[Part 1] The Basics of Generative AI

▣ Chapter 1: Introduction to Generative AI
1.1 What does this book cover?
1.2 What is generative AI?
1.3 What can be created?
__1.3.1 Entity Extraction
__1.3.2 Text Generation
__1.3.3 Image Creation
__1.3.4 Code Generation
__1.3.5 Logical problem solving skills
__1.3.6 Music Creation
__1.3.7 Video Creation
1.4 Enterprise Use Cases
1.5 If you do not use generative AI
1.6 How does generative AI differ from traditional AI?
1.7 What approach should companies take?
1.8 Architectural Considerations
1.9 Process for Introducing Generative AI in Businesses

▣ Chapter 2: Introduction to Large-Scale Language Models
2.1 Base Model Overview
2.2 LLM Overview
2.3 Transformer Architecture
2.4 Training Cutoff
2.5 Types of LLM
2.6 Small-scale language models
2.7 Open Source vs.
Commercial LLM
__2.7.1 Commercial LLM
__2.7.2 Open Source LLM
2.8 Key Concepts of LLM
__2.8.1 prompt
__2.8.2 Token
__2.8.3 Token Calculation
__2.8.4 Embedding
__2.8.5 Model Configuration
__2.8.6 Context Window
__2.8.7 Prompt Engineering
__2.8.8 Model Adaptation
__2.8.9 Emergent Behavior

▣ Chapter 3: Working with APIs - Text Generation
3.1 Model Categories
__3.1.1 Dependencies
__3.1.2 Model query
3.2 Completed API
__3.2.1 Completed Extension
__3.2.2 Azure Content Safety Filter
__3.2.3 Multiple Completion
__3.2.4 Randomness Control
__3.2.5 Randomness Control Using top_p
3.3 Advanced Completion API Options
__3.3.1 Streaming Complete
__3.3.2 Factors affecting token probability: logit_bias
__3.3.3 Presence and Frequency Penalties
__3.3.4 Log probability
3.4 Conversational Completion API
__3.4.1 System Roles
__3.4.2 Reason for completion
__3.4.3 Conversational Completion API for Non-Chat Scenarios
__3.4.4 Conversation Management
__3.4.5 Best Practices for Token Management
__3.4.6 Additional LLM Providers

▣ Chapter 4: From Pixels to Photos - Image Creation
4.1 Vision Model
__4.1.1 Variational Autoencoder
__4.1.2 Generative Adversarial Networks
__4.1.3 Vision Transformer Model
__4.1.4 Diffusion Model
__4.1.5 Multimodal Model
4.2 Image generation through stable diffusion
__4.2.1 Dependencies
__4.2.2 Creating an Image
4.3 Creating images through other providers
__4.3.1 OpenAI DALL·E 3
__4.3.2 Bing Image Creator
__4.3.3 Adobe Firefly
4.4 Editing and Enhancing Images Using Stable Diffusion
__4.4.1 Creating an Image Using the Image-to-Image API
__4.4.2 Using the Masking API
__4.4.3 Resizing images using the upscale API
__4.4.4 Image Creation Tips

▣ Chapter 5: What else can artificial intelligence create?
5.1 Code Generation
__5.1.1 Can I trust the code?
__5.1.2 GitHub Copilot
__5.1.3 How Copilot Works
5.2 Additional Code Related Work
__5.2.1 Code Description
__5.2.2 Test Generation
__5.2.3 Code Reference
__5.2.4 Code Refactoring
5.3 Other code generation tools
__5.3.1 Amazon CodeWhisperer
__5.3.2 Code Llama
__5.3.3 Tab Nine
__5.3.4 Self-check
__5.3.5 Best Practices for Code Generation
5.4 Video Creation
5.5 Audio and Music Creation

[Part 2] Advanced Techniques and Applications

▣ Chapter 6: Prompt Engineering Guide
6.1 What is Prompt Engineering?
__6.1.1 Why Prompt Engineering is Needed
6.2 Fundamentals of Prompt Engineering
6.3 Learning and Prompting in Context
6.4 Prompt Engineering Techniques
__6.4.1 System Messages
__6.4.2 Zero-shot learning, few-shot learning, and multi-shot learning
__6.4.3 Use clear syntax
__6.4.4 How to make learning in context work well
__6.4.5 Inference: Chain of Thought
__6.4.6 Self-consistent sampling
6.5 Image Prompting
6.6 Prompt Injection
6.7 Prompt Engineering Challenges
6.8 Best Practices

▣ Chapter 7: Augmented Search Creation: The Secret Weapon
7.1 What is RAG?
7.2 Advantages of RAG
7.3 RAG Architecture
7.4 Search Engine System
7.5 Understanding Vector Databases
__7.5.1 What is a vector index?
__7.5.2 Vector Search
7.6 RAG Task
7.7 Solutions to Chunking-Related Challenges
__7.7.1 Chunking Strategy
__7.7.2 Factors Influencing Chunking Strategy
__7.7.3 Handling unknown complexity
__7.7.4 Sentence-level chunking
__7.7.5 Chunking using natural language processing
7.8 PDF Chunking

▣ Chapter 8: Chatting with Data
8.1 Benefits of Companies Leveraging Their Own Data
__8.1.1 Pros and Cons of Large Context Windows
__8.1.2 Building a Chat Application Using Data
8.2 Using a Vector Database
8.3 Planning for Information Retrieval
8.4 Data Retrieval
__8.4.1 Finder Pipeline Best Practices
8.5 Searching with Redis
8.6 Comprehensive chat implementation based on RAG
8.7 Using Azure OpenAI on your data
8.8 Benefits of Integrating Enterprise Data into RAG

▣ Chapter 9: Model Customization through Model Adaptation and Fine-Tuning
9.1 What is model adaptation?
__9.1.1 Basics of Model Adaptation
__9.1.2 Advantages and Challenges in Business
9.2 When to Fine-Tune LLM
__9.2.1 Key steps in LLM fine-tuning
9.3 Fine-tuning the OpenAI model
__9.3.1 Preparing the Data Set for Fine-Tuning
__9.3.2 LLM Evaluation
__9.3.3 Fine Tuning
__9.3.4 Fine-tuning training metrics
__9.3.5 Fine-tuning with Azure OpenAI
9.4 Deploying the Fine-Tuned Model
__9.4.1 Inference: Fine-tuned Model
9.5 LLM Training
__9.5.1 Pre-training
__9.5.2 Map Fine Tuning
__9.5.3 Compensation Modeling
__9.5.4 Reinforcement Learning
__9.5.5 Direct Policy Optimization
9.6 Model Adaptation Techniques
__9.6.1 Low-Rank Adaptation
9.7 RLHF Overview
__9.7.1 RLHF's Challenges
__9.7.2 Extending the RLHF Implementation

[Part 3] Distribution and Ethical Considerations

▣ Chapter 10: Application Architecture for Generative AI Apps
10.1 Generative AI: Application Architecture
__10.1.1 Software 2.0
__10.1.2 The Age of Co-Pilots
10.2 Generative AI: Application Stack
__10.2.1 Integrating the Generative AI Stack
__10.2.2 Generative AI Architecture Principles
__10.2.3 Generative AI Application Architecture: A Detailed View
10.3 Orchestration Layer
__10.3.1 Benefits of the Orchestration Framework
__10.3.2 Orchestration Framework
__10.3.3 Operations Management
__10.3.4 Prompt Management
10.4 Grounding Layer
__10.4.1 Data Integration and Preprocessing
__10.4.2 Embedding and Vector Management
10.5 Model Hierarchy
__10.5.1 Model Ensemble Architecture
__10.5.2 Model Serving
10.6 Response Filtering

▣ Chapter 11: Scaling Up: Best Practices for Production Deployments
11.1 Challenges of Production Deployment
11.2 Deployment Options
11.3 Managed LLM via API
11.4 Best Practices for Production Deployments
__11.4.1 Metrics for LLM Inference
__11.4.2 Delay Time
__11.4.3 Scalability
__11.4.4 PAYGO
__11.4.5 Quotas and Rate Limits
__11.4.6 Quota Management
__11.4.7 Observability
__11.4.8 Security and Compliance Considerations
11.5 Generative AI Operational Considerations
__11.5.1 Reliability and Performance Considerations
__11.5.2 Managed ID
__11.5.3 Caching
11.6 LLMOps and MLOps
11.7 Checklist for Production Deployment

▣ Chapter 12: Evaluation and Benchmarks
12.1 LLM Evaluation
12.2 Existing evaluation indicators
__12.2.1 BLEU
__12.2.2 ROUGE
__12.2.3 BERTScore
__12.2.4 Example of existing indicator evaluation
12.3 LLM Task-Specific Benchmarks
__12.3.1 G-Eval: A Measurement Approach for NLG Evaluation
__12.3.2 Example of LLM-based evaluation metrics
__12.3.3 HELM
__12.3.4 HEIM
__12.3.5 HellaSWAG
__12.3.6 Understanding Large-Scale Multitask Languages
__12.3.7 Using Azure AI Studio for Evaluation
__12.3.8 DeepEval: An LLM Evaluation Framework
12.4 New Evaluation Benchmarks
__12.4.1 SWE-bench
__12.4.2 MMMU
__12.4.3 MoCa
__12.4.4 HaluEval
12.5 Human Rating

▣ Chapter 13: A Guide to Ethical Generative AI: Principles, Cases, and Pitfalls
13.1 Generative AI Risks
__13.1.1 LLM Restrictions
__13.1.2 Hallucinations
13.2 Understanding Generative AI Attacks
__13.2.1 Prompt Injection
__13.2.2 Example of unsafe output processing
__13.2.3 Model Denial of Service Attack
__13.2.4 Data Poisoning and Backdoors
__13.2.5 Sensitive information leak
__13.2.6 Over-reliance
__13.2.7 Model Hijacking
13.3 Responsible AI Lifecycle
__13.3.1 Identifying Risk Factors
__13.3.2 Measurement and Evaluation of Risk Factors
__13.3.3 Mitigation of hazard factors
__13.3.4 Transparency and Explainability
13.4 Red Team
__13.4.1 Red Team Example
__13.4.2 Red Team Tools and Techniques
13.5 Content Safety
__13.5.1 Azure Content Safety
__13.5.2 Google Perspective API
__13.5.3 Content Filter Evaluation

▣ Appendix
A: This book's GitHub repository
B: Responsible AI Tools
__B.1 Model Card
__B.2 Transparency Documentation
__B.3 HAX Toolkit
__B.4 Responsible AI Toolbox
__B.5 Learning Interpretation Tool
__B.6 AI Fairness 360
__B.7 C2PA
C: Azure OpenAI User Guide
__C.1 Create a Microsoft account and prepare an Azure subscription
__C.2 Creating Azure OpenAI Resources
__C.3 Model Deployment
__C.4 Verifying Keys and Endpoints
D: Required Dependency Components Installation Manual
__D.1 Installing Visual Studio Code
__D.2 Installing Conda
__D.3 Installing Python
__D.4 Installing Git

★ What this book covers ★

◎ A Practical Overview of Generative AI Applications
◎ Architectural patterns, integration guides, and best practices for generative AI
◎ Latest techniques such as RAG, prompt engineering, and multimodality
◎ Challenges and risks of generative AI, such as hallucinations and jailbreaks
How to Integrate Generative AI into Your Business and IT Strategies