CAI1001C
AI Thinking
Introduction to Generative AI
Disclaimer
Miami Dade College has adapted the content on the subsequent slides from the
AI for Current Workforce program, which is part of Intel® Digital Readiness
Programs.
Topics
1. What is Generative AI?
2. What are the different types of Generative AI?
3. How Generative AI learns and works?​
4. How to apply Generative AI tools to create content​?
5. What are the ethical considerations of using Generative AI?
3
To get you started with
Generative AI, let's do a simple
activity.
Activity: Guess the Real Image
vs. AI Generated Image.
Activity: Guess the Real Image vs. AI Generated
Image
Examine the images and determine whether either of the images is
a real image or an AI-Generated image.
6
Activity: Guess the Real Image vs. AI Generated
Image
Examine the images and determine whether either of the images is
a real image or an AI-Generated image.
7
Activity: Guess the Real Image vs. AI Generated
Image
Examine the images and determine whether either of the images is
a real image or an AI-Generated image.
8
Activity: Guess the Real Image vs. AI Generated
Image
Examine the images and determine whether either of the images is
a real image or an AI-Generated image.
9
Activity: Guess the Real Image vs. AI Generated
Image
Examine the images and determine whether either of the images is
a real image or an AI-Generated image.
10
Activity: Guess the Real Image vs. AI Generated
Image
Examine the images and determine whether either of the images is
a real image or an AI-Generated image.
11
Activity: Guess the Real Image vs. AI Generated
Image
Examine the images and determine whether either of the images is
a real image or an AI-Generated image.
12
Activity: Guess the Real Image vs. AI Generated
Image
Examine the images and determine whether either of the images is
a real image or an AI-Generated image.
13
Now that you have had a chance to try
the Guess the Real Image vs. AI
Generated Image activity, let's look at
the concepts behind the generation of
these images.
Supervised Learning And Discriminative Modeling
The classification of data elements into categories or labels was initially
taught to the machine learning models by humans.
15
Unsupervised Learning And Generative Modeling
In unsupervised or self-supervised learning, the machine learning model
takes unlabeled datasets and figures out patterns and inherent
structures within them — without human intervention.
16
Now that we have a basic understanding
of the difference between Generative
Modeling and Discriminative Modeling,
let's take a closer look at Generative AI.
Introduction to Generative AI
What is Generative AI?
Exploring the Power of Artificial Creativity
▪ Generative artificial intelligence (AI) refers
to the algorithms that generate new data
that resembles human-generated content,
such as audio, code, images, text,
simulations, and videos.
▪ This technology is trained with existing
data and content, creating the potential for
applications such as natural language
processing, computer vision, the
metaverse, and speech synthesis.
Introduction to Generative AI
1. Definition of Generative AI​
2. Types of Generative AI (GANs, VAEs, etc.)​.
3. Examples of Generative AI (e.g., art, music, text, images, videos)​
4. Benefits of using Generative AI​
5. Limitations of using Generative AI
6. Hands-on activity: Exploring Generative AI with a case study
20
What is Generative AI
At The Very Core, The Primary Aim Of Generative AI Is To Make Things Easier!
Faster
Adoption
Increased
Productivity
Greater
Accuracy
Faster
Delivery
21
Generative AI vs Conventional AI
In contrast to other forms of AI, Generative AI is specially made to produce
new and unique content rather than merely processing or categorizing
already-existing data. Here are some significant variations:
Goal
Training
• Generative AI creates new content, whereas
conventional AI analyzes, processes, and classifies
data.
• Generative AI models use vast libraries of samples to
train neural networks and other complicated
structures to produce new content based on those
patterns.
• Conventional AI employs fewer complex algorithms
and training methods.
22
Generative AI vs Conventional AI
In contrast to other forms of AI, Generative AI is specially made to produce
new and unique content rather than merely processing or categorizing
already-existing data. Here are some significant variations:
Output
Applications
• Generative AI output is fresh, innovative, and often
unexpected.
• Conventional AI produces more predictable output
based on existing data.
• Generative AI benefits art, music, literature, gaming,
and design.
• Conventional AI is used in banking, healthcare, image
recognition, and language processing.
23
Brief history of Generative AI
Evolution Of AI And Its Impact On Creative Fields
In the early 2000s,
researchers began to
focus on developing
generative models like
Generative Adversarial
Networks (GANs) that
could produce highly
realistic images, text,
and music.
Generative AI can be
traced back to the mid20th century when early
pioneers Alan Turing and
John McCarthy laid the
foundation for the
development of intelligent
machines.
In the 1980s and
1990s, researchers
began exploring the
use of machine
learning techniques
and neural networks
to create more
sophisticated
generative models.
In recent years,
Generative AI has made
significant progress
and has been applied to
a wide range of creative
fields such as music,
art, and literature.
24
Brief history of Generative AI
Let’s Now Zoom In And Look At The Timeline For Various Fields:
2023
Copywriting gets better
(e.g. able to write
scientific paper,
professional email)
PRE-2020
Spam detection,
Translation, Basic Q&A
2020
Copy writing
Text Field Evolution Timeline
25
Brief history of Generative AI
Let’s Now Zoom In And Look At The Timeline For Various Fields:
2022
Art, Logos, Photography
generation
2023
Product design, architecture
Image Field Evolution
Timeline
26
Brief history of Generative AI
Let’s Now Zoom In And Look At The Timeline For Various Fields:
2022
Longer form,
Better
accuracy
PRE 2020
One Line auto
complete
2020
Multi Line
Generation
2023
More
languages
More Verticals
Code Field Evolution Timeline
27
Brief history of Generative AI
Let’s Now Zoom In And Look At The Timeline For Prominent Fields
28
Types of Generative AI
Generative AI comes in a variety of forms, each with unique advantages and uses. Some of the
most typical varieties are listed below:
GANs (Generative Adversarial Networks)
• GANs are neural networks that collaborate to produce fresh data
• Made up of two neural networks: Generator Network & Discriminator Network
• The generator network produces the data, while the discriminator network analyses
the data and provides feedback.
• Until the generator can generate data that is identical to real data, the two networks
collaborate in a feedback loop.
VAEs (Variational Auto encoders)
• Another class of generative models is VAEs. In order to produce fresh data, VAEs
learn the distribution of the data and then sample from it.
29
Types of Generative AI
RNNs (Recurrent Neural Networks)
• RNNs are a special class of neural networks that excel at handling sequential data, like music or text.
• They function by ingesting past inputs and applying that knowledge to anticipate future inputs.
Auto encoder
• These are Neural networks that have been trained to learn a compressed representation of data
• They function by compressing data first, then decompressing that compressed data to restore it to
its original form.
• Auto encoders can be applied to denoising or picture compression applications.
Deep Belief Networks
• Deep Belief Networks excel at simulating complex, high-dimensional data
• They operate by extracting a hierarchy of features from the data, beginning with simple ones and
progressing to more intricate ones.
30
Examples of Generative AI
Generative AI has many applications, from art and music to language and
natural language processing. Here are some examples of how generative AI
is being used in various fields:
▪
▪
Art: Generative AI is being used to
create unique works of art.
For example, The Next
Rembrandt project used data
analysis and 3D printing to create
a new painting in the style of
Rembrandt
31
Examples of Generative AI
Generative AI has many applications, from art and music to language and
natural language processing. Here are some examples of how generative AI
is being used in various fields:
▪ Music: Generative AI is being used
to create new music, either by
composing original pieces or by
remixing existing ones.
▪ For example, AIVA is an AI
composer that can create original
pieces of music in various genres.
32
Examples of Generative AI
Generative AI has many applications, from art and music to language and
natural language processing. Here are some examples of how generative AI
is being used in various fields:
▪ Language: Generative AI is
being used to generate new
language, such as chatbots
that can hold conversations
with users or natural
language generation systems
that can produce written
content.
33
Benefits of using Generative AI
Creativity:
• Generative AI can assist creatives in pushing the boundaries in
making creative processes more efficient and personalized. This
can be particularly valuable in fields such as art, design, and
music.
Efficiency:
• Generative AI can automate content creation processes, which
can save time and reduce costs compared to traditional manual
processes.
Personalization:
• Generative AI can be used to create personalized content for
individual users based on their preferences and behaviors, such
as customized product recommendations or personalized news
articles.
34
Benefits of using Generative AI
Exploration:
• Generative AI can be used to explore new design spaces or
optimize complex systems, such as designing new drugs or
improving industrial processes.
Accessibility:
• Generative AI can democratize access to content creation tools,
making it easier for people with limited resources or technical
expertise to produce high-quality content.
Scalability:
• Generative AI can be used to generate large volumes of content
quickly and efficiently, making it a scalable solution for
businesses and organizations that need to produce large
amounts of content.
35
Limitations Of Using Generative AI
01
Data Bias
▪ If generative AI is trained on biased or incomplete data, the output may be similarly
biased or flawed. This can lead to inaccurate or problematic results in certain
applications, such as in facial recognition or natural language processing.
02
Uncertainty
▪ Generative AI can produce unexpected and often unpredictable results, which can be
both a benefit and a drawback.
03
Computational Demands
▪ Generative AI requires significant computational resources to train and generate its
output, which can be expensive and time-consuming.
36
Limitations Of Using Generative AI
04
Intellectual Property Issues
▪ As generative AI becomes more sophisticated, it may raise issues around intellectual
property and ownership. If an AI generates a work that is like existing work, it may be
difficult to determine who owns the resulting work, leading to potential legal issues.
37
Let us now understand the benefits
and limitations of Generative AI with
the help of a case study
Benefits of using Generative AI: A Case Study
Business Challenges In Content Creation
9.5% of company’s
revenue budgeted on
marketing to earn
visibility of products
and services
of
which
80% directed toward
digital marketing
of
which
46% spent on content
creation including
strategic planning,
resource allocation,
writing etc.
Benefits of using Generative AI: A Case Study
Business Challenges In Content Creation
TYPES OF CONTENT
CREATED
Podcast
Courses/…
Blogs
Social
Media
Email/
Newsletter
s
Courses
100
Blogs (500
words)
40
Social
Media
0,5
Email/News
letter
0,25
Ads
0
20
40
60
80
Time taken in hours per piece
100
40
Benefits of using Generative AI: A Case Study
Business Challenges In Content Creation
Let us filter all the other types of content and focus on the case of
professional email content creation in details
▪ In a recent study by
Litmus, it was found that
people spent just nine
seconds, on average,
looking at an email.
41
Benefits of using Generative AI: A Case Study
Business Challenges In Content Creation
▪ In the same study by Litmus, it was
also found that 30% of emails, on
average, are looked at for less than
two seconds, 41% are looked at for
between two and eight seconds, and
only 29% are looked at for more than
eight seconds.
42
Benefits of using Generative AI: A Case Study
Business Challenges In Content Creation
This means to increase productivity,
creativity, and uniqueness in digital
marketing, we need to focus on:
Creating
Unique
Pieces That
Catches
The
Attention
Reducing
The
Content
Creation
Time
Generate
New Ideas
43
Benefits of using Generative AI: A Case Study
Business Challenges In Content Creation
Let us now understand how Generative AI is
going to increase productivity for this particular
use case:
▪ By using a Generative AI tool ChatGPT, emails
can be written in a much shorter span of time.
▪ Chat GPT is a language model, a type of
artificial intelligence trained on natural
language data to predict the probability of the
next word or sequence of words in a sentence.
44
Benefits of using Generative AI: A Case Study
Business Challenges In Content Creation
▪ The language model can be instructed to
produce a content specifically for target
audiences.
▪ In a matter of seconds, hundreds of different
email content can be generated, all of which
may then be scheduled for subsequent
distribution.
▪ Generative AI tools can also be used to read
messages and modify grammar.
45
Benefits of using Generative AI: A Case Study
Business Challenges In Content Creation
▪ ChatGPT can also be used as a quicker and
simpler version of a Google search and can
also be used to generate new ideas and
content that other people can’t get.
▪ ChatGPT can be used to respond to existing
customers. If a product receives some
negative customer feedback, ChatGPT can
be asked to craft a good response to it.
46
Benefits of using Generative AI: A Case Study
Some Of The Challenges And Limitations Of Using Generative AI
It Needs Guidance
▪ ChatGPT is an incredibly powerful tool, but it often needs to be finetuned before getting the correct output. There have been occasions
where ChatGPT does not give the correct answer unless it is given
specific context and prompts.
It’s Not Updating
• ChatGPT has only been pre-trained up to 2023, which means that it
does not have access to any information on more recent events. To
create relevant and up-to-date content, you’re still going to need a
human content creation team.
It Needs Structuring
And Editing
• The generated content would still need to be edited, fact-checked,
and structured by a human. The technology can provide accurate
information, but there have been scenarios where it has given
incorrect information.
● Barring few limitations, this is how Generative AI can increase
productivity not just with content creation but with many other fields as
well!
● Let us now go into more detail and understand how Generative AI
works!
48
How Generative AI Works?
How Generative AI works?
1. What Is Generative Adversarial Networks (GANs)​?
2. What Is Variational Auto Encoders (VAEs)​?
3. Understanding The Concept Of “Latent Space"​
4. How Generative AI Learns And Creates New Data​
5. Hands-on Activity: Create A Simple Generative AI Model Using Python
50
Generative Adversarial Networks (GANs)​
Two neural networks, called a generator network and a discriminator network,
make up a Generative Adversarial Network (GAN), a form of deep learning
model. The generator network is in charge of producing fresh data which is like
a set of training data, and the discriminator network is in charge of telling the
two apart.
Bringing Monalisa to life with GAN
51
Now that you have a basic
understanding of GANs, let's
try out a hands-on activity
using GAN Paint.
Hands-on Activity: GAN Paint
GAN Paint is an interactive tool that allows
you to create images using a generative
adversarial network (GAN). GANs are a type
of machine learning model that can be used
to generate realistic images from scratch.
▪ Scan the QR Code on the right to access GAN
Paint!
GAN Paint: http://gandissect.res.ibm.com/ganpaint.html
53
Hands-on Activity: GAN Paint
▪ GAN Paint directly activates and
deactivates neurons in a deep network
trained to create pictures.
▪ Each left button ("door", "brick", etc.)
represents 20 neurons. The software
shows that the network learns about trees,
doorways, and roofs by drawing.
▪ Switching neurons directly shows the
network's visual world model.
54
Hands-on Activity: GAN Paint
▪ To use GAN Paint, you will first need to
select a base image from the website's
library. You can then use the brush tool to
add objects and textures to the image. As
you paint, the GAN network will learn to
generate more realistic images.
▪ You are encouraged to experiment with
GAN Paint and see what you can create.
Have fun!
55
Now that you have seen how
GANs can be used to create
images, let's look at an
example of how GANs can be
used to swap clothes on
people.
Generative Adversarial Networks (GANs)​
Cloth Swapping With GANs
▪ The Generator creates realistic visuals
by swapping a person's clothing. The
Generator creates photos that seem
as if garments were smoothly
switched.
▪ The Discriminator distinguishes
between actual photographs of
individuals wearing authentic garments
and the Generator's swapped images.
The Discriminator functions as a
"fashion critic" to verify garment swaps
and spot irregularities in created
photos.
57
Generative Adversarial Networks (GANs)​
● The discriminator improves at
recognizing realistic data as the
generator improves. This
feedback loop produces highquality data that matches the
training data.
● Computer-generated yellow
squares. The results are
amazing! Especially when you
learn picture set b contains nonexistent persons.
58
Generative Adversarial Networks (GANs)​
● GANs have generated images, videos, music, and even 3D worlds. They
have demonstrated encouraging outcomes in numerous areas but are still a
study subject with many hurdles.
● Week 12 of AI for Computer Vision covers GANs and their applications.
Image to image translations
59
Variational Auto Encoders (VAEs)​
●
A variational encoder is a type of algorithm
that helps in the extraction of useful
information from massive amounts of data.
●
Imagine you have a collection of images of
cats and dogs, but you're not sure which ones
are of cats, and which are of dogs. The
variational encoder can assist you in solving
the mystery by examining the images and
seeing patterns particular to each species.
60
Variational Auto Encoders (VAEs)​
● Let’s assume VAE is a magician with a box that
transforms items.
● The magician wants to create new items like flowers.
● The magician takes multiple photos of the flower from
various angles and distances to create a vast dataset.
● He recalls essential flower properties to construct a
limited collection of variables like a VAE encodes data
into a low-dimensional latent space.
● Next, the magician utilizes the compressed variables
to create new flowers, as a VAE decodes latent space
data into a picture.
● He may include randomness and unpredictability to
make the generating process more distinctive.
61
Variational Auto Encoders (VAEs)​
● The basic idea of a VAE is to train a neural
network to map data from the input space to the
latent space (low dimensional representation of
input data) and then to use a decoder to map
the data back to the input space.
● However, unlike a standard autoencoder, the
encoder of a VAE maps the input data to a
probability distribution over the latent space
rather than a single point in the space. The
decoder then samples from this distribution to
generate a new point in the input space.
62
Variational Auto Encoders (VAEs)​
● VAEs have several applications,
including image and video
generation, data compression, and
anomaly detection.
● VAEs are effective in generating
high-quality
images
that
are
indistinguishable from real images.
● Can also be used to generate new
images that vary smoothly and
continuously in the latent space.
Face images generated with a Variational
Autoencoder
63
Concept Of "Latent Space"​
● The auto encoder initially encodes
images into a smaller set of integers that
reflect their key properties.
● Then, it utilizes these numbers to create
a new picture that is similar yet different.
● The latent space is essentially a hidden
algorithm layer that has all the
necessary information to recreate the
original pictures.
● The system can find essential
characteristics and patterns by learning
to represent pictures in this latent space.
64
Concept Of "Latent Space"​
● The latent space can be used for various
tasks, such as data compression,
anomaly detection, and data generation.
● For example, in data generation, the
decoder can generate new data by
sampling from the learned distribution in
the latent space.
● The smoothness and continuity of the
latent space make it possible to generate
data that varies smoothly and
continuously in the high-dimensional
space.
65
Hands-on Activity: Create A Simple Generative AI
Model Using Python
● For the hands on we will be
using Generative Adversarial
Networks (GANs) to generate
images that never existed
before.
● For this ,we will be using the
Stanford Dog Dataset.
● Go to the link and download
the dataset from this link.
● Give the location of the
directory in the notebook
where you will be saving the
pictures.
66
Using Generative AI​
Using Generative AI
1. Overview of popular Generative AI tools (e.g., RunwayML, GPT4, StyleGAN, etc.)​.
2. How to use Generative AI tools in real-world scenarios – coding,
data generation, content creation, analysis, etc.​
3. Hands-on activity: Students will use a Generative AI tool to carry out
real-world tasks.
68
Generative AI tools
Popular Python Generative AI Libraries
● TensorFlow: Open-source machine learning library
TensorFlow includes generative models. It builds and
trains autoencoders, GANs, and VAEs using a highlevel interface.
● PyTorch: Another popular open-source machine
learning framework, PyTorch offers a simple and
versatile interface for constructing and training
generative models.
● Keras: A Python-based neural networks API that runs
on TensorFlow, Theano, or CNTK. It lets you design
and train neural networks, including generative models,
easily.
Generative AI tools
There are many generative AI tools available today that enable users to create
and experiment with generative models. Here are some popular tools:
● Artbreeder : Artbreeder is a
web-based tool that enables
users to generate new images
by combining different GAN
models. Users can select and
combine different GAN models
to create new and unique
images.
70
Hands-on Activity: Generate Images With Text Prompt
Go to artbreeder.com.
Click on ‘Try it now’
Give cool text prompt and see how AI generates a
picture from those prompts.
71
Generative AI tools
There are many generative AI tools available today that enable users to create
and experiment with generative models. Here are some popular tools:
▪ Runway ML: Runway ML is a
platform for creating, training, and
deploying generative models. It
provides a user-friendly interface
for building and training various
types of generative models,
including GANs, VAEs, and image
classifiers.
72
Generative AI tools: RunwayML
RunwayML platform supports various machine learning frameworks, such as
TensorFlow, PyTorch, and Keras.
In addition to model creation and training, RunwayML also provides a range
of deployment options which makes it easy to share and use your models
with others.
RunwayML can be connected with other tools and platforms, such as Adobe
Creative Suite and Web browsers, which enables users to integrate their
generative models into different applications and workflows.
73
Hands-on Activity: Explore AI Magic Tools Of Runway
ML
Go to https://runwayml.com/.
Explore the AI Magic Tools
Take any tool of your choice and generate
new content with it.
74
Generative AI tools: Huggingface
● Hugging Face is a platform
that offers tools that let users
create, train, and use machine
learning (ML) models using
open-source technology.
75
Generative AI tools: Chat GPT
I asked Chat GPT to introduce itself. And here is the response!
76
Generative AI tools: Google Bard
I asked Bard to introduce itself. And here is the response!
77
Generative AI tools: ChatGPT & Google Bard
Access the revolutionary Generative AI tools by scanning the corresponding
QR code below!
ChatGPT:
https://chat.openai.com/
Google Bard:
https://bard.google.com/
78
Hands-on Activity: Chit-Chat with ChatGPT & Bard
● Sign up & Login into both
ChatGPT and Bard.
● Chat with the ChatGPT and ask
it to write a paragraph on How it
Works? - ChatGPT
● Similarly, Chat with Bard and
ask it to write a paragraph on
How it Works? - Bard
79
Let's see what ChatGPT and
Bard can do when we give
them some prompts.
Hands-on Activity: Chit-Chat with ChatGPT & Bard
Here are 6 prompts that can be tried on ChatGPT
and Bard:
1. Write a summary of the history of the internet.
2. Explain how to code a simple website.
3. Write a blog post about the latest trends in
artificial intelligence.
4. Create a presentation about the benefits of
cloud computing.
5. Write a research paper about the future of
technology.
6. Design an app that solves a real-world problem.
81
Hands-on Activity: Chit-Chat with ChatGPT & Bard
Document the findings from above activity on Chat GPT vs Bard based on the
parameters below:
●
●
●
●
●
●
●
Parameter 1: Human-Like Response.
Parameter 2: Training Dataset and Underlying Technology.
Parameter 3: Authenticity of Response.
Parameter 4: Access to the Internet.
Parameter 5: User Friendliness and Interface.
Parameter 6: Text Processing: Summarization, Paragraph Writing, Etc.
Parameter 7: Charges and Price.
82
Now that you have had a chance
to experiment with ChatGPT and
Bard, let's look at how these
Generative AI tools can be used
in real-world scenarios.
How To Use Generative AI Tools in Real-world Scenarios
Generative AI tools can be used in a variety of real-world scenarios, from
creating art and design to improving business operations. Here are some
examples of how to use generative AI tools in real-world scenarios:
Art and design
• Generative AI tools like Artbreeder can be used by artists and designers to create unique and
visually stunning images, designs, and patterns.
• For example, designers can use these tools to create custom clothing patterns or generate
new textures and materials for 3D modeling.
Content creation
• Generative AI tools can be used to automate content creation tasks such as writing product
descriptions or generating social media posts.
• Tools like GPT-3 can generate text that is indistinguishable from human-written text,
making them useful for generating large volumes of content in a short amount of time.
84
How To Use Generative AI Tools in Real-world Scenarios
Product design and optimization
• Generative AI tools can be used to create new product designs or optimize existing designs.
• For example, engineers can use these tools to generate new designs for complex machinery
or optimize existing designs for better performance and efficiency.
Business operations
• Generative AI tools can be used to optimize business operations such as inventory
management and supply chain optimization.
• For example, businesses can use these tools to predict demand for products, optimize
inventory levels, and predict future supply chain disruptions.
Personalization
• Generative AI tools can be used to create personalized experiences for customers.
• For example, e-commerce companies can use these tools to generate personalized product
recommendations based on a customer's browsing and purchase history.
85
How To Use Generative AI Tools in Real-world Scenarios
The table shows popular Generative AI tools that can be used in various fields.
86
How To Use Generative AI Tools in Real-world Scenarios
The table shows popular Generative AI tools that can be used in various fields.
87
Now that we have explored the
potential benefits of using
generative AI tools, it is important
to consider the ethical
implications of this technology.
Generative AI​ Ethics
Generative AI Ethics
1. The ethical considerations of using Generative AI​
2. The potential negative impact on society – (impact on jobs)​
3. The importance of responsible use of Generative AI​
4. How to avoid potential bias in Generative AI
90
Ethical considerations of using Generative AI​
While Generative AI offers many benefits, there are also several ethical
considerations that should be considered when using this technology.
Ownership
• There are questions about who owns the content generated
by generative AI. This is particularly relevant in creative fields
such as music, literature, or art, where generative AI can create
original works that blur the lines between human and machine
authorship.
Human Agency
• Generative AI raises questions about human agency and
control. As technology becomes more sophisticated, it may
become increasingly difficult to distinguish between content
generated by humans and machines, which could lead to a loss
of human autonomy and agency.
91
Ethical considerations of using Generative AI​
Bias
• Generative AI can replicate and amplify existing biases present
in the data used to train the model.
• This can lead to harmful or discriminatory outcomes, especially
if the generated content is used in high-stakes applications
such as hiring, loan approvals, or criminal justice.
Misinformation
• Generative AI can be used to create fake news or deepfakes,
which can be used to spread misinformation and manipulate
public opinion.
• This can have serious consequences for democracy and trust in
institutions.
Privacy
• Generative AI can potentially be used to generate sensitive
personal information, such as credit card numbers, social
security numbers, or medical records.
• This could be used for malicious purposes.
92
The Potential Negative Impact On Society
▪ Misinformation: Generative AI can be
used to create fake news or deep fakes
that can spread misinformation and
manipulate public opinion. This can lead
to a breakdown in trust in institutions and
individuals and can have serious
consequences for democracy.
93
The Potential Negative Impact On Society
● Job displacement: As generative AI becomes more advanced, it may be
able to automate certain types of creative work, such as writing, graphic
design, or music composition. This could lead to job displacement for
humans who previously performed these tasks.
94
The Potential Negative Impact On Society
Job displacement:
● Goldman Sachs economists believe
generative AI could replace up to onefourth of current jobs globally, or 300
million.
● Office and administrative work most at risk,
followed by jobs in law, architecture, and
engineering.
● OpenAI Research Says 80% of U.S.
Workers' Jobs Will Be Impacted by GPT.
● This study also found that around 19
percent of workers will see at least 50
percent of their tasks impacted.
95
The Potential Negative Impact On Society
Job Displacement:
▪ One silver lining is with so much
technological research in progress, new
jobs will also be created.
▪ The need for domain experts, technologists,
and developers is higher than ever.
▪ However, the risk is access to technology is
not equal.
▪ Hence if we can take care of equitable
access and try to bridge the digital divide,
then it’s not longer a risk and it’s a great
opportunity instead.
96
The Potential Negative Impact On Society
Here is the table of 9 jobs that Chat GPT-4 can potentially replace, along with the
human traits being replaced:
Past Job
Human Trait Replaced
Recruiter
Interviewing and Assessment
Translator
Language Proficiency
Call Centre Operator
Persuasion and Communication
Data Entry Clerk
Speed and Accuracy
Travel Agent
Planning and Coordination
Market Research Analyst
Analytical Skills
News Reporter
Fact-checking and Writing
Proofreader
Attention to detail
Copywriter
Creativity and Writing
97
The Potential Negative Impact On Society
▪ Privacy and security: Generative AI
has the potential to generate sensitive
personal information, such as social
security numbers or medical records,
which could be used for malicious
purposes.
▪ It could also be used to create
deepfakes of individuals that could be
used for blackmail or other malicious
purposes.
98
As the Generative AI revolution
continues to reshape our world,
ethical and responsible AI
development is more critical than
ever!
Responsible Use Of Generative AI
Addressing the Ethical Considerations
▪ Ensuring that the training data used are
diverse and representative.
▪ The outputs are scrutinized for bias and
misinformation.
▪ Prioritizing user privacy and consent.
▪ Having clear guidelines around ownership
and attribution of generative content.
▪ Engaging in public discussions around the
social and ethical implications of this
technology to ensure that it is developed and
used in ways that are beneficial to society.
The Importance Of Responsible Use Of Generative AI​
● Responsible use of generative AI is crucial for ensuring that this
technology is developed and used in ways that benefit society. Here are
some reasons why responsible use of generative AI is important:
Ethics:
• The development and use of generative AI raise important
ethical questions around bias, discrimination, privacy,
ownership, and human agency. Responsible use of
generative AI requires careful consideration of these
ethical questions and a commitment to addressing them
in a transparent and accountable manner.
101
The Importance Of Responsible Use Of Generative AI​
Impact:
Trust:
• Generative AI has the potential to have a significant impact
on society, both positive and negative. Responsible use of
generative AI can help maximize the positive impacts of the
technology while minimizing the negative ones.
• Trust is a critical component of any technology, and
generative AI is no exception. Responsible use of
generative AI can help build trust with users, customers,
and the general public by demonstrating that the
technology is being developed and used in a responsible
and ethical manner.
102
The Importance Of Responsible Use Of Generative AI​
Regulation:
▪ As generative AI becomes more widely used, there is a
growing need for regulatory oversight to ensure that the
technology is being used in ways that are safe, ethical, and
beneficial for society.
▪ Responsible use of generative AI can help shape these
regulations in a way that balances the interests of all
stakeholders.
Innovation:
▪ Responsible use of generative AI can help foster innovation
by encouraging developers to explore new use cases for the
technology that is both socially responsible and
commercially viable.
103
In short, responsible use of
Generative AI is essential for
ensuring that this technology is
developed and used in ways that
benefit society!
By emphasizing ethics,
creating trust, limiting
negative repercussions,
defining legislation, and
encouraging innovation, we
may maximize Generative
AI’s potential to improve
society!
105
How To Avoid Potential Bias In Generative AI
Here are some steps that can be taken to minimize bias in Generative AI:
• One of the main sources of
bias in generative AI is biased
training data.
• To avoid this, developers
should use diverse training
data that represents a range
of perspectives and
experiences.
• This can help ensure that the
model is not skewed towards
any group or viewpoint.
Diverse
Training
Data
• Regular testing of the
generative AI model can
help identify and correct any
biases that may be present.
• This can be done by
evaluating the outputs of
the model against a diverse
set of inputs and measuring
any discrepancies or
inconsistencies.
Regular
Testing
• The algorithms used in
generative AI should be
transparent and
explainable so that biases
can be identified and
corrected.
• This can help ensure that
the outputs of the model
are fair and unbiased.
Transparen
t
Algorithms
106
How To Avoid Potential Bias In Generative AI
Here are some steps that can be taken to minimize bias in Generative AI:
• Developers should consider the
ethical implications of their
generative AI models
throughout the development
process. This includes thinking
about potential biases and
taking steps to minimize them.
It also includes thinking about
the impact of the model on
society.
Ethical
Considerations
• Having external auditors review
the generative AI model can
help identify any potential
biases that may have been
missed during the development
process. This can provide an
independent and objective
evaluation of the model's
fairness and accuracy.
• Collaboration and engagement
with a diverse range of
stakeholders, including experts
in ethics, diversity, and social
impact, can help ensure that
generative AI models are
developed in a way that is fair
and unbiased.
External Audits
Collaboration
And
Engagement
107
In the realm of Generative AI,
Incorporating Responsible AI
Guardrails helps to balance between
innovation and accountability!
This Generative AI revolution have
alerted Governments worldwide to
produce a term called
Constitutional AI!
109
Constitutional AI
What is Constitutional AI?
•
•
Constitutional AI refers to the development
of AI systems that align with a country’s
constitution to ensure they operate within
legal and ethical frameworks, protecting
individual rights and promoting the public
good.
It involves creating transparent and
accountable AI systems subject to
appropriate legal and regulatory oversight.
Finally, by implementing the above-mentioned
strategies, you can fine-tune AI models effectively to
address misinformation, biases, and ethical concerns,
while maintaining their usefulness in various
applications.
111
Key Takeaways
1. Generative AI is a potent technology that enables machines to generate
plausible content and outputs.
2. Using patterns acquired from existing data, generative AI models, such
as GANs and VAEs, are utilized to generate new data.
3. GANs are a popular form of generative AI model that consists of a
generator and a discriminator that compete with one another.
4. Training data serves a crucial role in generative AI model efficiency.
5. Building a straightforward GAN is a hands-on exercise that entails
configuring a GAN with Python and a deep learning framework and
training it to generate samples from a dataset.
112
Key Takeaways
6. There are numerous industries in which generative AI is applicable,
including the arts, fashion, and entertainment.
7. Important ethical considerations and challenges are associated with
generative AI, such as the risk of bias and misconduct.
8. Continuous research and advancements are made in generative AI, and
there are numerous learning resources available.
9. Utilizing generative AI technology responsibly and ethically is crucial for
its continued development and positive impact.
113
Reflections
1. How do you think generative AI can revolutionize the creative industry,
such as art and fashion, by enabling the generation of unique and
innovative designs?
1. Considering the ethical challenges associated with generative AI, what
are your thoughts on establishing guidelines or regulations to ensure
responsible use of these technologies? How can we balance the
potential benefits and risks?
114
Bibliography
1. Generative AI, Explained by Humans. (n.d.).
https://lingarogroup.com/blog/generative-ai-explained-by-humans
2. A Beginner’s Guide to Generative AI. (n.d.). Pathmind.
https://wiki.pathmind.com/generative-adversarial-network-gan
3. Editor. (2022, October 13). Generative AI Models Explained. AltexSoft.
https://www.altexsoft.com/blog/generative-ai/
4. Team, T. A., & Team, T. A. (2021, November 4). Generative AI and GANs.
Towards AI. https://towardsai.net/p/l/generative-ai-gans
5. Goldman Sachs: Generative AI Could Replace 300 Million Jobs. (2023, April
12). AI Business. https://aibusiness.com/nlp/goldman-sachs-generative-aicould-replace-300-million-jobs
Bibliography
1. OpenAI Research Says 80% of U.S. Workers’ Jobs Will Be Impacted by
GPT. (n.d.). OpenAI Research Says 80% of U.S. Workers’ Jobs Will Be
Impacted by GPT. https://www.vice.com/en/article/g5ypy4/openai-researchsays-80-of-us-workers-will-have-jobs-impacted-by-gpt
2. Tiu, E. (2020, February 4). Understanding Latent Space in Machine
Learning. Medium. https://towardsdatascience.com/understanding-latentspace-in-machine-learning-de5a7c687d8d
Thank you for your contribution
today. I look forward to the next
module with all of you.