Unit 1: Introduction [6 Hours]
Artificial Intelligence focuses on building machines and computer programs that can perform tasks that normally require human intelligence. This unit introduces the meaning of intelligence and Artificial Intelligence, the major foundations and applications of AI, different ways of understanding AI through thought and behavior, and the concept of intelligent agents. It also covers agent structures, PEAS description, different types of agents, and the environments in which agents operate.
Tested Topics
- Intelligent Agent Types — 3x (2025, 2023, 2021)
- Agent Properties and Characteristics — 1x (2021)
- Artificial Intelligence Definition — 1x (2025)
- Foundations / Components of AI — 1x (2025)
- PEAS Framework — 1x (2021)
- Turing Test — 1x (2021)
- Understanding AI Based on Thought Process and Behavior — 1x (2026)
1.1 Artificial Intelligence
Intelligence
Intelligence is the ability to acquire knowledge, understand information, learn from experience, solve problems, reason about situations, and adapt to changing environments.
For example, when a student encounters a programming problem they have never seen before, they may analyze the problem, recall previously learned concepts, develop possible solutions, test them, and correct mistakes. These abilities are examples of intelligent behavior.
Intelligent Behavior
Intelligent behavior refers to actions that demonstrate abilities such as reasoning, learning, problem solving, decision making, perception, communication, and adaptation.
A system does not necessarily have to think exactly like a human to behave intelligently. For example, a chess-playing program may examine thousands of possible moves and select the move most likely to win. Its internal process is different from human thinking, but its behavior can still be considered intelligent.
Artificial Intelligence
Artificial Intelligence (AI) is a branch of computer science concerned with developing systems capable of performing tasks that normally require human intelligence, such as learning, reasoning, problem solving, perception, understanding language, and decision making.
Examples of AI systems include virtual assistants, recommendation systems, self-driving vehicles, language translators, medical diagnosis systems, and game-playing programs.
Let's Clarify
Suppose you use an online translation system to translate an English sentence into Nepali. A traditional program would require programmers to explicitly write rules for every possible sentence. An AI-based system can instead learn patterns from large amounts of language data and use those learned patterns to translate sentences it has never seen before.
The ability to use knowledge or learned patterns to deal with new situations is one of the major ideas behind Artificial Intelligence.
Understanding AI Based on Thought Process and Behavior
Artificial Intelligence can be understood using two major dimensions: whether we are interested in a system's thought process or its behavior, and whether the system attempts to imitate humans or act rationally.
This produces four commonly used approaches to understanding AI:
| Approach | Main Idea | Example |
|---|---|---|
| Thinking Humanly | Develop systems whose thinking process resembles human thinking. | A cognitive model designed to simulate how humans solve a puzzle. |
| Acting Humanly | Develop systems whose observable behavior resembles human behavior. | A chatbot communicating naturally enough that a person may think they are talking to another human. |
| Thinking Rationally | Develop systems that use logical reasoning to reach correct conclusions. | An expert system applying logical rules to determine a diagnosis. |
| Acting Rationally | Develop agents that select actions expected to achieve the best result. | A navigation system choosing the best route to reach a destination. |
1. Thinking Humanly
The thinking humanly approach attempts to understand how humans actually think and then create computer systems that follow similar cognitive processes.
This approach is closely related to cognitive science and psychology. Human problem solving may be studied through observation, experiments, or brain activity, and the resulting knowledge can be used to develop computational models.
2. Acting Humanly
The acting humanly approach judges intelligence according to the external behavior of a machine. If the machine behaves sufficiently like a human, it may be considered intelligent.
The best-known idea associated with this approach is the Turing Test.
3. Thinking Rationally
The thinking rationally approach focuses on correct reasoning. A system uses logical rules and available facts to derive conclusions.
For example, consider:
- All humans are mortal.
- Ram is a human.
- Therefore, Ram is mortal.
The conclusion is obtained using logical reasoning.
4. Acting Rationally
The acting rationally approach focuses on selecting actions that provide the best expected result according to available information.
A rational agent does not necessarily imitate human thinking. Its objective is to choose an appropriate action that helps achieve its goal.
For example, a self-driving vehicle may detect another vehicle suddenly entering its lane and immediately apply the brakes. The action is selected because it provides the safest expected outcome.
Summary
- Intelligence includes learning, reasoning, problem solving, decision making, and adaptation.
- AI develops computer systems capable of performing tasks associated with human intelligence.
- AI can be understood as thinking humanly, acting humanly, thinking rationally, or acting rationally.
- Modern AI often emphasizes rational agents that perceive their environment and select suitable actions.
Test Yourself
- What is Artificial Intelligence?
- What do you mean by intelligent behavior?
- Explain understanding AI based on thought process and behavior with example. (2026)
- Differentiate between thinking humanly and acting humanly.
- Differentiate between thinking rationally and acting rationally.
Hints:
- AI is the field of developing systems capable of tasks requiring intelligence such as learning, reasoning, perception, and decision making.
- Intelligent behavior demonstrates reasoning, learning, problem solving, adaptation, perception, or decision making.
- Explain the four approaches: thinking humanly, acting humanly, thinking rationally, and acting rationally, with a simple example of each.
- Thinking humanly focuses on the internal cognitive process; acting humanly focuses on observable human-like behavior.
- Thinking rationally focuses on correct logical reasoning; acting rationally focuses on choosing actions that produce the best expected outcome.
Hard AI vs. Soft AI
Hard AI refers to the idea of creating machines with genuine human-like intelligence and understanding. Such a machine would not merely simulate intelligent behavior but would possess general intellectual capabilities comparable to humans.
Soft AI focuses on developing systems that perform specific intelligent tasks without claiming that the machine actually possesses human consciousness or understanding.
| Hard AI | Soft AI |
|---|---|
| Aims for genuine human-like intelligence. | Aims to simulate intelligence for particular tasks. |
| Associated with general intelligence. | Usually designed for limited or specific problems. |
| Would understand and reason across many domains. | Works within the domain for which it is designed. |
| Largely a theoretical goal. | Widely used in present-day applications. |
Strong AI vs. Weak AI
Strong AI refers to AI capable of general human-level intellectual abilities, allowing the system to understand, learn, reason, and apply knowledge across different types of problems.
Weak AI, also called narrow AI, is developed to perform a particular task or a limited range of tasks.
Examples of weak AI include spam filters, recommendation systems, voice assistants, translation systems, and chess-playing programs.
Common Mistake
Do not assume that a system is Strong AI simply because it performs one task better than a human. A chess program may defeat the world's best chess players but still be unable to perform unrelated tasks such as preparing a meal or understanding a completely different problem. Excellent performance in one narrow domain does not automatically mean general intelligence.
1.2 Foundations of Artificial Intelligence
Artificial Intelligence is not based on computer science alone. It developed from ideas contributed by several fields concerned with reasoning, learning, decision making, language, computation, and intelligent behavior.
Major Foundations of AI
- Philosophy: contributed ideas about reasoning, knowledge, logic, mind, and intelligence.
- Mathematics: provides logic, probability, statistics, optimization, and mathematical methods used by AI algorithms.
- Economics: contributes decision theory, utility, and methods for selecting actions when resources and outcomes are uncertain.
- Neuroscience: studies how the brain processes information and inspired ideas such as artificial neural networks.
- Psychology: contributes knowledge about human learning, perception, cognition, memory, and problem solving.
- Computer Engineering: provides the hardware and computational systems needed to execute AI algorithms efficiently.
- Control Theory: contributes concepts related to feedback, stability, and systems that continuously adjust their actions.
- Linguistics: provides knowledge about language structure and meaning, forming an important foundation of Natural Language Processing.
Let's Clarify
Consider a voice assistant. Linguistics helps it understand language, mathematics and probability help it choose likely interpretations, psychology contributes ideas about human communication, computer engineering provides the hardware, and computer science provides the algorithms that connect everything together.
This is why AI is considered an interdisciplinary field.
Summary
- AI developed from several disciplines rather than from computer science alone.
- Major foundations include philosophy, mathematics, economics, neuroscience, psychology, computer engineering, control theory, and linguistics.
- Each discipline contributes concepts needed for reasoning, learning, perception, decision making, or communication.
Test Yourself
- Define Artificial Intelligence. Explain its components/foundations. (2025)
- Why is Artificial Intelligence considered an interdisciplinary field?
- Explain the contribution of mathematics and psychology to AI.
Hints:
- Start with the AI definition, then explain major foundations such as philosophy, mathematics, economics, neuroscience, psychology, engineering, control theory, and linguistics.
- AI combines concepts from several disciplines needed for reasoning, learning, decision making, perception, and communication.
- Mathematics contributes logic, probability and optimization; psychology contributes knowledge about human cognition, learning and problem solving.
1.3 Applications of Artificial Intelligence
Artificial Intelligence is used wherever computers need to analyze information, recognize patterns, make predictions, understand language, or automatically select suitable actions.
Major Applications of AI
- Healthcare: disease diagnosis, medical image analysis, drug discovery, and patient monitoring.
- Education: intelligent tutoring systems, personalized learning, automated evaluation, and educational assistants.
- Finance: fraud detection, credit risk analysis, algorithmic trading, and financial forecasting.
- Transportation: autonomous vehicles, traffic prediction, route optimization, and driver assistance systems.
- Natural Language Processing: chatbots, language translation, text summarization, and speech recognition.
- Computer Vision: facial/object recognition, medical image analysis, surveillance, and industrial inspection.
- Robotics: industrial robots, warehouse robots, service robots, and autonomous machines.
- Entertainment: game-playing agents, content recommendation, and computer-generated media.
- E-commerce: product recommendation, demand prediction, customer support, and personalized advertisements.
- Agriculture: crop monitoring, disease detection, yield prediction, and automated farming systems.
Summary
- AI has applications in healthcare, education, finance, transportation, robotics, agriculture, entertainment, and e-commerce.
- Common AI capabilities include prediction, pattern recognition, language understanding, perception, and automated decision making.
- The same fundamental AI techniques can be applied to many different real-world domains.
1.4 Intelligent Agents
What is an Agent?
An agent is an entity that perceives its environment through sensors and acts upon that environment through actuators.
A human agent uses eyes, ears, nose, and other sensory organs as sensors, while hands, legs, and speech can act as actuators. A robotic agent may use cameras and other sensors to observe its surroundings and motors to perform actions.
Intelligent Agent
An intelligent agent is an autonomous system that perceives its environment, processes the received information, and selects appropriate actions to achieve its objectives.
The basic interaction can be represented as:
Environment → Sensors → Agent → Actuators → Environment
The information received by an agent at a particular moment is called a percept. The complete history of everything the agent has perceived is called its percept sequence.
Structure of an Intelligent Agent
An intelligent agent can be represented as:
Agent = Architecture + Agent Program
- Architecture: the physical or computational platform on which the agent operates.
- Agent Program: the program that maps percepts or percept sequences into actions.
For example, in a robotic vacuum cleaner, the physical robot and its sensors form the architecture while the program deciding whether to move, turn, or clean is the agent program.
Properties of Intelligent Agents
- Autonomy: operates without continuous human control and uses its own experience or knowledge to make decisions.
- Reactivity: observes changes in the environment and responds appropriately.
- Pro-activeness: does not only react; it can take initiative to achieve its goals.
- Rationality: selects actions expected to produce the best result according to available information.
- Learning: improves behavior from experience or feedback.
- Social Ability: may communicate or cooperate with humans or other agents.
Let's Clarify
Think about a navigation application. It receives your location and road information as percepts. It calculates possible routes and recommends one. If traffic conditions change, it reacts and may select another route.
The system therefore follows the same general pattern as an intelligent agent: observe → decide → act → observe again.
Types of Intelligent Agents
AI agents can be classified according to how they choose their actions and how much information they use while making decisions.
1. Simple Reflex Agent
A simple reflex agent selects an action only on the basis of the current percept. It does not consider previous percepts or maintain an internal representation of the world.
It generally operates using condition-action rules:
IF condition → THEN action
For example:
IF room is dirty → THEN clean the room
IF room is clean → THEN move to another room
Simple reflex agents work effectively when the environment is simple and completely observable, but they perform poorly when important information cannot be obtained from the current percept alone.
2. Model-Based Reflex Agent
A model-based reflex agent maintains an internal state representing aspects of the environment that cannot currently be observed.
It uses both:
- the current percept, and
- knowledge about how the environment changes.
For example, a robot may remember the location of an obstacle even after that obstacle is temporarily outside its camera's field of view.
3. Goal-Based Agent
A goal-based agent selects actions according to whether they help achieve a specified goal.
Instead of simply reacting to the current situation, it may consider several possible future actions and select a path that leads toward the desired state.
For example, a navigation agent whose goal is to reach Pokhara may compare different roads before selecting a route.
4. Utility-Based Agent
A utility-based agent not only tries to reach a goal but evaluates how desirable different possible outcomes are using a utility function.
This becomes useful when:
- several solutions can achieve the same goal, or
- the agent must make trade-offs between factors such as cost, time, safety, or comfort.
For example, two routes may both reach Pokhara. A goal-based agent may accept either because both achieve the destination. A utility-based agent can compare travel time, road condition, fuel consumption, and safety and select the route with the highest utility.
Goal-Based vs. Utility-Based Agent
| Goal-Based Agent | Utility-Based Agent |
|---|---|
| Attempts to achieve a defined goal. | Attempts to achieve the most desirable outcome. |
| Primarily considers whether the goal is achieved. | Measures how good an outcome is. |
| May treat multiple goal-achieving solutions similarly. | Can rank several possible solutions. |
| Uses goal information. | Uses a utility function. |
| Example: selecting any route that reaches a destination. | Example: selecting the safest and fastest route among several alternatives. |
5. Learning Agent
A learning agent improves its performance through experience. Instead of relying entirely on knowledge given during development, it can learn from previous actions and feedback.
A learning agent commonly contains four components:
- Performance Element: selects the external actions performed by the agent.
- Learning Element: improves the performance element using experience.
- Critic: evaluates how well the agent is performing and provides feedback.
- Problem Generator: suggests exploratory actions that may lead to useful new experiences.
For example, an email spam filter can improve as it receives feedback about which messages are spam and which are legitimate.
Summary
- Simple reflex agent: acts only according to the current percept.
- Model-based agent: maintains an internal state of the environment.
- Goal-based agent: selects actions that help achieve a goal.
- Utility-based agent: selects the most desirable action according to utility.
- Learning agent: improves its performance using experience.
PEAS Description of an Agent
PEAS is a framework used to describe the task environment of an intelligent agent.
- P – Performance Measure: determines how the success of the agent is evaluated.
- E – Environment: represents the surroundings in which the agent operates.
- A – Actuators: mechanisms through which the agent performs actions.
- S – Sensors: mechanisms through which the agent obtains information about the environment.
PEAS for an English to Nepali Translating Agent
| PEAS Component | English to Nepali Translation Agent |
|---|---|
| Performance Measure | Translation accuracy, grammatical correctness, preservation of meaning, fluency, and response speed. |
| Environment | English text, Nepali language rules, vocabulary, documents, websites, and users. |
| Actuators | Display translated Nepali text, produce translated document, or generate speech output. |
| Sensors | Keyboard/text input, uploaded documents, microphone, or other input interfaces. |
Test Yourself
- What is an agent? List the different types of agents used in AI. (2025)
- Compare goal-based and utility-based agents with example. (2023)
- Explain the properties of an intelligent agent.
- How does a simple reflex agent work? (2021)
- Design a PEAS framework for an English to Nepali translating agent. (2021)
Hints:
- Agent = entity that perceives through sensors and acts through actuators. Types: simple reflex, model-based, goal-based, utility-based, and learning agents.
- Goal-based checks whether an action reaches the goal; utility-based additionally compares how desirable each outcome is.
- Explain autonomy, reactivity, pro-activeness, rationality, learning ability, and social ability.
- A simple reflex agent uses the current percept and condition-action rules without remembering percept history.
- PEAS = Performance Measure, Environment, Actuators, Sensors. Define suitable values for the translation task.
Types of Agent Environments
The behavior of an intelligent agent depends strongly on the characteristics of the environment in which it operates.
Deterministic vs. Stochastic
In a deterministic environment, the next state is completely determined by the current state and the action performed by the agent.
In a stochastic environment, the result of an action may involve uncertainty and cannot always be predicted exactly.
Example: A simple puzzle can be deterministic, while driving in real traffic is stochastic because the actions of other drivers cannot be predicted with complete certainty.
Static vs. Dynamic
A static environment does not change while the agent is deciding what action to perform.
A dynamic environment may continue changing while the agent is processing information or making a decision.
Example: A crossword puzzle is mostly static, whereas road traffic is dynamic.
Observable vs. Semi-Observable
In an observable environment, the agent's sensors can obtain the information required to understand the current state of the environment.
In a semi-observable environment, the agent receives only part of the information needed to completely determine the current state.
Example: A chess board is highly observable because the pieces are visible, while a self-driving vehicle has limited information about intentions of other drivers and conditions beyond its sensor range.
Single-Agent vs. Multi-Agent
A single-agent environment contains one primary agent performing the task.
A multi-agent environment contains multiple agents whose actions may cooperate with or compete against one another.
Example: Solving a crossword puzzle can be considered a single-agent task, while football or competitive chess involves multiple agents.
| Environment Type | Meaning | Example |
|---|---|---|
| Deterministic | Action produces a predictable result. | Puzzle |
| Stochastic | Outcome contains uncertainty. | Road traffic |
| Static | Environment does not change while deciding. | Crossword puzzle |
| Dynamic | Environment changes continuously. | Driving |
| Observable | Required state information is available to the agent. | Chess board |
| Semi-Observable | Only partial information about the state is available. | Autonomous driving |
| Single Agent | One primary agent performs the task. | Crossword solver |
| Multi Agent | Several agents interact. | Football game |
Extra: Topic Asked in Exams
Turing Test (2021 Exam)
The Turing Test, proposed by Alan Turing, is a test used to evaluate whether a machine can demonstrate human-like intelligent behavior during a conversation.
In the basic setup, a human evaluator communicates through a text-based interface with a human and a machine without seeing either participant. If the evaluator cannot reliably distinguish the machine from the human based on their responses, the machine is considered to have passed the test.
Capabilities Required to Pass the Turing Test
- Natural Language Processing: to communicate successfully using human language.
- Knowledge Representation: to store information learned or provided during interaction.
- Automated Reasoning: to answer questions and derive conclusions from available information.
- Machine Learning: to learn from experience and adapt to new situations.
Common Mistake
The Turing Test does not directly prove that a machine thinks or understands information exactly like a human. It evaluates whether its conversational behavior is sufficiently human-like that an evaluator cannot reliably distinguish it from a human.
Test Yourself
- What is the Turing Test? Mention the characteristic features that an agent should have to pass the Turing Test. (2021)
- Which approach to understanding AI is most closely associated with the Turing Test?
Hints:
- Explain the human evaluator, hidden human and machine, and then mention NLP, knowledge representation, automated reasoning, and machine learning.
- The Turing Test is primarily associated with the acting-humanly approach.
Exam Questions
- Explain Understanding AI based on thought process and behavior with example. [5] (2026)
- Define Artificial Intelligence. Explain its components. (2025)
- What is an agent? List the different types of agents used in AI. (2025)
- Compare Goal Based and Utility Based agent with example. (2023)
- What is Turing Test? Mention the characteristic features that an agent should have to pass Turing Test. (2021)
- How does a simple reflex agent work? Using your own assumptions, design PEAS framework for an English to Nepali Translating Agent. (2021)
Whole Chapter Summary
- Artificial Intelligence develops systems capable of performing tasks associated with human intelligence, including learning, reasoning, perception, problem solving, and decision making.
- AI can be understood through four approaches: thinking humanly, acting humanly, thinking rationally, and acting rationally.
- The foundations of AI include philosophy, mathematics, economics, neuroscience, psychology, computer engineering, control theory, and linguistics.
- AI is used in areas such as healthcare, education, transportation, finance, robotics, language processing, computer vision, agriculture, and e-commerce.
- An agent perceives its environment using sensors and affects the environment through actuators.
- The main agent types are simple reflex, model-based reflex, goal-based, utility-based, and learning agents.
- A simple reflex agent reacts only to the current percept, while a model-based agent maintains information about the environment.
- A goal-based agent attempts to achieve a target state, while a utility-based agent compares the desirability of different possible outcomes.
- PEAS stands for Performance Measure, Environment, Actuators, and Sensors and is used to specify an agent's task environment.
- Agent environments may be deterministic or stochastic, static or dynamic, observable or semi-observable, and single-agent or multi-agent.
- The Turing Test evaluates whether a machine can produce sufficiently human-like conversational behavior.