Neuro-Symbolic AI: Design transparent and trustworthy systems that understand the world as you do

Preface
1
The Evolution and Pitfalls of AI
The basic idea behind AI
The evolution of AI
Philosophy
Logic
Mathematics
Cognitive science
A short history of AI
Subfields of AI
ML
Computer vision
Natural language processing
Robotics
Knowledge representation
Problem-solving and reasoning
Planning
Evolutionary computing
The pitfalls of AI
Is AI limitless?
How important is the data?
Can we get training data?
Have we got good data?
Can a high-performance AI still fail?
Summary
2
The Rise and Fall of Symbolic AI
Defining Symbolic AI
Humans, symbols, and signs
Enabling machine intelligence through symbols
The concept of intelligence
Towards Symbolic AI
From symbols and relations to logic rules
The fall of Symbolic AI
Symbolic AI today
Expert systems
Natural language processing
Constraint satisfaction
Explainable AI
The sub-symbolic paradigm
Summary
Further reading
3
The Neural Networks Revolution
Artificial neural networks modeling the human brain
A simple artificial neural network
Introducing popular neural network architectures
Recurrent neural networks
Competitive networks
Hopfield networks
Delving into deep neural networks
Convolutional neural networks
Long short-term memory networks
Autoencoders
Deep belief networks
Generative networks
Transformers
The rise of data
The complexities and limitations of neural networks
Summary
4
The Need for Explainable AI
What is XAI?
Why do we need XAI?
XAI case studies
The state-of-the-art models in XAI
Accumulated Local Effects
Anchors
Contrastive Explanation Method
Counterfactual instances
Explainable Boosting Machine
Global Interpretation via Recursive Partitioning
Integrated gradients
Local interpretable model-agnostic explanations
Morris Sensitivity Analysis
Partial dependence plot
Permutation importance
Protodash
SHapley Additive exPlanations
Summary
5
Introducing Neuro-Symbolic AI – the Next Level of AI
The idea behind NSAI
Modeling human intelligence – insights from child psychology
The ingredients of an NSAI system
The symbolic ingredient
The neural ingredient
The neuro-symbolic blend
Exploring different architectures of NSAI
Neuro-Symbolic Concept Learner
Neuro-symbolic dynamic reasoning
Dissecting the NLM architecture
Summary
Further reading
6
A Marriage of Neurons and Symbols – Opportunities and Obstacles
The benefits of combining neurons and symbols
Data efficiency
High accuracy
Transparency and interpretability
The challenges of combining neurons and symbols
Knowledge and symbolic representation
Multi-source knowledge reasoning
Dynamic reasoning
Query understanding for knowledge reasoning
Research gaps in neuro-symbolic computing
Summary
7
Applications of Neuro-Symbolic AI
Application 1 – health – computational drug
Application details
Problem statement
The role of NSAI
Application 2 – education – student strategy prediction
Application details
Problem statement
The role of NSAI
Application 3 – finance – bank loan risk assessment
Application details
Problem statement
The role of NSAI
Summary
Further reading
8
Neuro-Symbolic Programming in Python
Environment and data setup
Solution 1 – logic tensor networks
Loading the dataset
Modifying the dataset
Creating train and test datasets
Defining our knowledge base and NN architecture
Defining our predicate, connectives, and quantifiers
Setting up evaluation parameters
Training the LTN model
Analyzing the results
Solution 2 – prediction stacking
Experiment setup and loading the data
Data preparation
Training our NSAI model
Analyzing the results
Prediction interpretability and logic tracing
Summary
Further reading
9
The Future of AI
Looking at fringe AI research
Small data
Novel network architectures
New ways of learning
Evolution of attention mechanisms
World model
Hybrid models
Exploring future AI developments
Quantum computing
Neuromorphic engineering
Brain-computer interaction
Bracing for the rise of AGI
Preparing for singularity
Popular media
Exploring the expert views
Singularity challenges
Summary
Further reading
Index
Other Books You May Enjoy