Artificial Intelligence: An Accessible Approach
Pages: est 350-400
|This book is in the process of being written, with expected completion in early to mid 2023.|
Table of contents
From surveillance tracking systems to conversational bots, disease diagnosis to addictive content recommenders, Artificial Intelligence is playing a bigger role in our daily lives and our digital society than ever – but few people understand how it works. Automating Minds explores and exposes the modern Artificial Intelligence field from an insider’s perspective for the general public.
After understanding fundamental principles in machine learning and neural networks, you’ll learn about the inner workings of important AI applications like video call background blurring, intelligent question answering systems, content recommendation on platforms like YouTube, art generation, and much more. You’ll also learn how to think critically about sources, ethics, and responses to bias and fairness in AI systems – an increasingly important topic in the field. Moreover, to move with the constantly progressing nature of research in AI, Automating Minds guides you through the AI research development space and how to extract high-level information from research papers for yourself instead of reading it second-hand from often incomplete popular science coverage.
Written for a wide audience, Automating Minds will provide you the concepts and knowledge to understand, think about, and interact with the quickly evolving world of AI development and research more deeply.
- This book addresses a gap in the market: there are no books that seek to make AI accessible to the general public while not being afraid to dive into technical concepts and internal dynamics.
- The book uses an application-first pedagogical style, relying upon familiar AI applications the reader has likely interacted with or heard about to make concepts more accessible, relevant, and engaging.
- A significant section of the book is dedicated towards helping the reader obtain basic literacy with AI research literature, which allows them to keep up with AI developments and to do so directly from the primary literature rather than second-hand from often inaccurate popular science coverage.
- The book engages in important discussion on AI history, philosophy, and fairness/responsibility issues while remaining grounded in technical concepts and research.
Tentative Table of Contents
Page count approximation: 350-400 pages
Chapter 1: History of Artificial Intelligence
Chapter Goal: To understand patterns, trends, developments, and controversies of artificial intelligence throughout history and the key discoveries that have contributed to the state of machine learning today.
- Data Modeling throughout History
- Development of the Computer and Basic Computability Theory
- Alan Turing’s Arguments for the Intelligence of Machines
- IBM DeepBlue and Elementary AI
- Symbolic AI and Expert Systems
- AI Winters
- Discovery of Neural Network
- ImageNet and the Image Recognition Race
- Natural Language Processing and Transformers
- AI and Society Today
Chapter 2: Automating Intelligence: Machine Learning Concepts
Chapter Goal: To understand core concepts, ideas, and principles in machine learning, with a focus on different mathematical and computational interpretations of ‘learning’.
- Learning Modalities
- Data Hierarchies, Organization, and Practices
- Simple Machine Learning Algorithms
- Learning as Iterative Parameter Update
- Learning as Discovering Feature Space Mapping
- Searle’s Chinese Room
Chapter 3: Beginnings of a Mind: Fundamental Neural Network Theory
Chapter Goal: To understand the inspiration, design, and properties of feed-forward artificial neural networks; as well as both early and modern applications.
- Building a Neural Network – Assembling Perceptrons and Layers
- Optimizing a Neural Network – Backpropagation & Gradient Descent
- Neural Network Properties – Universal Approximation Theorem, Simple Feature Space Learning
- Early Applications of Neural Networks (Protein Structure Prediction, Handwritten Digit Recognition)
- YouTube Recommendation – Building Deep Learning Systems
- Deep Double Descent – Bizarre Phenomena in Deep Learning
Chapter 4: Seeing and Drawing: Computer Vision
Chapter Goal: To understand key mechanisms and concepts supporting neural networks with image inputs and/or image outputs, and their applications and societal implications.
- Image Recognition – Image-Domain Classification
- Facial Recognition – Image Embeddings and Few-Shot Learning
- Blurring Video Call Backgrounds – Semantic Segmentation
- AI, Artist – Generative and Reconstructive Models
- DeepFake – Socially Dangerous Generative Models
Chapter 5: Reading and Speaking: Natural Language Processing
Chapter Goal: To understand key mechanisms and concepts supporting neural networks with text inputs and/or text outputs, and their applications and societal implications.
- Spam Removal – Text Classification
- Hands-Free AI on the Internet – Unsupervised/Self-Supervised NLP
- Detecting Toxicity on Social Media – Subjective Text Classification
- Article-Writing AI – Generative Text Models
- Intelligent Writing Improvement/Grammarly – Suggestion Models
- Translation – Neural Machine Translation
- Hey, Alexa/Siri/Google – Question Answering Systems
Chapter 6: Ethical, Responsible, Fair AI
Chapter Goal: To understand statistical interpretations and measurements of fairness, problematic AI behavior and its possible causes, and AI model attacks.
- Statistical Fairness Metrics and Ideas
- Problematic Behavior in AI Applications
- Bias Dataset Development and Data Crowdsourcing Challenges
- Explaining and Interpreting AI
- Inherent Model Bias or Dataset-Embedded Bias Debate
- Adversarial Learning and Model Corruption
Chapter 7: Reading AI Research Papers for Yourself
Chapter Goal: To be exposed to the AI research space and process, and to be able to read most deep learning papers and understand the key findings at a very high level. No of pages: 40 Sub - Topics:
- The AI Research Space and Process
- How AI Research Papers are Structured
- Case Studies – Step by Step Analysis of Example AI Research Papers
Chapter 8: Into the Future: Exciting Topics in Deep Learning Research
Chapter Goal: To explore various exciting research topics pushing deep learning’s frontiers forward.
- Cross-modality networks (text to speech, speech to text, image to text, text to image, converting tabular data into images)
- Model Compression
- Foundation Model Debate
- The Interpolation or Extrapolation Debate
- Neural Architecture Search
- Modern Neural Network Learning Theory and Representation