Unit 1:
What is artificial intelligence?
● Making computers that think?
● The Automation of activities we associate with human thinking, like decision-making, learning,
etc.
● The art of creating machines that perform functions that require intelligence when performed by
people?
● The study of mental faculties through the use of computational models?
Artificial Intelligence (AI)—Early History and Foundational Work:
First Work in AI:
The first work that is most generally recognized as a significant contribution to the field of artificial
intelligence (AI) was done by Warren McCulloch and Walter Pitts in 1943.
Sources of Inspiration:
McCulloch and Pitts drew upon the following three major sources:
1. Knowledge of basic physiology and the functioning of neurons in the brain
2. A formal analysis of propositional logic
3. Turing’s theory of computation
Key Findings:
● They proposed a model of artificial neurons.
● These neurons could be characterized as being "on" or "off," much like binary switches.
● They demonstrated that
○ Some network of connected neurons could compute any computable function.
● Their model showed that neural networks could simulate logical reasoning.
● They also suggested that suitably defined networks could be capable of learning.
Significance of Their Work:
● Their work laid the theoretical foundation for later developments in neural networks and machine
learning.
● It marked the first mathematical model that tried to simulate how the human brain might perform
logical operations.
● This was a pivotal moment in AI history, marking a significant milestone that bridged biology,
logic, and computation.
Acting Humanly: The Turing Test Approach
The Turing Test was proposed by Alan Turing in 1950 as a way to define and evaluate machine
intelligence.
Turing suggested that if a computer could engage in a conversation with a human—through written
text—such that the human cannot distinguish whether the responses are coming from a machine or a
human, then the computer can be said to exhibit intelligent behavior.
In simple terms, a machine passes the Turing Test if it can imitate human responses so well that a human
interrogator cannot reliably tell whether the answers are from a person or a computer.
The Birth of AI:
The term AI was coined in 1956 by a computer scientist, John McCarthy, at the Dartmouth conference
organized by him and three other scientists. He is often known as the “Father of AI.” In 1958, LSP was
developed. There were so many discoveries going on in the field of AI from the year 1952 to 1969.
AI winter (1974-1980):
In the 1970s, AI was subjected to criticism and financial setbacks. The expectation from AI was quite
high due to heavy optimism, but the researchers have failed to materialize the promised results.
A boom of AI (1980-1987):
● Year 1980: After the AI winter, AI came back with "Expert System." Expert systems were
programmed that emulate the decision-making ability of a human expert.
● In the year 1980, the first national conference of the American Association of Artificial
Intelligence was held at Stanford University.
Year 1997: In the year 1997, IBM Deep Blue beat world chess champion Gary Kasparov and became the
first computer to beat a world chess champion.
Year 2005: Stanley, the self-driving car developed by Stanford University, wins the DARPA Grand
Challenges.
Intelligence and AI:
Intelligence: Intelligence is the ability to acquire knowledge and use it, or we can say, “Ability to acquire
and apply knowledge and skills.”
We call ourselves Homo sapiens—man the wise—because our intelligence is so important to us. For four
thousand years, we have tried to understand how we think; that is, how a mere handful of matter can
perceive, understand, predict, and manipulate a world far larger and more complicated than itself. The
field of artificial intelligence, or AI, goes further still; it attempts not just to understand but also to build
intelligent entities.
Artificial: Produced by human art or effort, rather than originating naturally.
AI is defined as
● AI is the study of ideas that enable computers to be intelligent.
● AI is the part of computer science concerned with the design of computer systems that exhibit
human intelligence.
Sub-Fields: It includes various sub-fields like:
● Machine learning: where machines learn from experience.
● Natural Language Processing: which is about interaction with human languages.
● Computer vision: Where AI interprets and understands the visual form of data.
Applications of AI:
● Astronomy
● Healthcare
● Gaming
● Finance
● Data Security
● Social Media
● Travel & Transport
● Automotive Industry
● Robotics
● Entertainment
● Agriculture
● E-commerce
● Education
Goals of AI:
To make computers more useful by letting them take over dangerous or tedious tasks from
humans.
● Learning
● Reasoning
● Planning
● Perception
● Natural Language Processing
● Ability to manipulate/move/act in a given situation.
● Understand principles of human intelligence.
Problem-Solving in Artificial Intelligence
Problem-solving is a core capability of artificial intelligence (AI), where machines are designed to
perform tasks that normally require human intelligence and reasoning.
AI systems are capable of addressing complex, real-world problems by simulating human cognitive
processes such as analysis, reasoning, and decision-making.
Examples of AI in Problem-Solving:
● Speech Recognition:
Understanding and processing spoken language (e.g., voice assistants like Siri or Google
Assistant).
● Image Recognition:
Identifying and interpreting images (e.g., facial recognition, object detection).
● Decision Making:
Analyzing data to make informed decisions (e.g., fraud detection, medical diagnosis).
● Language Translation:
Converting one language into another using natural language processing (e.g., Google Translate).
Type of AI:
Artificial intelligence (AI) can be classified into four categories based on capabilities and functionalities:
Reactive Machines:
● These are the most basic type of AI systems.
● They do not store past experiences or learn from them.
● Reactive machines operate solely based on the current input and respond to situations
accordingly.
● Example: IBM’s Deep Blue, the chess-playing computer that defeated world champion Garry
Kasparov.
Limited Memory:
● These AI systems can use past data to make better decisions.
● They store limited information for a short duration and learn from historical data to improve over
time.
● Most current AI applications, such as chatbots, virtual assistants, and self-driving cars, fall under
this category.
Theory of Mind:
● This is a more advanced form of AI that is still under research.
● It would require the AI to understand human emotions, beliefs, intentions, and needs, and then
use this understanding to interact and make decisions.
● This type of AI aims to simulate human-like social intelligence.
Self-Aware AI:
● This is the most advanced and hypothetical form of AI, representing the future of artificial
intelligence.
● Machines with self-awareness would have consciousness, self-awareness, and emotional
understanding.
● Such systems would be able to form their own beliefs, desires, and goals.
● Currently, this type of AI is theoretical and does not yet exist.
Technologies Based On Artificial Intelligence:
Machine learning: A subfield of AI that uses algorithms to enable systems to learn from data and make
predictions or decisions without being explicitly programmed.
Natural language processing (NLP): A branch of AI that focuses on enabling computers to understand,
interpret, and generate human language.
Computer vision: A field of AI that deals with the processing and analysis of visual information using
computer algorithms.
Robotics: AI-powered robots and automation systems that can perform tasks in manufacturing,
healthcare, retail, and other industries.
Neural Networks: A type of machine learning algorithm modeled after the structure and function of the
human brain.
Expert System: AI systems that mimic the decision-making ability of a human expert in a specific field.
Chatbots: AI-powered virtual assistants that can interact with users through text-based or voice-based
interfaces.
Underlying Problems in Artificial Intelligence
Despite its rapid growth and potential, artificial intelligence (AI) still faces several challenges that limit its
widespread adoption and effectiveness. Some of the major underlying problems are:
1. Computing Power
AI algorithms, especially those used in machine learning and deep learning, require significant
computational resources. These algorithms are often power-hungry, needing advanced hardware such as
multi-core CPUs and high-performance GPUs.
● The high cost of computing infrastructure discourages many developers and researchers.
● As AI models become more complex, the demand for computing power continues to grow.
2. Trust Deficit
One of the biggest concerns in AI is the lack of transparency in how deep learning models make
decisions.
● The internal workings of many AI systems are often considered a "black box", making it difficult
to understand how certain inputs lead to specific outputs.
● This lack of interpretability creates a trust gap, especially among non-experts and
decision-makers.
● Explaining AI decisions to a layperson remains a major challenge.
3. Data Privacy and Security
AI systems rely heavily on large volumes of data to train models effectively.
● While massive datasets are available, much of this data comes from personal user interactions
collected globally.
● This raises concerns over privacy breaches and the misuse of sensitive data.
● Ensuring data is used ethically and securely is essential to gain user trust and comply with legal
standards.
4. Bias Problem
● AI systems rely heavily on data collected from various sources.
● Unfortunately, much of this data can be biased, reflecting the views or behaviors of a limited
group.
● These biases may arise from factors such as religion, ethnicity, gender, community, or culture.
● Biased data can lead to unfair or discriminatory decisions made by AI systems.
● Addressing this issue requires the development of algorithms that can detect and reduce bias
effectively.
5. Ethical Concerns
● AI systems, especially those involving machine learning and deep learning, require massive
computational power.
● This demand for high-performance hardware increases the cost and makes AI development less
accessible to many.
● Furthermore, ethical concerns arise when AI is used without transparency or accountability,
especially in decision-making systems.
6. Job Displacement
● One of the most discussed issues with AI is its ability to automate tasks that were traditionally
performed by humans.
● This may result in job loss in various industries.
● To adapt to these changes, there is a growing need for reskilling and upskilling the workforce to
stay relevant in an AI-driven economy.
Advantages and disadvantages of artificial intelligence:
Pros:
● More powerful and more useful computers.
● New and improved interfaces.
● Solving new problems.
● Better handling of information.
● Relieves information overload.
● Conversion of information into knowledge.
Cons:
● Increased costs.
● Difficulty in Software Development.
● Lack of Experienced Programmers.
● Limited Practical Products.
—----------------------------------------------------------------------------------------------------------------------------
Agents and Environment:
Agent: An agent is anything that can be viewed as perceiving its environment through sensors and acting
upon that environment through actuators.
Human Agent: Eyes, ears, and other organs for sensors; hands, legs, mouth, and other body parts for
actuators.
Robotic Agent: Camera and infrared range finders for sensors, various motors for actuators.
Agent Function: Maps any given percept sequence to an action.
[ F: P *→A]
Percept: This refers to the agent’s perceptual inputs at any given instant.
Percept Sequence: This is the complete history of everything the agent has ever perceived. An AI agent’s
decisions or actions at any moment can be influenced by the entire percept sequence it has experienced so
far. However, it can't base decisions on information it hasn’t perceived or sensed.
Agent Function: This is a rule or set of rules that determines what action an AI agent takes, based on its
percept sequences.
Agent Program: This is the practical, real-world implementation of the agent function.
Concept of Rationality in AI:
Rational Agent: A rational agent does the right thing, conceptually speaking; every entry in the table for
the agent function is filled out correctly.
Action Selection: Based on these precepts, a rational agent selects an action that’s expected to maximize
its performance measure. Using its percept history and innate knowledge.
Performance Measure: It defines the success of an agent’s behavior in a particular environment.
Designing Performance Measures: Ideally, these are created based on desired outcomes in the
environment, not preconceived ideas about the agent's behavior.
Understanding Task Environments and Rational Agents:
Task Environments: A task environment specification includes the performance measure, the external
environment, the actuators, and the sensors. In designing an agent, the first step must always be to specify
the task environment as fully as possible.
Task environments vary along several significant dimensions. They can be fully or partially observable,
single-agent or multi-agent, deterministic or stochastic, episodic or sequential, static or dynamic,
discrete or continuous, and known or unknown.
For example, the task environment of a chess game includes the rules of the game, the current state of the
board, and the goal of checkmating the opponent's king. The “problem” in this case is determining the
best move to make at each step to win the game. In the chess example, a rational agent would be an AI
system that can analyze the state of the game, consider all possible moves, and choose the one that
maximizes its chance of winning.
Rational Agent: A rational agent, then, is the “solution” to the problem: it's an AI system designed to
make the best possible decisions given the circumstances of the task environment.
PEAS:
Acronymically, we call it PEAS (Performance, Environment, Actuators, Sensors). Description of
PEAS for agent: Taxi Driver (Human Agent):
Agent Type Performance Environment Actuators Sensors
Measure
Taxi Driver Safe, Fast, Legal, Roads, other Steering, Camera, sonar,
Comfortable trip, traffic, accelerator, brake, speedometer, GPS
Maximize profits pedestrians, and signal, horn, odometer, engine
customers display sensors, keyboard
accelerator
Properties of task environments:
Fully Observable vs. Partially Observable:
Fully Observable: If an agent's sensors give it access to the complete state of the environment at each
point in time, then we say that the task environment is fully observable.
Partially Observable: Not all aspects are visible. The agent doesn’t have complete information about the
environment’s state because of noisy and inaccurate sensors or because parts of the state are simply
missing from the sensor data.
Unobservable: If the agent has no sensors at all, then the environment is unobservable.
Single Agent vs. Multi-Agent:
Single Agent: Only one agent is acting in the environment, like playing solitaire.
Multi-Agent: Multiple agents are interacting with the environment and potentially with each other, like
playing chess.
Deterministic vs. Stochastic:
Deterministic: The environment is deterministic. A task environment is said to be deterministic if the
current state and actions performed in the current state completely determine the next state.
Stochastic: The next state cannot be entirely predicted and might depend on probability, like poker.
Episodic/Sequential: This is a sequential environment. An environment is said to be episodic if it
consists of independent episodes and actions performed in one episode don't affect the other episodes.
In a sequential environment, the actions performed in the current state will affect the next state.
Static vs. Dynamic:
Static: The environment does not change while the agent is deciding, like in a game of tic-tac-toe.
Dynamic: The environment can change while the agent is making a decision, like in stock trading.
Types of Agents:
Four basic kinds of agent programs that embody the principle underlying almost all intelligent systems
are:
● Simple Reflex Agents.
● Model-Based Reflex Agents.
● Goal-Based Agents.
● Utility-Based Agents.
Simple Reflex Agent:
● They choose action only based on the current percept.
● Ignore the rest of the precept history
● Based on the condition-action (if-then”) rule.
● The task environment should be fully observable and limited intelligence.
Example: If the temperature is > 30°C, then switch on the AC
Model-Based Reflex Agents:
● Model-based reflex agents find a rule whose condition matches the current situation or percepts.
● Works on “how the world works”—internal model.
● Works in a partially observable environment.
● Consider the stored percept history.
Example: If the temperature is > 30°C && if the room is not empty, then switch on the AC.
Goal-Based Agents:
● Goal-based agents make decisions based on their current distance from the goal, and every
action the agent takes aims to reduce the distance from the goal.
● Extension of the model-based reflex agent.
● Incorporate searching and planning.
Utility-Based Agents:
● Utility-based agents are designed to make decisions that optimize their performance
by evaluating the preferences (or utilities) for each possible state.
● Focus on the utility, not the goal state.
● Deals with human emotions, e.g., happy/unhappy, satisfied/unsatisfied.
