Lecture#01

Artificial intelligence basic course by Sir IRFAN

MALIK
Here we are going to learn about the basics of artificial intelligence, through Sir
Irfan Malik and sir shiraz sahib.

What is artificial intelligence?

"The simulation of human intelligence in machines that are programmed to think, learn, and make
decisions."

Or

It’s actually the branch of computer science that deals with the methods that allow computers to
learn without explicit programming.

There are three types of artificial intelligence

ANI : ARTIFICIAL NARROW INTELLIGENCE(weak ai)

 What it is: AI designed to perform a single task or a narrow range of tasks very well.
It does not possess general intelligence or understanding outside its specific task.
 Capabilities: Limited to pre-defined tasks.
 Examples:

 Siri, Alexa (voice assistants).

 Spam filters in emails.
 Recommendation systems (e.g., Netflix or Amazon suggestions).

AGI: ARTIFICIAL GENERAL INTELLIGENCE (Strong AI)

 what it is: A more advanced form of AI that can perform any intellectual task that a human
can do. It has the ability to learn, understand, and apply knowledge across a wide range of tasks.
 Capabilities:

 Thinks, reasons, and learns like a human.

 Capable of generalizing knowledge to solve new, unseen problems.

 Examples:

 AGI does not exist yet, but it's the ultimate goal of AI research.
  Hypothetical example: A robot that can cook, clean, write poems, and design a rocket—
just like a human!

ASI: ARTIFICIAL SUPER INTELLIGENCE (Beyond Human Intelligence)

 What it is: AI that surpasses human intelligence in all areas, including creativity,
decision-making, and problem-solving. It would be capable of things humans cannot
even imagine.
 Capabilities:
o Superhuman in every way—knowledge, reasoning, speed, creativity, etc.
o Can design and improve itself, potentially leading to rapid advancements.
 Examples:
o ASI is theoretical and has not been developed yet.
o Depictions in science fiction, like the AI in movies such as Her or Ex Machina.

COMPARISON OF TYPES

Type Level of Intelligence Examples

ANI (Narrow AI) Focused on one task Siri, Google Maps, spam filters
AGI (General Human-like intelligence Hypothetical—future goal of AI research
AI)
ASI (Super AI) Superior to human intelligence Theoretical—depicted in sci-fi

How to do AI?

Most popular way to do AI is ML machine learning.

It’s a branch of AI which focuses on methods that can learn from examples and experiences instead of
relying on hard coded rules and make predictions on new data.

Machine Learning is a way of teaching computers to learn from data, just like how we humans learn
from experience.

Instead of programming every step, we let the computer figure out patterns and make decisions on its
own.

DIFFERENCE IN ML AND CLASSICAL AI

Classical AI Traditional algorithm answers

Rules and data
 MACHINE LEARNING
ANSWERS AND DATA MACHINE LEARNING RULES

What is algorithm?

An algorithm is set of step by step instructions designed to perform a specific task or solve a
particular problem.

In computer science, algorithms are implemented in programming languages to automate

tasks, from simple ones like sorting numbers to complex ones like training AI models.

TYPES OF MACHINE LEARNING

01: Supervised learning (learning with labeled data)

In supervised learning, the computer learns from labeled data.

Think of it like a teacher giving the computer examples with the correct answers and asking it to
learn the relationship between the input and the output.

Example:

 Daily Life: Imagine teaching a child to recognize apples and bananas. You show pictures
(inputs) and tell them whether it’s an apple or a banana (labels). Later, they can identify
fruits on their own (output).
 Tech Example: Predicting house prices based on features like location, size, and number
of rooms.

02: Unsupervised learning (discover patterns in unlabeled data)

In unsupervised learning, there are no labels. The computer has to find patterns or groupings in
the data on its own.
Example:

 Daily Life: Imagine a basket of mixed fruits. You don’t label them, but the child figures
out that fruits can be grouped by size, color, or shape.
 Tech Example:
o Grouping customers into clusters based on shopping habits (customer
segmentation).
o Organizing similar-looking images together.

03: Reinforcement learning (learn to act based on feedback/rewards)

In reinforcement learning, the computer learns by trial and error.

It gets rewards for good actions and penalties for bad actions, and it aims to maximize its reward
over time.

Example:

 Daily Life: Think of training a dog. If it sits when told, you give it a treat (reward). If it
doesn’t, no treat.
 Tech Example:
o Teaching a robot to walk or a self-driving car to navigate traffic.
o AI learning to play games like chess or video games by improving its strategy
over time.

04: semi supervised learning (labeled + unlabeled)

This is a mix of supervised and unsupervised learning. Some of the data is labeled, and the rest is
not. The computer uses the labeled data to help make sense of the unlabeled data.

Example:

 Daily Life: Imagine you label only a few fruits in the basket, and the child uses this
information to group the rest on their own.
 Tech Example: Detecting fake news articles when you have a small dataset of labeled
articles and a large set of unlabeled ones.

05: self supervised learning (a newer type)

This is a special type of learning where the machine generates labels from the data itself.

It's particularly common in large-scale AI models like Chat GPT.

Example:
  Daily Life: Think of a puzzle where you remove one piece and try to predict what it
looks like based on the rest of the picture.
 Tech Example: Models like GPT learning to predict the next word in a sentence by
looking at previous words.

Summarized table of ML types

Type Data Provided Example Use Case

Supervised Labeled data Predicting house prices, spam detection
Unsupervised Unlabeled data Customer segmentation, anomaly
detection
Reinforcement Rewards & penalties Teaching robots, self-driving cars
Semi- Mix of labeled & unlabeled Fake news detection
Supervised data
Self-Supervised Data labels itself Large AI models, image recognition

There are two types of supervised learning;

Regression and Classification…

1. Regression

Regression is used when the output we want to predict is a continuous value—a number that can
have an infinite range.

Example:

 Daily Life: Predicting the temperature for tomorrow based on historical weather data. The
output could be something like 25.5°C.
 Tech Example:
o Predicting house prices based on location, size, and age.
o Estimating stock market prices.

Key Point:

The goal is to find a line (or curve) that best fits the data and can predict future values.
2. Classification

Classification is used when the output is a category or a label. The task is to classify the input
into one of several predefined groups.

Example:

 Daily Life: Identifying whether an email is "spam" or "not spam."

 Tech Example:
o Diagnosing diseases as "positive" or "negative" based on medical test results.
o Classifying images as "cat" or "dog."

Key Point:

The goal is to assign a label to each input, often by drawing boundaries between classes in the
data.

Regression Classification
Predicts continuous values Predicts categories or labels
Example: House price = Example: Email = "spam" or "not spam"
$350,000
Output: Numbers Output: Labels or groups

23:01:25