditionally. Beyond Algorithms offers a more nuanced, nontechnical introduction from experienced practitioners. Filled with guiding principles and real-world examples, this book offers a balanced view of AI's strengths and limitations and provides practical advice for problem selection, project management, and expectation setting for AI initiatives." — Emily Riederer, Senior Analytics Manager at Capital One

"Beyond Algorithmsis one of those wonderful story-driven IT books that distils decades worth of real world experience into intoxicating wisdom that is refreshingly easy to consume. The book plows through and around the hype of the 'all powerful' deep neural nets to provide an engineering approach for the field. It won’t tell you how to establish the Singularity, but it’s ‘A to I checklist’, accuracy and monitoring advice, and ‘doability method’ could make you an AI delivery hero." — Richard Hopkins, FREng FIET CEng, former President of IBM’s Academy of Technology

“Important and timely. A practical guide for realizing business value through AI capabilities! Rather than focusing on hard to access details of the mathematics and algorithms behind modern AI, this book provides a roadmap for a diversity of stakeholders to realize AI business solutions that are reliable, responsible, and sustainable.” — Matthew Gaston, AI Engineering Evangelist

From the Publisher