Unit 1: Introduction
Artificial Intelligence and Its Applications
AI is the field of computer science that enables machines to mimic human intelligence.
Applications include:
Healthcare: Diagnostics, drug discovery
Finance: Fraud detection, algorithmic trading
Retail: Chatbots, recommendation engines
Transportation: Autonomous vehicles
Agriculture: Yield prediction, smart irrigation
AI Techniques
Search algorithms: BFS, DFS, A*
Machine Learning: Supervised, unsupervised, reinforcement learning
Logic and reasoning: Deductive and inductive methods
Knowledge representation: Semantic nets, frames
Levels of Models
Reactive Machines: No memory (e.g., Deep Blue)
Limited Memory: Uses past experiences (e.g., self-driving cars)
Theory of Mind: Understands emotions (in development)
Self-aware: Conscious machines (theoretical)
Criteria of Success
Accuracy, efficiency, adaptability, scalability, explainability.
Intelligent Agents
Agents that perceive the environment and act rationally to achieve goals.
Nature of Agents
Types:
Simple Reflex
Model-based Reflex
Goal-based
Utility-based
Learning Agents
Learning Agents
Can improve over time. Components:
Learning Element
Performance Element
Critic
Problem Generator
Advantages and Limitations of AI
Advantages: Speed, accuracy, automation, 24/7 availability
Limitations: Lack of common sense, high cost, bias in data
Impact and Examples
Positive: Enhanced productivity, automation, innovation
Negative: Job loss, surveillance, ethical dilemmas
Application Domains
Healthcare, education, robotics, customer service, military, smart homes.
The AI Ladder
1.
Collect: Gather and organize data
2.
Organize: Clean and curate data
3.
Analyze: Use AI models
4.
Infuse: Integrate AI into workflows
Career Advice
Learn programming (Python), math (linear algebra, probability), AI tools (TensorFlow, PyTorch), and keep up with research.
Hotbeds of AI Innovation
USA (Silicon Valley), China, Canada (Toronto), UK, India (Bangalore), Germany.
Unit 2: Problem Solving Techniques
State Space Search
Represents problems as a set of states with transitions (actions).
Control Strategies
Uninformed: BFS, DFS
Informed: Greedy, A*
Heuristic Search
Uses domain-specific knowledge. E.g., A* combines cost-so-far and estimated cost-to-goal.
Problem Characteristics
Deterministic/Stochastic
Static/Dynamic
Discrete/Continuous
Single/Multi-agent
Production System Characteristics
Set of rules, working memory, control strategy, rule applier.
Search Algorithms
Generate and Test: Randomly generate and check solutions
Hill Climbing: Move towards better state
Best First Search: Uses a heuristic
A Search*: Combines path cost and heuristic
Constraint Satisfaction Problem (CSP)
Solve problems with constraints on variables. E.g., Sudoku.
Means-End Analysis
Reduce the difference between the current and goal state by identifying actions.
Game Playing
Minimax: Choose best move assuming opponent plays optimally
Alpha-Beta Pruning: Skip branches that won't affect final decision
Iterative Deepening
Combines depth-first and breadth-first by progressively deepening the search.
Unit 3: Logic
Propositional Logic
Uses variables (true/false), connectives: AND, OR, NOT, ,
Predicate Logic
More expressive; includes objects, functions, quantifiers ( , )
Resolution
A rule of inference used in automated theorem proving
Clause Form
Convert logical sentences into CNF (conjunctive normal form) for resolution
Unification Algorithm
Finds a substitution that makes different logical expressions identical
Unit 4: Knowledge Representation Schemes and Reasoning
Mapping Between Facts and Representations
Transform real-world knowledge into computable data.
Approaches to Knowledge Representation
Logical: Propositional, Predicate logic
Procedural: Encodes knowledge as procedures
Declarative: States facts and rules
Procedural vs Declarative
Procedural: “How” knowledge
Declarative: “What” knowledge
Forward vs Backward Reasoning
Forward: From facts to conclusions
Backward: From goals to supporting facts
Matching
Compare data with rules/templates
Conflict Resolution
Decide which rule to apply when multiple are applicable
Non-monotonic Reasoning
Allows withdrawal of conclusions when new facts are added
Default Reasoning
Assume a value unless contradicted
Statistical Reasoning
Handle uncertainty using probability (e.g., Bayesian networks)
Fuzzy Logic
Handle partial truth values between 0 and 1
Weak and Strong Slot-Filler Structures
Weak: Simple attribute-value pairs
Strong: Rich semantics (e.g., frames)
Semantic Networks
Graph of concepts and relations
Frames
Data structures for stereotyped situations
Conceptual Dependency
Represents meaning of sentences using a set of primitives
Scripts
Predefined causal chain of events (e.g., restaurant script)
Unit 5: Planning
The Planning Problem
Find a sequence of actions to achieve a specific goal
Planning with State Space Search
Explores sequences of actions from initial to goal state
Partial Order Planning
Not all steps are ordered; more flexible
Planning Graphs
Graphical representation of action sequences and effects
Planning with Propositional Logic
Uses logical inference to determine valid plans
Analysis of Planning Approaches
Compares efficiency, scalability, and applicability of methods
Hierarchical Planning
Breaks problems into subproblems (HTN – Hierarchical Task Network)
Conditional Planning
Plans include conditions and branches based on outcomes
Continuous and Multi-Agent Planning
Continuous: Actions over time
Multi-agent: Coordination among multiple agents
