AI Defini ons
AI stands for Ar ficial Intelligence, which refers to machines or programs that can perform
tasks usually requiring human intelligence, like learning, reasoning, or problem-solving. For
example, a chatbot answering your ques ons or a self-driving car deciding when to stop are
applica ons of AI.
Founda ons of AI
AI is built on ideas from various fields:
1. Mathema cs: Logic (reasoning), probability (handling uncertainty), and sta s cs
(making predic ons).
2. Computer Science: Algorithms (step-by-step procedures) and data structures (ways
to organize informa on).
3. Psychology: Understanding how humans think and learn.
4. Neuroscience: Studying how the brain processes informa on.
5. Philosophy: Exploring concepts like decision-making, ethics, and intelligence.
History of AI
1950s: Alan Turing proposed the Turing Test to define intelligent machines. Early AI
focused on simple logic and problem-solving.
1956: The Dartmouth Conference officially launched AI research.
1970s-1980s: Development of rule-based expert systems (e.g., diagnosing diseases).
1990s-2000s: AI began solving real-world problems (e.g., IBM’s Deep Blue defeated
chess champion Garry Kasparov).
2010s-Present: Advances in machine learning and deep learning allow AI to perform
complex tasks like recognizing faces, understanding speech, and driving cars.
State of the Art
Modern AI technologies include:
1. Natural Language Processing (NLP): AI that understands and generates human
language (e.g., chatbots, voice assistants like Siri).
2. Computer Vision: AI that interprets images and videos (e.g., facial recogni on, object
detec on).
3. Robo cs: Robots powered by AI to perform tasks like cleaning or surgery.
4. Genera ve AI: AI systems like ChatGPT and DALL-E that can create text, art, or music.
Intelligent Agents
Agents and Environments
Agent: A system that senses its environment (input) and takes ac ons (output) to
achieve its goals. Example: A vacuum cleaner robot senses dirt and cleans it.
Environment: The world around the agent. For example, the vacuum cleaner’s
environment is the house.
Good Behavior
An agent behaves ra onally when it acts in a way that maximizes its success based on its
knowledge and goals. For example, a chess-playing AI chooses moves that increase its
chances of winning.
The Nature of Environments
Environments can be:
1. Sta c vs. Dynamic: Sta c environments don’t change (e.g., a chessboard), while
dynamic environments do (e.g., traffic for self-driving cars).
2. Discrete vs. Con nuous: Discrete environments have dis nct parts (e.g., board
games), while con nuous ones have endless possibili es (e.g., flying a drone).
3. Fully vs. Par ally Observable: Fully observable environments give the agent all the
informa on (e.g., chess), while par ally observable ones don’t (e.g., poker).
Structure of Agents
Agents can have different designs:
1. Simple Reflex Agents: React based on current condi ons. Example: A thermostat
turning on/off based on room temperature.
2. Goal-Based Agents: Consider future goals. Example: A GPS system plans the best
route to a des na on.
3. Learning Agents: Improve their performance over me by learning from experience.
Example: A chatbot ge ng be er at answering ques ons.
Problem Solving
Problem-Solving Agents
These agents plan sequences of ac ons to achieve goals. For example, a delivery robot plans
the shortest path to deliver a package.
Example Problems
1. Puzzles: Solving a Sudoku or maze.
2. Real-world Problems: Naviga ng traffic, scheduling flights, or controlling a robot
arm.
Searching for Solu ons
Finding the steps to achieve a goal. For example, in a maze, the solu on is the path from
start to finish.
Uninformed Search Strategies
Search methods that don’t use addi onal knowledge about the problem:
1. Breadth-First Search: Explores all op ons one level at a me.
2. Depth-First Search: Explores one path deeply before backtracking.
Avoiding Repeated States
To save me, agents remember states they’ve already visited to avoid revisi ng them.
Example: In a maze, marking visited paths.
Searching with Par al Informa on
When the agent doesn’t have complete knowledge, it uses explora on or makes
assump ons. For example, a robot exploring an unknown building maps it as it moves.
