1. What is Artificial Intelligence?
Artificial Intelligence (AI) is a branch of computer science concerned with building systems capable of performing tasks that normally require human intelligence.
These tasks include:
- Learning
- Reasoning
- Problem solving
- Decision making
- Natural language processing
- Perception
- Planning
Examples of AI
- Chatbots
- Recommendation systems
- Speech recognition
- Image recognition
- Expert systems
- Autonomous vehicles
- Game-playing programs
2. Goals of Artificial Intelligence
Major goals of AI include:
- Developing intelligent machines
- Automating complex tasks
- Simulating human reasoning
- Learning from data and experience
- Solving complex problems
- Making intelligent decisions
3. Intelligent Agents
An agent is an entity that perceives its environment through sensors and acts upon that environment using actuators.
Basic Model
Environment → Sensors → Agent → Actuators → Environment
Examples
| Agent | Sensors | Actuators |
|---|---|---|
| Robot | Camera, sensors | Motors |
| Self-driving car | Camera, radar | Steering, brakes |
| Vacuum cleaner | Dirt sensor | Wheels, suction |
4. Types of Intelligent Agents
Important types include:
1. Simple Reflex Agent
Acts according to the current percept.
Condition → Action
Example:
If dirt is detected → clean it.
2. Model-Based Reflex Agent
Maintains an internal model/state of the environment.
3. Goal-Based Agent
Chooses actions based on a desired goal.
4. Utility-Based Agent
Selects actions based on a utility or preference measure.
5. Learning Agent
Improves its performance through experience.
Exam Tip:
Simple Reflex → Current percept
Model-Based → Internal state
Goal-Based → Goal
Utility-Based → Utility
Learning Agent → Experience
5. Problem Solving in AI
AI problem solving involves finding a sequence of actions that transforms an initial state into a goal state.
A problem can generally be represented using:
- Initial state
- Actions/operators
- Transition model
- Goal test
- Path cost
Example: 8-Puzzle
Initial State → Possible Moves → Search → Goal State
The AI searches through possible states until it finds a solution.
6. State Space Representation
A state space represents all possible states that can be reached from an initial state through valid actions.
For example, in the 8-puzzle:
- Each board configuration = State
- Legal movement = Operator
- Starting configuration = Initial state
- Desired configuration = Goal state
7. Search Algorithms in AI
Search algorithms are fundamental to AI problem solving.
They can broadly be divided into:
Uninformed Search
- Breadth First Search (BFS)
- Depth First Search (DFS)
- Uniform Cost Search
- Depth-Limited Search
- Iterative Deepening Search
Informed Search
- Best First Search
- Greedy Best First Search
- A* Search
8. Breadth First Search (BFS)
BFS explores nodes level by level.
It generally uses a queue (FIFO) data structure.
Characteristics
- Complete under standard finite-branching assumptions
- Optimal when all step costs are equal
- Can require large memory
Example
A / \ B C / \ / \ D E F G
BFS order:
A → B → C → D → E → F → G
9. Depth First Search (DFS)
DFS explores one branch as deeply as possible before backtracking.
It can be implemented using a stack or recursion.
For the above tree, one possible DFS traversal is:
A → B → D → E → C → F → G
Advantage
Requires less memory than BFS in many search-tree settings.
Limitation
It may follow a very deep or infinite path and does not generally guarantee the shortest solution.
10. Uniform Cost Search
Uniform Cost Search expands the node with the lowest path cost.
It is useful when actions have different costs.
Key point:
UCS chooses the least-cost path rather than simply the shallowest node.
11. Heuristic Search
A heuristic is a function that estimates the cost or distance from a current state to the goal.
It is commonly represented as:
h(n)
where h(n) is the estimated cost from node n to the goal.
12. Greedy Best First Search
Greedy Best First Search chooses the node with the smallest heuristic value.
It uses:
f(n) = h(n)
It can be fast but does not generally guarantee an optimal solution.
13. A* Search Algorithm
A* combines the cost already incurred with an estimated cost to the goal.
The evaluation function is:
f(n) = g(n) + h(n)
Where:
-
g(n)= cost from the initial state to noden -
h(n)= estimated cost fromnto the goal -
f(n)= estimated total solution cost throughn
Important Exam Point
A* can be complete and optimal under appropriate conditions, particularly when the heuristic is admissible (and for graph search, suitable consistency conditions are typically used).
14. Knowledge Representation
Knowledge Representation (KR) is concerned with representing information about the world in a form that an AI system can use for reasoning.
Common techniques include:
- Logic
- Semantic networks
- Frames
- Production rules
- Ontologies
15. Propositional Logic
Propositional logic represents statements that are either true or false.
Example:
P = "It is raining."
P can have the value:
True or False
Important logical operators:
| Operator | Meaning |
|---|---|
| AND (∧) | Conjunction |
| OR (∨) | Disjunction |
| NOT (¬) | Negation |
| → | Implication |
| ↔ | Biconditional |
16. Predicate Logic
Predicate logic provides more expressive representation than propositional logic.
Example:
Human(Socrates)
This represents that Socrates is a human.
Predicate logic can represent:
- Objects
- Properties
- Relationships
- Quantifiers
Important quantifiers:
∀ = For all
∃ = There exists
17. Semantic Network
A semantic network represents knowledge using:
Nodes + Links
Example:
Animal | is-a | Bird / \ has can wings fly
It represents relationships between concepts.
18. Frames
A frame represents knowledge using a structured collection of attributes or slots.
Example:
Frame: Student Name: Rahul Age: 22 Course: Computer Science
Frames are useful for representing structured knowledge.
19. Production Rules
Production systems represent knowledge using rules of the form:
IF condition THEN action/conclusion
Example:
IF temperature > 38°C
THEN patient may have fever
Production rules are widely associated with rule-based expert systems.
20. Expert Systems
An expert system is an AI system designed to solve problems in a particular domain using expert knowledge.
Main components include:
Knowledge Base
Contains domain knowledge.
Inference Engine
Applies reasoning mechanisms to the knowledge.
User Interface
Allows interaction between the user and the system.
Explanation Facility
May explain how a conclusion was reached.
21. Forward Chaining
Forward chaining is data-driven reasoning.
It starts with known facts and applies rules to derive new facts.
Facts ↓ Rules ↓ New Facts ↓ Conclusion
22. Backward Chaining
Backward chaining is goal-driven reasoning.
It starts with a goal and works backward to determine which facts/rules can support that goal.
Goal ↓ Required Conditions ↓ Facts
Easy Difference
Forward Chaining → Facts → Conclusion
Backward Chaining → Goal → Facts
23. Machine Learning
Machine Learning (ML) is a branch of AI in which systems learn patterns from data and improve their performance without being explicitly programmed for every individual case.
Major types include:
- Supervised Learning
- Unsupervised Learning
- Reinforcement Learning
24. Supervised Learning
The training data contains input-output pairs / labeled examples.
Examples:
- Classification
- Regression
Classification
Predicts a category.
Example:
Email → Spam / Not Spam
Regression
Predicts a numerical value.
Example:
House features → Estimated price
25. Unsupervised Learning
The training data does not provide target labels.
Common tasks include:
- Clustering
- Dimensionality reduction
- Association analysis
Example
Customer data can be grouped into different customer segments without predefined labels.
26. Reinforcement Learning
In reinforcement learning, an agent interacts with an environment and learns from rewards and penalties.
Basic model:
Agent → Action → Environment → Reward
The objective is generally to learn a policy that maximizes cumulative reward.
Bihar STET & BPSC Artificial Intelligence MCQs
Q1. Artificial Intelligence is primarily concerned with:
A. Designing only computer hardware
B. Making machines capable of intelligent behavior
C. Increasing monitor resolution
D. Managing computer networks only
Answer: B. Making machines capable of intelligent behavior
Q2. An AI agent perceives its environment using:
A. Actuators
B. Sensors
C. Compilers
D. Databases
Answer: B. Sensors
Q3. Which component allows an agent to act on its environment?
A. Sensor
B. Actuator
C. Compiler
D. Knowledge base
Answer: B. Actuator
Q4. Which agent maintains an internal representation of the environment?
A. Simple reflex agent
B. Model-based reflex agent
C. Random agent
D. Static agent
Answer: B. Model-based reflex agent
Q5. Which search algorithm generally uses a FIFO queue?
A. DFS
B. BFS
C. A*
D. Hill climbing
Answer: B. BFS
Q6. Which search algorithm generally uses a stack?
A. BFS
B. DFS
C. Uniform Cost Search
D. Greedy Search
Answer: B. DFS
Q7. BFS explores a search tree:
A. Depth-wise
B. Level-wise
C. Randomly
D. From goal to root only
Answer: B. Level-wise
Q8. Which search algorithm uses the evaluation function f(n) = g(n) + h(n)?
A. DFS
B. BFS
C. A*
D. Hill Climbing
Answer: C. A*
Q9. In A* search, g(n) represents:
A. Estimated cost to the goal
B. Cost from the initial state to node n
C. Total number of nodes
D. Heuristic error
Answer: B. Cost from the initial state to node n
Q10. In A* search, h(n) represents:
A. Cost already incurred
B. Estimated cost from node n to the goal
C. Number of children
D. Branching factor
Answer: B. Estimated cost from node n to the goal
Q11. Which search uses only the heuristic function for node selection?
A. Greedy Best First Search
B. BFS
C. DFS
D. Uniform Cost Search
Answer: A. Greedy Best First Search
Q12. A heuristic function is used to:
A. Estimate the cost to reach the goal
B. Encrypt data
C. Store images
D. Compile programs
Answer: A. Estimate the cost to reach the goal
Q13. Which technique represents knowledge using nodes and relationships?
A. Semantic network
B. Stack
C. Queue
D. Hash table
Answer: A. Semantic network
Q14. Which logic uses predicates and quantifiers?
A. Boolean algebra only
B. Predicate logic
C. Machine language
D. Assembly language
Answer: B. Predicate logic
Q15. What does the symbol ∀ represent?
A. There exists
B. For all
C. Not equal
D. Implication
Answer: B. For all
Q16. Which reasoning technique is goal-driven?
A. Forward chaining
B. Backward chaining
C. Random search
D. BFS
Answer: B. Backward chaining
Q17. Forward chaining is:
A. Goal-driven
B. Data-driven
C. Random-driven
D. Image-driven
Answer: B. Data-driven
Q18. Which component of an expert system contains domain knowledge?
A. Knowledge base
B. Compiler
C. Operating system
D. Sensor
Answer: A. Knowledge base
Q19. Which component of an expert system applies rules to derive conclusions?
A. Inference engine
B. User interface
C. Sensor
D. Actuator
Answer: A. Inference engine
Q20. Which type of machine learning uses labeled training data?
A. Unsupervised learning
B. Supervised learning
C. Reinforcement learning
D. Random learning
Answer: B. Supervised learning
Q21. Which of the following is a supervised learning task?
A. Classification
B. Clustering
C. Random sampling
D. Data compression
Answer: A. Classification
Q22. Which machine learning technique is commonly used to group unlabeled data?
A. Classification
B. Clustering
C. Regression
D. Supervised learning
Answer: B. Clustering
Q23. Which type of learning uses rewards and penalties?
A. Supervised learning
B. Unsupervised learning
C. Reinforcement learning
D. Rule-based learning
Answer: C. Reinforcement learning
Q24. In reinforcement learning, the learner is called:
A. Agent
B. Compiler
C. Database
D. Parser
Answer: A. Agent
Q25. Which of the following is an example of regression?
A. Spam/Not Spam
B. Cat/Dog classification
C. Predicting house price
D. Grouping customers
Answer: C. Predicting house price
Important AI Revision Points
| Topic | Key Point |
|---|---|
| AI | Simulation/automation of intelligent behavior |
| Agent | Perceives and acts on environment |
| Sensor | Receives environmental information |
| Actuator | Performs actions |
| BFS | Level-order search; FIFO queue |
| DFS | Depth-first search; stack/recursion |
| UCS | Expands lowest path-cost node |
| Heuristic | Estimates remaining cost |
| Greedy Best First | f(n)=h(n) |
| A* | f(n)=g(n)+h(n) |
| Semantic Network | Nodes + relationships |
| Frame | Structured knowledge representation |
| Forward Chaining | Data-driven |
| Backward Chaining | Goal-driven |
| Expert System | Knowledge base + inference engine |
| Supervised Learning | Labeled data |
| Unsupervised Learning | Unlabeled data |
| Reinforcement Learning | Reward/penalty |
| Classification | Predicts category |
| Regression | Predicts numerical value |
| Clustering | Groups similar data |