AI -Basics of artificial intelligence lecture notes

ARTIFICIAL
INTELLIGENCE
TAUGHT BY:
PROF. DEEPIKA SHARMA

COURSE OUTCOMES
1. Describe the concept of Artificial Intelligence, Intelligent agents and Learning
agents
2. Identify issues in problem solving and apply the appropriate search methods
3. Describe and select the different knowledge representation methods
4. Implement AI planning techniques, decision-making models for real world
problems.
5. Use appropriate domain knowledge and develop an Expert system
6. Explore AI applications across various domains, ethical challenges, societal impact
and emerging trends shaping the future of AI

UNIT- I
Introduction to Intelligence and Problem Solving

The Imitation Game
cogito, ergo sum : 'I think therefore I am.'

What is AI exactly????
• Artificial Intelligence:
• -making machines think and reason like humans
• - It's Not Just Robots
• -solve problems, make decisions, understand language, recognize
patterns, and even create new things
• Weak AI targets a specific task where analyzing large amounts
of data assists in problem solving.
• Strong AI is defined by self-learning technologies that are
used to solve arbitrary tasks, at the level of human intelligence.
• Super AI surpasses intelligence of human in solving-problem, creativity,
and overall abilities.

AI enables systems to perform tasks that require human-like decision-
making, such as learning from data, identifying patterns, making
informed choices and solving complex problems

Typical AI Problems
• Problem of perception – perceptual tasks are
difficult as they require analog signals and require
large no. of things to perceived at once
• Problem of communication – to understand
sentences we need to know language as well as
the topic itself
• Problem of Simple – expertise tasks are easier to
perform rather than common sense tasks as they
require less knowledge which is easier to
represent

AI Technique : intelligence requires knowledge!
• A method that exploits knowledge that should be represented in such
a way that –
1. Knowledge captures generalizations
2. It can be understood by people who provide it
3. Easily modifiable to reflect real world changes
4. Not be built for one-two tasks only
5. Reduce amount of knowledge by narrowing the range of
possibilities

Let’s Play Tic-Tac-Toe…..
• Traditional way - Rule-Based Programming
• You, the programmer, explicitly write down
every possible move and a priority for those
moves.
• "If this, then do that. Else, if this other thing,
then do that.“
• The AI Programming Approach -computer
learns to play by itself
• The AI observes the current "state" (the board
configuration) and chooses an "action" (where
to place its mark).

Question – Answer Time
• https://start.csail.mit.edu/index.php

Intelligent Agents
• Intelligent agents represent a subset of AI systems demonstrating
intelligent behaviour, including adaptive learning, planning, and
problem-solving.
• It can make decisions, learn from experience, and adapt to changing
conditions.
• An ideal rational agent is one that always chooses actions that
maximize its performance measure, considering its current percepts,
percept sequence, and built-in knowledge.
• Give some examples????

Agents and Environment
• An agent in AI is an entity that can perceive its environment through sensors
and act upon it using actuators to achieve specific goals.
• A human agent has sensory organs such as eyes, ears, nose, tongue and skin
parallel to the sensors, and other organs such as hands, legs, mouth, for
effectors.
• A robotic agent replaces cameras and infrared range finders for the sensors,
and various motors and actuators for effectors.
• A software agent has encoded bit strings as its programs and actions.
• The environment is the external context or space in which an agent
operates. It provides the stimuli the agent perceives and reacts to.

• Performance Measure of Agent − It is the
criteria, which determines how successful an
agent is.
• Behavior of Agent − It is the action that agent
performs after any given sequence of percepts.
• Percept − It is agents perceptual inputs at a
given instance.
• Percept Sequence − It is the history of all that
an agent has perceived till date.
• Agent Function − It is a map from the precept
sequence to an action.

Structure of Intelligent Agents
1. Simple Reflex Agents
• They choose actions only based
on the current percept.
• They are rational only if a
correct decision is made only
on the basis of current precept.
• Their environment is
completely observable.
• Condition-Action Rule − It is a
rule that maps a state
(condition) to an action.

Example
Automatic Doors:
• Automatic doors open when a sensor detects someone approaching
and remain closed otherwise. They react to the presence or absence
of a person without remembering past interactions or predicting
future traffic.
Basic Traffic Light Systems:
• In a simple traffic light system, lights change based on a timer or
sensor input. The system doesn't remember past traffic patterns or
try to anticipate future congestion.

2. Model Based Reflex Agents
• They use a model of the world to
choose their actions. They maintain
an internal state.
• Model − knowledge about how the
things happen in the world.
• Internal State − It is a representation
of unobserved aspects of current
state depending on percept history.
• Updating the state requires the
information about −
• How the world evolves.
• How the agents actions affect the world.

Example
Navigation Systems (GPS):
• While not fully autonomous, GPS
systems use a model of road
networks and traffic patterns to
provide the best route to a
destination. This involves
predicting travel times and
suggesting alternative routes
based on real-time traffic
conditions.
Game AI:
• In some video games, AI
opponents might use a model of
the player's behavior to
anticipate their actions and
respond strategically. For
example, they might predict the
player's movement and position
themselves to intercept or avoid
the player.

3. Goal Based Agents
• They choose their actions
in order to achieve goals.
• Goal-based approach is
more flexible than reflex
agent since the knowledge
supporting a decision is
explicitly modeled, thereby
allowing for modifications.
• Goal − It is the description
of desirable situations.

Example
Self-Driving Cars:
• Goal: To reach a specific destination.
• Actions: The car analyzes its environment (other vehicles,
pedestrians, traffic signals, road conditions), plans a route, and
controls the steering, acceleration, and braking to navigate safely and
efficiently.
• Decision Making: The agent considers various factors like traffic laws,
road hazards, and the desired route to determine the optimal actions
at each point.

4. Utility Based Agents
• They choose actions based
on a preference (utility) for
each state.
• Goals are inadequate when −
• there are conflicting goals,
out of which only few can be
achieved.
• Goals have some uncertainty
of being achieved and you
need to weigh likelihood of
success against the
importance of a goal.

Example
Recommendation Systems:
• Goal: Provide users with relevant and engaging content or products.
• Utility Function: The agent considers factors like:
• User Satisfaction: Predicting what the user will enjoy based on their past
behavior, preferences, and ratings.
• Engagement: Encouraging the user to spend more time on the platform.
• Revenue: For e-commerce, maximizing sales by recommending specific
products.
• Decision Making: The agent suggests items or content that maximize
the overall utility score, balancing user preferences with business
goals.

5. Learning Agents
• New experiences are added to
their initial knowledge base, which
occurs autonomously.
• This learning enhances the agent’s
ability to operate in unfamiliar
environments.
• Learning agents may be utility or
goal-based in their reasoning
• Main elements of learning agents:
1. Learning element: It is
responsible for making
improvements by learning from
the environment.
2. Critic: The learning element takes
feedback from critics which
describes how well the agent is
doing with respect to a fixed
performance standard.
3. Performance element: It is
responsible for selecting external
action.
4. Problem Generator: This
component is responsible for
suggesting actions that will lead to
new and informative experiences.

Example
Chatbots and Virtual Assistants:
• AI-powered chatbots and virtual
assistants like Siri and Alexa
employ learning agents to
enhance their conversational
abilities.
• These agents learn from user
interactions, refining their
responses and improving their
understanding of natural language.
Fraud Detection Systems:
• Fraud detection systems use
learning agents to analyze
transaction data and identify
fraudulent patterns.
• These agents continuously adapt
to new fraud tactics, improving
their accuracy in detecting
fraudulent activities.

Nature of Environments
• Fully Observable vs. Partially Observable:
In a fully observable environment, the
agent has access to all relevant information to
make decisions. Example: Chess, where all
pieces and their positions are visible. In a
partially observable environment, only limited
or incomplete information is available. Example:
Driving in fog, where visibility is restricted.
• Deterministic vs. Stochastic:
A deterministic environment ensures that
the outcome of an action is predictable based
on the agent’s input. Example: Tic-tac-toe,
where actions lead to consistent results. In
stochastic environments, outcomes are
uncertain and influenced by random factors.
Example: Poker, where the opponent’s hidden
cards introduce unpredictability.
• Static vs. Dynamic:
Static environments remain unchanged
during the agent’s decision-making process.
Example: Crossword puzzles, where clues and
solutions remain constant. Dynamic
environments evolve over time, requiring the
agent to adapt. Example: Traffic systems,
where the agent must respond to changing
traffic conditions.
• Discrete vs. Continuous:
Discrete environments have distinct,
countable states or actions. Example: Turn-
based board games like checkers. Continuous
environments involve a range of possible
states or actions. Example: Autonomous drone
navigation, where movements are fluid and
ongoing.

The one and only Turing Test
• Two persons and a machine to be
evaluated participate in the test. Out
of the two persons, one plays the role
of the tester. Each of them sits in
different rooms. The tester is unaware
of who is machine and who is a
human. He interrogates the questions
by typing and sending them to both
intelligences, to which he receives
typed responses.
• This test aims at fooling the tester. If
the tester fails to determine machines
response from the human response,
then the machine is said to be
intelligent.
• Has Anything Ever Passed the Turing
Test?
• While no machine has ever passed the
Turing Test flawlessly, several
machines have fooled judges to some
extent.
• In 1966, MIT professor Joseph
Weizenbaum created a machine
named ELIZA, which is regarded as
one of the first computers to have
fooled a judge.
• The chatbots Eugene Goostman and
GPT-4 also passed the Turing Test in
2014 and 2024, respectively.

Here are some questions to ask
• What is your most memorable childhood event and how has that
impacted you today?
• Describe yourself using only colors and shapes.
• Describe why time flies like an arrow but fruit flies like a banana?
• How do you feel when you think about your upbringing and what
makes you feel that way?
• What historical event changed you the most and where were you
when it happened?
• Which of the previous questions was the most difficult to answer and
why?

Limitation of Turing Test
• Turing’s test is not a sufficient
indicator of artificial intelligence.
• Not only does Turing’s test fail to
account for whether or not a
machine understands its input
and output,
• it also accounts for neither a
machine’s ability to recognize
patterns nor its ability to apply
common knowledge or sense.

MACHINE LEARNING (ML)
• It is a field of study in artificial intelligence concerned with the
development and study of statistical algorithms that can learn from
data and generalize to unseen data, and thus perform tasks without
explicit instructions.
The process includes:
• Data Preprocessing: Cleaning and preparing data for analysis.
• Feature Selection: Identifying the most relevant data attributes.
• Model Training: Applying algorithms to datasets to find patterns.
• Model Testing: Evaluating accuracy using separate test data.

DEEP LEARNING (DL)
• Deep Learning is a subfield of
machine learning that is
concerned with algorithms
inspired by the brain’s
structure and functions
known as artificial neural
networks.

Data Science (DS)
• Data science is the science of analyzing raw data using statistics and
machine learning techniques with the purpose of drawing conclusions
about that information..
• Data Science is about finding patterns in data, through analysis, and
make future predictions.

Aspect Artificial Intelligence (AI) Machine Learning (ML) Deep Learning (DL) Data Science (DS)
Definition
The broad concept of creating machines
that can mimic human intelligence to
perform tasks like reasoning, learning,
problem-solving, and perception.
A subset of AI that enables systems to
learn from data and improve without
being explicitly programmed.
A specialized subset of ML that
uses multi-layered neural
networks (inspired by the human
brain) to learn from vast amounts
of data.
An interdisciplinary field that uses
scientific methods, processes,
algorithms, and systems to extract
knowledge and insights from data in
various forms (structured,
unstructured, noisy).
Scope
The overarching field; the "big dream" of
intelligent machines.
A method to achieve AI; focuses on
learning from data.
A powerful technique within ML;
excels at complex pattern
recognition.
Broader than AI in terms of data
usage; focuses on the entire data
lifecycle to extract insights.
Learning Method
Logic, rules, reasoning, knowledge
representation.
Learns from data to identify patterns
and make predictions (e.g., supervised,
unsupervised, reinforcement learning).
Learns hierarchical
representations of data through
multiple layers of artificial neural
networks.
Uses various analytical and statistical
methods, including ML, to analyze
data and derive actionable insights.
Complexity
Varies widely, from simple rule-based
systems to highly advanced learning.
Can range from simple algorithms to
complex models.
Generally involves highly complex
models with many layers,
requiring significant computational
power.
Deals with the entire complexity of
data, from collection and cleaning to
analysis and communication of
results.
Hardware Needs Low to moderate. Moderate.
High (often requires GPUs for
training large models).
Moderate to high, depending on the
scale and complexity of data being
processed.

Aspect Artificial Intelligence (AI) Machine Learning (ML) Deep Learning (DL) Data Science (DS)
Examples
Self-driving cars, Siri/Alexa,
chess engines, robotics.
Email spam filters,
recommendation engines
(Netflix, Amazon), predicting
house prices.
Facial recognition, natural
language processing
(advanced translation,
chatbots), image generation.
Business intelligence
dashboards, predictive
analytics for market trends,
customer behavior analysis,
fraud detection systems.
Relationship AI is the umbrella term. ML is a subset of AI. DL is a subset of ML.
Data Science uses AI/ML/DL
techniques as tools to
achieve its goals.

CASE STUDIES TO DO-
•AI for everyone
•Building AI projects
•AI for Society
•Ethics for AI
• The Problem Statement
• What AI technology is used?
• What is the impact?
• Measurable results
• Qualitative results
• The downside of AI
• Summary

Fun Weekend Activity
The Problem: Get the LLM to explain the concept of "recurrent neural
networks" in two distinct ways:
• Version A (Academic & Formal): The explanation should sound like
it's coming from a university professor delivering a lecture, using
appropriate technical jargon and a structured, formal tone.
• Version B (Casual & Analogous): The explanation should sound like a
friendly peer or tutor, using simple language, relatable analogies, and
a conversational, encouraging tone.

The Problem: Provide students with a moderately complex, factual
paragraph (e.g., a short case study on a recent tech innovation or a
brief description of a new programming paradigm). Their task is to
prompt the LLM to:
• Summarize the provided paragraph.
• Crucially, the summary must ONLY use information explicitly stated
in the given text. No outside knowledge or inferences allowed.
• The summary must be between 50 and 70 words.

The Problem: Prompt the LLM to write a Python function that
calculates the nth Fibonacci number, but with the following specific
requirements:
• It must use an iterative approach (no recursion).
• It must include comprehensive docstrings explaining its purpose,
arguments, and return value.
• It must include error handling for non-integer or negative inputs.

AI -Basics of artificial intelligence lecture notes

More Related Content

Similar to AI -Basics of artificial intelligence lecture notes

Recently uploaded

AI -Basics of artificial intelligence lecture notes