Artificial Intelligence - UNIT - 1 Topic : 2. AI_PROBLEMS
UNIT - 1
2. AI PROBLEMS
1. Introduction
AI problems refer to tasks or scenarios that require intelligence to
solve—such as planning, reasoning, learning, decision-making, and perception.
These problems can range from simple tasks (like playing tic-tac-toe) to
complex ones (like autonomous driving or medical diagnosis).
2. Characteristics of AI
Problems
Understanding the nature of AI problems helps in designing
effective solutions.
a. Complexity
- May
involve large search spaces or huge amounts of data.
- Problems
may be computationally expensive (NP-hard).
b. Uncertainty
- AI
often works with incomplete, noisy, or ambiguous
data.
- Example:
Speech recognition under background noise.
c. Dynamic Environment
- The
environment can change during problem-solving.
- Example:
Self-driving cars reacting to traffic.
d. Goal-Driven
- AI
problems typically involve achieving a defined goal or objective.
- Example:
Chess AI tries to win the game.
e. Real-Time Response
- Some
problems require decisions in real-time.
- Example:
AI in robotics or missile systems.
3. Types of AI Problems
AI problems can be broadly classified into the following categories:
3.1 Search Problems
These involve finding a path or solution from a start state to a goal
state.
- Examples:
- Solving mazes
- Pathfinding in maps
- Puzzle-solving (e.g., 8-puzzle)
- Techniques:
- Breadth-First Search (BFS)
- Depth-First Search (DFS)
- A* Algorithm
3.2 Game Playing
AI is used to create agents that can play and often win in competitive
games.
- Examples:
- Chess, Go, Checkers, Tic-tac-toe
- Characteristics:
- Adversarial (opponent involved)
- Involves strategy and planning
- Techniques:
- Minimax algorithm
- Alpha-Beta pruning
3.3 Planning Problems
Involves creating a sequence of actions to achieve a goal from a
given initial state.
- Examples:
- Robot navigation
- Task scheduling
- Techniques:
- STRIPS (Stanford Research Institute Problem
Solver)
- Partial Order Planning
3.4 Reasoning and Logic
Problems
Require the system to infer new facts from known information
using logic.
- Examples:
- Medical diagnosis
- Legal reasoning
- Techniques:
- Propositional logic
- Predicate logic
- Inference rules
3.5 Natural Language Understanding
(NLU)
AI systems understand and process human languages like English.
- Examples:
- Chatbots
- Language translation
- Challenges:
- Ambiguity
- Syntax and semantics
- Techniques:
- Parsing
- Semantic analysis
- NLP algorithms (e.g., transformers)
3.6 Perception and Robotics
Involves sensing and interpreting the physical world to interact
meaningfully.
- Examples:
- Vision systems in autonomous vehicles
- Robots detecting and grasping objects
- Techniques:
- Image processing
- Computer vision
- Sensor fusion
3.7 Learning Problems
Involve learning from data instead of being explicitly
programmed.
- Examples:
- Spam email detection
- Movie recommendation
- Techniques:
- Supervised Learning
- Unsupervised Learning
- Reinforcement Learning
4. Problem Formulation in
AI
Before solving a problem, it must be formulated properly. This
includes defining:
- Initial state – Starting point
- Actions –
Possible operations
- Transition model – Outcome of actions
- Goal test –
Whether the goal is achieved
- Path cost –
Measure of solution cost
A well-formulated problem helps AI algorithms work efficiently and
effectively.
5. Example Problem:
8-Puzzle
Problem:
- A 3x3 board with tiles numbered 1 to 8 and one
blank.
- The goal is to move the tiles to reach a specific
goal configuration.
Formulation:
- State:
Current configuration of tiles
- Actions:
Move blank up/down/left/right
- Goal Test:
Matches goal configuration
- Path Cost:
Number of moves
Solution Methods:
- BFS (for shortest path)
- A* using heuristic like Manhattan Distance
|
Problem |
Domain |
Solution Technique |
|
Medical Diagnosis |
Healthcare |
Rule-based reasoning, ML |
|
Speech Recognition |
NLP |
Neural Networks, HMM |
|
Self-driving Cars |
Robotics |
Deep Learning, SLAM |
|
Fraud Detection |
Finance |
Anomaly Detection, ML |
|
Recommendation Systems |
E-commerce |
Collaborative Filtering |
7. Common Challenges in AI
Problem Solving
- State-space explosion: Too many possible states
- Ambiguity:
Incomplete or unclear input
- Uncertainty: Inaccurate
or noisy data
- Real-time constraints: Need fast decisions
- Scalability:
Problem grows with input size
- Ethical issues: Fairness, bias, and misuse
8. Key Terms to Remember
- State Space:
Set of all possible states
- Heuristic:
Rule of thumb for guiding search
- Agent:
Entity that perceives and acts
- Environment:
External system where the agent operates
- Goal State:
Desired end condition
9. Summary
- AI problems simulate real-world intelligent
behavior.
- They include search, logic, planning, language,
perception, and learning.
- Solving these problems requires proper
formulation and appropriate techniques.
- Challenges include uncertainty, complexity,
and dynamic environments.
Comments
Post a Comment