OLIVIER BERGERET is a technical leader at Amazon Web Services (AWS), working on database and analytics services. He has over 25 years of experience in data engineering and analytics. Since joining AWS in 2015, he's supported the launch of most of AWS AI services including Amazon SageMaker and AWS DeepRacer. He is a regular speaker and presenter at various data, AI and cloud events such as AWS re:Invent, AWS Summits and third-party conferences.

ASIF ABBASI is a Principal Solutions Architect at AWS and has spent the last 20 years working in various roles with focus around Data Analytics, AI/ML, DWH Strategic and Technical Implementations, J2EE Enterprise applications design/development and Project Management. Asif is an Amazon Certified SA, Hortonworks Certified Hadoop professional and Administrator, Certified Spark Developer, SAS Certified Predictive Modeler, along with being a Sun Certified Enterprise Architect and a Teradata Certified Master.

JOEL FARVAULT is a Principal Solutions Architect Analytics at Amazon Web Services. He has 25 years' experience working on enterprise architecture, data strategy, and analytics, mainly in the financial services industry. Joel has led data transformation projects on fraud analytics, business intelligence, and data governance. He is also a lecturer on Data Analytics at IA School, at Neoma Business School and at Ecole Superieure de Genie Informatique (ESGI). Joel holds several associate and specialty certifications on AWS.

Acknowledgments xiii

About the Authors xv

Foreword xvii

Introduction xix

Chapter 1: A Brief History of AI 1

The Precursors of the Mechanical or "Formal" Reasoning 2

The Digital Computer Era 4

Cybernetics and the Beginning of the Robotic Era 6

Birth of AI and Symbolic AI (1955-1985) 10

Subsymbolic AI Era (1985-2010) 14

Deep Learning and LLM (2010-Present) 16

Key Takeaways 17

Chapter 2: Machine Learning 19

What Is Machine Learning? 19

Types of Machine Learning 20

Supervised Learning 21

Unsupervised and Semi-Supervised Learning 22

Reinforcement Learning 23

Methodology for Machine Learning 24

Implementation of Machine Learning 26

Machine Learning Applications 27

Natural Language Processing (NLP) 27

Computer Vision 27

Recommender System 27

Predictive Analytics 28

Fraud Detection 28

Machine Learning Frameworks and Libraries 28

TensorFlow 28

PyTorch 31

Scikit-learn 34

Keras 35

Apache Spark MLlib 37

Future Trends in Machine Learning 40

Rise of Edge Computing and Edge AI 40

Convergence with Emerging Technologies 40

Advancements in Unsupervised Learning, Reinforcement Learning, and Generative Models 41

Increased Specialization and Customization 41

Explainable and Trustworthy AI 42

Key Takeaways 42

References 43

Chapter 3: Deep Learning 45

Deep Learning vs. Machine Learning 45

Computer Vision Example 46

Natural Language Processing Example 47

The History of Deep Learning 47

Understanding Deep Learning 52

Neurons 52

Weights and Biases 54

Layers 54

Activation Function(s) 55

An Introduction to the Perceptron 58

Overcoming Perceptron Limitations 59

FeedForward Neural Networks 60

Backpropagation 60

Parameters vs. Hyperparameters 62

Hyperparameters in Artificial Neural Networks 64

Loss Functions - a Measure of Success of a Neural Network 64

Optimization Algorithms 64

Neural Network Architectures 68

Putting It All Together 71

Deep Learning on AWS 71

Chipsets and EC2 Instances 71

AWS P5 Instances 72

AWS Inferentia 72

Amazon Elastic Inference 73

Pre-built Containers: Deep Learning AMIs and Containers 74

Deep Learning AMIs 74

Deep Learning Containers 74

Managed Services for Building, Training, and Deployment 74

Pre-trained Services 75

Key Takeaways 77

References 77

Chapter 4: Introduction to Generative AI 79

Generative AI Core Technologies 80

Neural Networks 80

Generative Adversarial Networks (GANs) 80

Variational Autoencoders (VAEs) 81

Recurrent Neural Networks (RNNs) and Long Short-Term Memory Networks (LSTMs) 82

Limitations of Recurrent Neural Networks 84

Transformer Models 85

Self-Attention 86

Parallelism 86

Diffusion Models 86

Autoregressive Models 87

Reinforcement Learning (RL) 87

Transfer Learning and Fine-Tuning 87

Optimization Algorithms 87

Transformer Architecture: Deep Dive 87

Deep Dive 89

Step 1: Tokenization (Preprocessing) 89

Step 2: Embedding 89

Step 3: Encoder 92

Step 4: Encoder Output to Decoder Input 97

Step 5: Decoder 98

Step 6: Translation Generation 99

Step 7: Detokenization 99

Terminology in Generative AI 99

Prompt 104

Inference 105

Context Window 106

Prompt Engineering 106

In-Context Learning (ICL) 107

Zero-Shot/One-Shot/Few-Shot Inference 108

Inference Configuration 109

Maximum Length 110

Diversity (Top P/Nucleus Sampling) 111

Top K 111

Randomness (Temperature) 112

System Prompts 112

Prompt Engineering 113

Key Elements of a Prompt 113

Designing Effective Prompts 114

Prompting Techniques 115

Zero-Shot Prompting 115

Few-Shot Prompting 115

Chain-of-Thought Prompting 116

Advanced Prompting Techniques 117

Self-Consistency 118

Tree of Thoughts (ToT) 119

Retrieval-Augmented Generation (RAG) 120

Automatic Reasoning and Tool-Use (ART) 122

ReAct Prompting 123

Coherence Enhancement 124

Progressive Prompting 126

Handling Prompt Misuse 127

Prompt Injection 127

Prompt Leaking 128

Mitigating Bias 129

Mitigating Bias in Prompt Engineering 130

Generative AI Business Value 133

Building Value Within Your Enterprises 135

Technology: Creating a Flexible and Strong System 136

People: Training and Adapting the Team 136

Processes: Good Management and Fair Use of AI 136

Why a Solid Foundation Is Crucial 136

Summary 137

References 137

Chapter 5: Introduction to Foundation Models 139

Definition and Overview of Foundation Models 139

Characteristics of Foundation Models 142

Examples of Foundation Models 144

Types of Foundation Models 147

The Large Language Model (LLM) 154

Natural Language Processing 155

Early Approaches to NLP 156

Evolution Toward Text-Based Foundation Model 160

Applications of Foundation Models 162

Challenges and Considerations 163

Infrastructure 163

Ethics 164

Areas of Evolution 165

Key Takeaways 167

References 168

Chapter 6: Introduction to Amazon SageMaker 169

Data Preparation and Processing 172

Data Preparation 172

Data Processing 173

Model Development 174

Model Training and Tuning 175

Model Deployment 177

Model Management 178

Security 179

Compliance and Governance 180

Model Explainability and Responsible AI 181

MLOps with Amazon SageMaker 181

Boost Your Generative AI Development with SageMaker JumpStart 182

No-Code ML with Amazon SageMaker Canvas 183

Amazon Bedrock 184

Choosing the Right Strategy for the Development of Your Generative AI Application with Amazon SageMaker 186

Conclusion 187

References 188

Chapter 7: Generative AI on AWS 191

AWS Services for Generative AI 192

Generative AI Trade-Off Triangle 192

How AWS Solves the Generative AI Trade-Off Triangle 192

Generative AI on AWS: The Fundamentals 193

Infrastructure for FM Training and Inference 194

Models and Tools to Build Generative AI Apps 194

Applications to Boost Productivity 195

Amazon Bedrock 196

Foundation Models with Bedrock 197

AI21 Labs - Jurassic 197

Amazon Titan 198

Anthropic's Claude 3 199

Cohere's Family of Models 201

Key Features of Cohere 201

Cohere Models on Amazon Bedrock 203

Meta's Family of Models - Llama 204

When to Use Which Model 207

Mistral's Family of Models 208

When to Use Which Model 209

[...]'s Family of Models - Stable Diffusion XL 1.0 209

Poolside Family of Models 210

Luma's Family of Models 211

Amazon's Nova Family of Models 212

Model Evaluation in Amazon Bedrock 213

Common Approaches to Customizing Your FMs 214

Amazon Bedrock Prompt Management 214

Amazon Bedrock Flows 216

Data Automation in Amazon Bedrock 219

GraphRAG in Amazon Bedrock 220

Knowledge Bases in Amazon Bedrock 222

How Knowledge Bases Work 223

Pre-Processing Data 224

Runtime Execution 224

Creating a Knowledge Base in Amazon Bedrock 225

Agents for Amazon Bedrock 225

How Agents Work 226

Components of an Agent at Build Time 226

Components of an Agent at Runtime 228

Guardrails for Amazon Bedrock 230

Security in Amazon Bedrock 231

Amazon Q 232

Amazon Q Business 232

Amazon Q in QuickSight 235

Amazon Q Developer 238

Amazon Q Connect 239

Amazon Q in AWS Supply Chain 241

Summary 241

Chapter 8: Customization of Your Foundation Model 243

Introduction to LLM Customization 244

Continued Pre-Training (Domain Adaptation Fine-Tuning) 244

Fine-Tuning 245

Prompt Engineering 245

Retrieval Augmented Generation (RAG) 246

Choosing Between These Customization Techniques 246

Cost of Customization 249

Customizing Foundation Models with AWS 250

Continuous Pre-Training with Amazon Bedrock 250

Creation of a Training and a Validation Dataset 250

Launch of a Continued Pre-Training Job 251

Analysis of Our Results and Adjustment of Our Hyperparameters 252

Deployment of Our Model 254

Use Your Customized Model 255

Instruction Fine-Tuning with Amazon Bedrock 257

Instruction Fine-Tuning with Amazon SageMaker JumpStart 257

Conclusion 260

Chapter 9: Retrieval-Augmented Generation 263

What Is RAG? 263

Background and Motivation 264

Overview of RAG 266

Building a RAG Solution 269

Design Considerations 269

Best Practices 270

Common Patterns 271

Performance Optimization 271

Scaling Considerations 272

The Future of RAG Implementations 273

Retrieval Module 274

Retrieval Techniques and Algorithms 276

Augmentation Module 278

Generation Module 280

RAG on AWS 282

Custom Data Pipeline to Build RAG 284

Core Components of a RAG Pipeline 284

Implementation Approaches 286

Basic Solution: LangChain Implementation 286

Advanced Solution: Spark-Based Pipeline 287

Data Ingestion...