Introduction to

Generative AI

Basics of Generative AI
"Introduction to Generative AI: Understanding
and Exploring its Foundations and Applications"
So let’s start with the Introduction to Generative AI. If you're unsure
what that means, you're in the perfect spot! We're going to cover four
main points:
First, what exactly is generative AI?
On high level - It's a type of artificial intelligence that can create
different kinds of content, such as text, images, audio, and synthetic
data.
So, what is AI in general?
Since we’ll be exploring generative AI, it’s good to have
some context. AI is a field within computer science, much
like physics or chemistry is a field within science. AI
involves creating intelligent systems that can think, learn,
and act independently.
I hope you are with me so Far?
Basically, AI is about developing machines that can think
and behave like humans.
Simple enough, right?
Now, let's talk about machine
learning
• Machine learning is a part of AI. It's a system that learns from input data to create a model. This
model can then make predictions based on new data, even if it has never seen that specific data
before.
This means machine learning lets computers learn without being explicitly programmed for every
task.
• It is a program or system that trains a model from input data the trained model can make useful
predictions from new never-before seen data drawn from the same one used to train the model
this means that machine learning gives the computer the ability to learn without explicit

• There are two main types of machine learning models:

Supervised and unsupervised. The main difference is that supervised
models use labeled data, which means the data comes with tags
like names, types, or numbers. Unsupervised models use unlabeled
data, which has no tags. The key difference between these models
lies in the nature of the input data.
So, what can you do with supervised
learning?
• Supervised Learning: In supervised learning, the input data comes with labels. These labels are
tags such as names, types, or numbers that provide additional information about the data.
Supervised learning algorithms use these labels to learn and make accurate predictions or
classifications on new, unlabeled data.

Example: Email spam detection is a common example of supervised learning. Here, the model is
trained on a dataset containing emails labeled as "spam" or "not spam." The model learns to
identify patterns and features that differentiate spam from non-spam emails. Once trained, it can
classify new incoming emails as either spam or not spam based on the patterns it has learned.

One of many Examples: In medical diagnostics, supervised learning can be used to predict
diseases. For instance, a model can be trained on patient data labeled with whether or not they
have a certain disease. The trained model can then predict the likelihood of new patients having
that disease based on their medical data.
And what can you do with unsupervised learning?

• Unsupervised Learning: In contrast, unsupervised learning works with unlabeled

data, which lacks these tags. Unsupervised learning algorithms try to identify
patterns or structures in the data without prior knowledge of the outcomes.
• Example: Customer segmentation is a typical use case for unsupervised learning. Businesses
can use unsupervised learning algorithms to analyze purchasing data and identify distinct
customer groups based on buying behavior. This helps in tailoring marketing strategies to
different customer segments.
• Example: In image recognition, unsupervised learning can be used for clustering similar
images. For instance, a large dataset of images can be analyzed to group similar images
together without any prior labels, which can be useful in organizing large collections of
photos.
• By using these machine learning models, we can achieve various tasks such as
classification, regression, clustering, and dimensionality reduction, depending on
the problem and data at hand.
 Supervised v/s Unsupervised
workflow
So to put this in a workflow
In supervised learning testing data
values X our input into the model the
model outputs a prediction and Compares
it to the training data used to train
the model if the predicted test data
values and actual training data values
are far apart that is called error the
model tries to reduce this error until
the predicted and actual values are
closer together this is a classic
optimization
Deep Learning
• Remember this image Let’s look at other one

• Let’s understand Deep Learning before we Dive in Gen AI more

What is Deep Learning
• Decades ago, when AI started to see a lot of popularity, there was much debate among
scientists and researchers about making a machine that would have the ability to learn,
adapt, and make decisions. The general idea was to have machines replicate the way the
human brain learns.
https://aws.amazon.com/what-is/deep-learning/
• The brain is primarily made up of a collection of neurons, all interconnected and firing
electric signals back and forth to help the mind interpret things, reason, make decisions,
etc. AI researchers at the time sought inspiration from this and tried to mimic the human
brain’s function by creating artificial neurons.
• Artificial neural networks draw inspiration from the human brain. Similar to our cognitive
organ, they comprise numerous interconnected nodes, or neurons. These neurons are
adept at learning tasks by analyzing data and making predictions. Deep learning models
typically boast multiple layers of neurons, allowing them to grasp complex patterns more
effectively than their traditional counterparts.
• Neural networks can operate with both labeled and unlabeled data, a technique referred to as semi-
supervised learning. In this method, the network is trained on a combination of labeled data,
providing foundational insights, and unlabeled data, aiding in broader understanding and
adaptation. The labeled data helps the neural network to learn the basic concepts of the tasks while
the unlabeled data helps the neural network to generalize to new examples
Now Let’s get in to Gen-AI
• Gen AI is subset of deep learning
• Which means it uses artificial
neural networks can process both labeled
and unlabeled data using supervised
unsupervised and semi-supervised
methods large language models are also a
subset of deep learning.
• Deep learning models or machine learning
models in general can be divided into
two types
• Generative and discriminative model
Discriminative Models
• A discriminative model is
a type of model that is used to classify
or predict labels for data points
discriminative models are typically
trained on the data set of labeled data
points and they learn the relationship
between the features of the data points
and the labels once a discriminative model is
trained it can be used to predict the
label for new data points a generative
 Generative Model
A generative model generates new
data instances based on a learned
probability distribution of existing
data generative models generate new
To summarize Discriminative GEN-AI, discriminates
after learning from exiting data, on other hand
Generative AI (as the name suggests) Generate’s
content’s including images.
Can you name a few examples from real world. Please
take that as a task and give some examples in sections
for the trainings and share with other students.
Technically:
To summarize generative models can
generate new data instances and
discriminative models discriminate
between different kinds of data
instances one more quick example the top
 Let’s play a Game!!!
What if someone dares you to discern
between generative and non-generative
content? Fear not, I've got you covered.
Let me help you here.
This illustration provides a handy method to
differentiate between the two.
Generative content isn't classified as such
when the output, also known as Y or label, is a
numerical value or falls into a specific
category, such as "spam" or "not spam," or
even a probability. On the other hand, content
is considered generative when the output
takes the form of natural language, such as
speech or text, audio files, or images
AI Journey
• The generative AI process can take training code labeled data and unlabeled data of all data types and build a
foundation model the foundation model can then generate new content it can generate text code images audio video
and more.
• We've come a long way from traditional programming to neural networks to generative models
• In traditional programming we used to have to hardcode the rules for example “to distinguishing a cat with paramters
like 1. type animal 2. legs four 3. ears two 4. fur yes etc etc.
• In the wave of neural networks we could give the networks pictures of cats and dogs and ask is this a cat and it would
predict a cat or not a cat.
• Now in the generative wave, we as users can generate our own content whether it be text images audio video or more
for example models like Palm or Pathways language model or Lambda language model for dialogue applications inset
very very large data from multiple sources across the internet and build Foundation language models we can use
simply by asking a question whether typing it into a prompt or verbally talking into the prompt itself so when you ask
it what's a cat it can give you everything it's learned
about a cat (through training)
GEN-AI definition
• In-short: GEN-AI is a type of artificial intelligence that creates new
content based on what it has learned from existing content
• Training: The process of learning from existing content is called training
• When given a prompt gen uses a statistical model to predict what an
expected response might be and this generates new content
• It learns the underlying structure of the data and can then generate
new samples that are similar to the data it was trained on
• A generative language model can take what has learned from the
examples it's been shown and create something entirely new based
on that information that's why we use the word generative
Type of Generative Models
A generative image model takes an
image as input and can output text
another image or video for example under
the output text you can get visual question and
answering while under output image an image
completion is generated and under output video
animation is
generated

Same way a generative language model

takes text as input and can output more
text an image audio or decisions for
example under the output text question
and answering is generated and under
output image a video is
generated
 Some Random Terms and Definition
The power of generative AI comes from the use of
Transformers.
Transformers produced the revolution in natural language processing
At a high level a Transformer model consists of an encoder and a decoder
The encoder encodes the input sequence and passes it to the decoder which
learns how to decode the representations for a relevant task

We will discuss all these terms in details in other modules of this training so
please hang-on there!!!
 Foundation Model
A foundation model which is a large AI model pre-trained on a vast quantity of data designed to be adapted or fine-tuned to a wide range of
Downstream tasks such as sentiment analysis image captioning and object recognition Foundation models have the potential to revolutionize many
Industries including Healthcare finance and customer service they can even be used to detect fraud and provide personalized customer
Closure
• AI have been here since long but with recent advancement in GPU
and hardware capabilities the imagination is taking shape
• With powerful chips and semiconductors and processors, many
companies have developed there foundational models and have
made it available for organizations to explore and use, instead of
developing there own FM’s