Introduction to Generative AI

Generative AI is an advanced form of AI that creates new content in the form of images,
text, and audio based on the data it has learnt from. Unlike traditional AI, which usually
classifies or predicts things, generative AI focuses on making new data that looks similar
to what it has been trained on. Let us understand the concept behind generative AI.

Concepts Behind Generative AI

Supervised and unsupervised learning are the key concepts behind understanding
generative AI.
Generative models are trained using these fundamental techniques.

Supervised Learning and Discriminative Modelling

In supervised learning, the learning of the machine (model) is done with the help of
labelled data, i.e., some data is already tagged with the correct answer. Consider the
following illustrations, where each image has a label.
Star
Rectangle
dog
Triangle dog

cat cat

Circle
Square
Labelled Datasets
Initially, the labelled data is used to teach machines (models), or you can say that
models are taught by humans with the help of data, and after learning, models help us
in identifying or generating new content. Let us understand it with the help of some
examples.
Example 1:
Suppose, you have trained a machine (model) with the help of a machine learning
algorithm and images of animals, with each image labelled to indicate the animal it
depicts: 'Elephant', 'Camel', and 'Cow'. This means the model was trained using the
supervised machine learning method.
After training, this model will be able to identify animals in new images that you provide
as input. Notice the given figure, where the provided input image allows the model to
correctly identify the 'Elephant', 'Camel', and 'Cow'.
Kips | Artificial Intelligence 179
 Input Raw Data Supervisor

Labelled Data

Training dataset Desired Output

Ouput
Processing
Elephant

Labels Camel

Cow
Elephant
Cow Camel Algorithm

Example 2:
Now, consider another example, where you trained a machine (model) with the help of
a machine learning algorithm and images of bicycles. At the time of training, all images
of bicycles were labelled, i.e., a supervised machine learning method was followed to
train the model. So, the model will be able to identify bicycles from the input you are
providing it. Notice the following illustration, where the model for the provided input
image can recognise bicycles.
Input Output
Characteristics or features of an item Label for the item

Bicycle

Both examples show how a model trained with supervised learning can accurately classify
new data based on the patterns it learns during training.
Supervised learning is a method that helps you understand how a machine learning
model is trained. Discriminative modelling is how the model uses what it has learnt to
identify the category or label of a particular input.
180 Artificial Intelligence | Kips
Discriminative models are also called predictive models because their role is to predict
the category or label of a given input based on the patterns learnt from the training
data. For instance, in example 2, the label "bicycle" has specific features, like wheels
and handlebars. With the help of these features, the model becomes more confident in
predicting that the object is a bicycle.

K i B o ’ s Fa c t

Training data is the initial dataset used to train an AI model.

Testing data is used to evaluate the performance of the AI model.

Unsupervised Learning and Generative Modelling

Unsupervised learning is also called self-supervised learning. In unsupervised learning,
the learning of the machine (model) is done with the help of unlabelled data. Consider
the following illustrations, where images do not have any labels.

Unlabelled Datasets

This means that the data that is fed to the machine is random. In unsupervised learning,
different algorithms are used to identify or learn the relationships, patterns, and trends
within data, which can then be used to make predictions in the future or generate new
data.
Example 1:
Suppose, you want to predict a model to identify whether it is 'Elephant', 'Camel', or
'Cow', you provide it with a dataset containing images or descriptions of these animals
without any corresponding labels (names). The model does not have predefined labels,
so it discovers patterns, groups similar data, or extracts meaningful insights from the data
on its own (without human intervention), i.e., in the absence of a labelled dataset, it learns
from experience.

Input Raw Data Output

Interpretation Algorithm Processing

Unknown Output
No Training Data Set
Model Training
Kips | Artificial Intelligence 181
Example 2:
Consider another example where you are supposedly providing various pictures of
bicycles to a machine (model). Over the time, the model learns the common features
and characteristics shared by bicycles. Once trained, the model can use this knowledge
to generate an image of a bicycle that it has never seen before. This ability to create new,
realistic images based on learnt patterns is a key feature of generative models in machine
learning.

Unstructured/unlabelled Emergent pattern/inherent Example that is similar

dataset structure to what is in the dataset

What is Generative AI?

As the name suggests, generative AI is the AI technology that can generate any content. It
uses AI algorithm to generate content, which could be text, images, music, speech, code,
or any other type of data.
Generative AI models learn patterns and structures from large datasets during training
and use this knowledge to generate new data that is similar, but not identical, to the
training data.
The content generated through generative AI resembles human-generated content. Even
sometimes, it becomes impossible to identify whether the content is real or generated by
AI.

KiBo's Activity 2 Writing Subjective Responses

Generative AI
Visit the following link to watch the videos:
https://www.youtube.com/watch?v=om7iYSucLrk
https://www.youtube.com/watch?v=26fJ_ADteHo

182 Artificial Intelligence | Kips

After watching the videos, answer the given questions.

A. Fill in the blanks.

1. Large language models are trained on huge libraries of _____________________________ .
2. To get more accurate output from generative AI tool, prompt should be in more
_____________________________ .

B. Answer the following questions.

1. What do you understand by generative AI?

____________________________________________________________________________________

____________________________________________________________________________________

____________________________________________________________________________________

____________________________________________________________________________________

2. Give a few examples of generative AI.

____________________________________________________________________________________

____________________________________________________________________________________

____________________________________________________________________________________

____________________________________________________________________________________

3. What do you know about DeepFake?

____________________________________________________________________________________

____________________________________________________________________________________

____________________________________________________________________________________

____________________________________________________________________________________

Generative AI Timeline
Advancements in AI technologies like deep learning, which emphasise neural networks,
gave birth to generative AI. Initially, AI models were used only for identifying or
classifying objects or images, predicting weather and stock prices, assessing the risk of
disease, understanding human language, etc.
Today, many applications of generative AI exist, such as text generation, image synthesis,
and the production of creative content that displays the fruits of years of research and
development.
Kips | Artificial Intelligence 183
The different models developed throughout the development of generative AI are
depicted in the following picture:
2022
2020 Imagen
Midjourney 2022
2016-2019 Stable Diffusion
2014 OpenAI Five
2011 Generative Adversarial 2020 2022
IBM's Watson Networks GPT-3 ChatGPT

2013 2015 2017 2019

Variational TensorFlow Transformers GPT-2 2023 2023
Autoencoders GPT-4 LLAMA
(VAEs) 2018 2023
2016 GPT Bing Chat 2023
Google DeepMind 2023 Adobe
AlphaGo Google Bard Firefly

Generative AI vs Conventional AI
Regular or conventional AI processes or categorises already-existing data. It predicts the
output based on predefined rules or patterns that it learns from the data. In contrast,
generative AI is specially made to produce new and unique content. It uses patterns learnt
from training data to generate new, similar data.
Here are some significant variations between generative AI and conventional AI:
Basis Generative AI Conventional AI
Goal Generative AI creates new content Conventional AI analyses, processes,
that resembles the training data. and classifies data.
Example: When you ask ChatGPT to Example: A stock price prediction
write an essay, it can write a complete model can only predict the future
essay on a given topic. It means that it value of any given stock.
generated the content that was based
on a query given by the user.
Training Generalised generative AI models Conventional AI models can be
are prepared using a huge amount of prepared using less data, limited
data and multiple complex Python Python libraries (TensorFlow, Keras,
libraries, like ChatGPT. scikit-learn), and minimal processing
Also, a large amount of processing capacity.
capacity is required because the data Example: Normally, a model for
is very large. weather prediction will require
Example: GPT-3 was trained with data in megabytes. It uses a single
hundreds of gigabytes of text data. processor for training.

184 Artificial Intelligence | Kips

 It uses multiple processors for
training. It includes Python libraries,
like TensorFlow, StyleGAN2-
ADA, GAN Lab, OpenAI Gym, and
Transformers.
Output Generative AI can produce fresh, Conventional AI produces more
innovative, and often unexpected predictable output based on existing
output. data.
Example: It can generate images or Example: The output of a model for
videos from text. It can also convert disease prediction will be either yes
text to audio. or no.
Applications Applications of generative AI include Conventional AI is used for
generating music, literature, code, personalised applications such as
design, images, videos, etc. assessing students' performance in
education, predicting product prices
in business, and identifying diseases
in the medical field.

KiBo's Activity 3 Critical Thinking

Generative AI vs. Conventional AI

Tick () whether the following is an application of generative AI or conventional AI.

Example Generative AI Conventional AI

1. An organisation using a question-answer system to help
its customers

2. Model for classifying cats and dogs

3. An interior designer using AI to create multiple design

concepts for a client’s living room

4. A model for identifying objects in an image

5. Model for generating music

6. An email service provider using AI to classify emails as

spam or legitimate
7. A streaming service using AI to suggest movies and
TV shows to users based on their viewing history and
preferences
8. A news agency using AI to produce news articles on
trending topics
9. Assisting artists by providing inspiration or creating
entirely new pieces of art

Kips | Artificial Intelligence 185

 10. A marketing team using AI to design and generate
multiple versions of advertisements tailored for different
demographics and platforms

Types of Generative AI
Generative AI comes in a variety of forms or types, each with its unique advantages and
uses. Some of its most typical forms are listed below:

Generative Adversarial Networks (GANs)

Generative Adversarial Network uses two neural networks that work in collaboration to
generate new data. The names of these two neural networks are generator network and
discriminator network. The task of the generator network is to produce new data, while
the task of the discriminator network is to analyse the generated data and provide feedback.
The generator network generates the data until it receives feedback from the discriminator
network that the generated data is similar to real data.
Examples: Generating fake images, creating portraits of non-existing people, converting
images from day to night, generating images based on textual descriptions, generating
realistic videos, etc.

Variational Autoencoders (VAEs)

Another class of generative models are VAEs. VAEs learn the distribution of data and
then generate new data with it. They take data (images, text, etc.) and condense it into
a smaller, more manageable form. This compressed version captures the essence of the
data. Then, the VAEs can use this compressed form to generate entirely new data points
that resemble the originals, but with a twist of creativity. This makes VAEs useful for
generating new content that looks similar to the original data on which they were trained.
Examples: Generating new images similar to the given training set, image reconstruction,
generating drafts for writers, generating new sounds and music composition, etc.

Recurrent Neural Networks (RNNs)

These are also the types of neural networks that are specifically designed to handle
sequential data, such as music or text. They use the feedback loop to remember previous
information.
Example: When you draft an email on Gmail, it helps you complete the sentences by
suggesting the next word or sentence.

186 Artificial Intelligence | Kips

Autoencoder
These neural networks are trained to learn a compressed representation of data. First,
the data is compressed and then decompressed to restore it to its original form. They
generate highly realistic samples.
Examples: Image compression, noise removal from images, artistic image creation, drug
discovery

Examples of Generative AI
Generative AI is used in many fields, like art, music, language, and natural language
processing, as shown below:

Art and Financial

Design Forecasting

Music Customer Behaviour

Composition Analysis

Content Demand
Creation Forecasting

Healthcare
Virtual Diagnosis and
Worlds Prognosis

Fashion and
Product Design Fraud
Detection

Let us explore some examples of generative AI in different fields.

Art Generation
Generative AI is being used to create unique works of art. The following are the different
examples of generative AI in art generation:
The Next Rembrandt Project uses data analysis and 3D printing to create a new
painting in the style of Rembrandt.
Note: Visit the following link to watch the video:
https://www.youtube.com/watch?v=IuygOYZ1Ngo)
AI models, like OpenAI's DALL.E, help generate unique visual concepts from textual
descriptions.

Kips | Artificial Intelligence 187

 Generative AI tools, like DeepArt and Prisma, can be used to apply the style of one
image to another.
AI filters can be used to convert images into sketches.
Different patterns and sketches can be generated with the help of generative AI.
Personalised or customised artworks can be generated based on user preferences.
AI can upscale low-resolution images and add colour to black-and-white photos.
3D models and animations can be created, facilitating digital sculpting and motion
graphics.

AI Tools for Art Generation

DALL.E 2 Runway
MidJourney DeepArt
(OpenAI) ML

NightCafe Deep Dream

Artbreeder Prisma
Studio Generator

Music Generation
Generative AI is also being used to create new music, create original compositions, and
even collaborate with AI to produce innovative music.
Note: Visit the link to understand how AI can compose a personalised soundtrack to your
life. https://www.youtube.com/watch?v=wYb3Wimn01s
The following are the different examples of generative AI in music generation:
Generative AI can be used to create melodies and harmonies.
Generative AI can help songwriters by generating lyrics based on specific themes,
keywords, or phrases.
AI can suggest instrumentation and orchestration arrangements for a specific song.
Generative AI can be used during live performances to create real-time music.
AI can create realistic environmental sounds.

AI Tools for Music Generation

AIVA (Artificial
Amper OpenAI
Intelligence
Music MuseNet
Virtual Artist)

Soundraw Melobytes Ecrett Music

188 Artificial Intelligence | Kips

Language Generation
Generative AI plays a pivotal role in language generation tasks, where it can produce
human-like text based on input data or user prompts.
Note: Visit the link to know about ChatGPT.
https://www.youtube.com/watch?v=wYb3Wimn01s
The following are the different examples of generative AI in language generation:
Written content for websites, blogs, and news articles can be generated.
Natural Language Generator systems can be used to generate reports, summaries,
and documents.
Chatbots interact with users in natural language. They may be used for customer
support.
Virtual assistants like Siri, Alexa, and Google Assistant use generative AI to understand
user commands and generate responses.
AI can be used for translation of multiple languages.

AI Tools for Language Generation

ChatGPT
Google's T5
(Generative
(Text-To- EleutherAI
Pre-trained BERT RoBERTa
Text Transfer GPT Models
Transformer
Transformer)
Models)

Benefits of Using Generative AI

Generative AI has impacted human life
in many positive ways. The benefits of Scalability
generative AI are as follows:
Creativity: Generative AI tools can Accessibility Creativity
generate different types of content in a
very attractive and creative manner. For Benefits
example, content can be generated in the
form of an image, video, or audio, which Exploration Efficiency
is a better medium of communication as
compared to textual information.
Personalisation

Kips | Artificial Intelligence 189

Efficiency: AI tools can generate summarised and precise information, which may be
very useful and save a lot of time.
Personalisation: Generative AI tools can generate content in the form of text, image, or
video as per the user's requirements. For example, if you have an imagination of a scene
that may have a red rose in a river, you can generate that scene.
Exploration: Generative AI can be used to explore new design spaces or optimise complex
systems, such as designing new medicines or improving industrial processes. It helps in
developing and giving strength to an idea because it can draw your imagination. Also,
generative AI tools can be used to generate huge datasets for the purpose of training and
testing other AI models.
Accessibility: Many generative AI tools are freely available and easy to use. By entering a
prompt into the tool, you get whatever you want. For example, tools like GPT-3 can write
articles, generate marketing copy, and draft emails.
Scalability: Generative AI can quickly and efficiently produce large amounts of content,
providing a scalable solution for businesses and organizations that need high-volume
content production.

Limitations of Using Generative AI

There are various benefits of using
generative AI. But every useful tool also
High Data
has some limitations. The limitations of Computational Bias
generative AI are as follows: Demands
Data bias: AI models are trained with the
help of a huge amount of data. If the data Ethical
used for training is biased, then the results Limitations Uncertainty
Issues
produced by the model may not be of any
use or may be inaccurate. Reuse
Limited
Uncertainty: Sometimes the results Understanding of
of Context Existing
produced by generative AI models are
Data
unpredictable and uncertain. For example,
if you ask a generative AI model about a
specific disease, it may provide incorrect or unrelated information, especially if it has not
been trained on a diverse and comprehensive dataset.
Reuse of existing data: Generative AI cannot create entirely new concepts or ideas from
scratch. It remixes and rearranges existing information to generate output.

190 Artificial Intelligence | Kips

Limited understanding of context: It cannot understand long contexts. The question that
you asked two hours ago may not be linked to the question that you are asking right now.
Ethical issues: The ability of generative AI to generate realistic content raises ethical
concerns. For example, creating DeepFakes or artificial media for malicious purposes can
have negative consequences.
High computational demands: It is easy to use generative AI models. But it is very
complex to develop generative AI models because they require huge amount of data and
processing capacity.

A I

LAB
Experiment 1: GAN Paint Hands-on Activity
GAN Paint can be used to draw pictures because it is trained to create pictures. Each left button
("door", "brick", etc.) represents 20 neurons. Switching these neurons directly shows the network's
visual world model. The software shows that the network learns about trees, doorways, and roofs
by drawing. Let us experiment with GAN Paint and see what you can create.
To use GAN Paint, do the following:
Visit the following link. https://ganpaint-v2.vizhub.ai/
On visiting the link, you will get the interface for GAN Paint, as shown below:

First, you need to select a base image from the website's library. For this, select any image
from the ones appearing in the bottom panel of the interface.
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.

Kips | Artificial Intelligence 191

 Generative AI Tools
Many generative AI tools are available today that enable users to create and experiment
with generative models. Here are some popular tools:

Artbreeder
Artbreeder is an AI-powered platform that enables users to generate new images by
combining different generative adversarial network (GAN) models. Users can select
and combine different GAN models to create new and unique images. This platform
enables users to create artwork, character designs, landscapes, and more by combining
and tweaking different elements.

A I

LAB
Experiment 2: Artbreeder Hands-on Activity
Let us generate images with text prompt using Artbreeder.
To use Artbreeder, do the following:
Visit the following link: artbreeder.com
The following interface appears. By default, you are on the welcome page. Click on the Create
tab located at the top-centre of the window.

You will get the different categories for generating images. Click on the New Image button
under the category of your choice. For example, scroll down the web page and click on the
New Image button under the Prompter category.

192 Artificial Intelligence | Kips

 Another window appears. Here, enter a text prompt in the Description box and see how
AI generates a picture from those prompts. For example, give the prompt, "A kid reading a
book".
Click on the generate button. You will get two options of images.
You can download the images by clicking on the download button or save them by clicking
on the Save to Library button.

Give more prompts and see how AI generates your imagination using your text prompts.

Runway ML
Runway ML is a creative platform that offers a collection of pre-trained machine learning
models that can be used for many tasks, like image and video generation, style transfer,
object detection, and more. It provides an easy-to-use interface for building, training,
and deploying various types of generative models, including GANs, VAEs, and image

Kips | Artificial Intelligence 193

classifiers. It enables users to incorporate AI into their creative workflows without needing
extensive programming knowledge.
Visit the following link and watch the video:
https://www.youtube.com/watch?v=trXPfpV5iRQ

A I

LAB
Experiment 3: Runway ML Hands-on Activity
Let us explore AI magic tools of Runway ML.
To use Runway ML, do the following:
Visit the following link: https://runwayml.com
You will get the welcome page of Runway ML platform. Click on the Try Runway for Free
button to explore it.

You will get the next page. It will ask you to login. Login with your Google account by clicking
on the Log in with Google button.
After login, you will get the following screen. Select any tool of your choice under Runway's
AI Tools to generate new content with it.

194 Artificial Intelligence | Kips

For example, click on the Text/Image to Video option to generate a video.

You will move on to the next window that allows you to enter a text prompt to generate
a video. Type the prompt, "A lion forest", in the Describe your shot... box and click on the
Generate 4s button. You will notice a progress bar that displays the progression of the video.
Wait until the progress bar is completed.

You will get the video once the progress bar is completed, as shown below. Click on the
Play button to watch the video. You can download the video by clicking on the
Download button.

In the same way, you can use other AI tools of Runway ML.

Kips | Artificial Intelligence 195

ChatGPT, Gemini AI, and Copilot
These are popular generative AI tools. They allow you to enter a prompt for content
creation. Nowadays, they are everyday AI companions. They let you search for specific
information, generate new text (such as emails and summaries), or create images based
on text prompts you write. For example, you can ask these tools to summarise a memo in
two sentences, and they will provide a concise written summary.

A I

LAB
Experiment 4: ChatGPT Hands-on Activity
To use ChatGPT, do the following:
Visit the following link:
https://chat.openai.com
ChatGPT platform asks you to login. Login with your Google account. After login, you will
get the interface of ChatGPT.
Type your text prompt in the Message ChatGPT textbox and press Enter. ChatGPT gives you
a response according to your prompt.

196 Artificial Intelligence | Kips

 Enter a prompt; for instance, "Introduce yourself", and you will get the following response
from ChatGPT:

Experiment 5: Gemini AI
To use Gemini AI, do the following:
Visit the following link: https://gemini.google.com
Gemini AI platform asks you to login. Login with your Google account. After login, you will
get the interface of Gemini AI.
Type your text prompt in the Enter a prompt here textbox and press Enter. Gemini AI gives
you a response according to your prompt.

Kips | Artificial Intelligence 197

 Enter a prompt; for example, "Today's news", and you will get the following response from
Gemini AI:

Note: In the same way, you can interact with Copilot by visiting the following link:
https://copilot.microsoft.com/

KiBo's Activity 4 Computational Thinking

Chit Chat with ChatGPT, Gemini, and Copilot

Interact with these tools using the given queries and observe the responses.
1. Write a summary of the history of the internet.
2. How to code a simple website?
3. Write a blog post about the latest trends in artificial intelligence.
4. Create a presentation on the benefits of cloud computing.
5. Write a research paper about the future of technology.
6. Design an application that solves a real-world problem.
7. Today's news.
Now, consider the following parameters and write your findings in the table provided below:

Parameters ChatGPT Response Gemini Reponses Copilot

Human-like
Response
Training Dataset and
Underlying Technology
Authenticity of
Response

198 Artificial Intelligence | Kips

 Access to the
Internet

User Friendliness and

Interface

Text Processing:
Summarisation,
Paragraph Writing, etc.

Price

More Generative AI Tools

The following are some more generative AI tools that can be used in various fields.

Video
Muse AI Visla AI Topas AI

Text
ChatGPT Notion AI Compose

Images
Midjourney Magic Studio Pebblely

Design
Viesus Piggy AI Galileo

Coding
Bugasura CodeGPT Rapit Ghostwriter

Audio
FineShare Boomy AI Playlist AI

Productivity
Briefly AI Socra AI Viesus

Writing
Jasper Ryter Anyword

Voice
Generator Murf Lovo Resemble

Data
Tableau Polymer Sisense

Kips | Artificial Intelligence 199

 Ethical Considerations of Using Generative AI
While generative AI offers many benefits, there are also several ethical considerations
that should be taken into account when using this technology.
Ownership: It is always a concern to determine who will be the owner of 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 work, although it may be based
on some existing content.
Human autonomy: Uncontrolled use of generative AI may create a question mark on
human autonomy. As technology becomes more sophisticated, it may become increasingly
difficult to distinguish between content generated by humans and machines.
Biased outcome: If the data used for training is biased, then the generative AI model will
also generate biased outcomes. This may lead to inaccurate results.
Misuse of tools: There are many generative AI tools, like Deepfake, that can be used to
generate fake images, videos, and content that can be misused for criminal activity.
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.

Potential Negative Impact on Society

Generative AI has the ability to create realistic and convincing content, but it presents
unique challenges to society. The following are some negative impacts of generative AI:
Misuse of the tools that can generate fake news, images, or videos may lead to many
negative consequences. There should be strict control over the use of these tools.
It may replace humans in some areas like content writing, video editing, and image
drawing. The challenge for humans will be to learn and adapt new technologies for
better work.
Generative AI can generate sensitive information, like medical data and security
numbers, which can also be misused.

Responsible Use of Generative AI

Generative AI has many benefits, but its creation and use should be done in a responsible
manner. The following responsibilities should be adhered to before creating or using any
generative AI tool:
Ensure that the training data used is diverse and representative.
200 Artificial Intelligence | Kips
 The outputs that are generating bias or misinformation should be filtered.
User privacy and consent should be given the highest priority.
Clear guidelines should be decided regarding the ownership and attribution of
generative content.
Public should be made aware of generative AI's ethical use and the implications of
misusing it.
Responsible use of generative AI is essential for ensuring that this technology is
developed and used in ways that benefit society.
By now, you must have understood that responsible use of generative AI is essential. It
ensures that this technology is developed and used in a way that benefits society. That
is why, by emphasising ethics, creating trust, limiting negative effects, defining law, and
encouraging innovation, we may maximise generative AI’s potential to improve society.

R c ap
Generative AI is an advanced form of AI that creates new content in the form of images, text, and
audio, based on the data it has learnt from.
Generative models are trained using supervised and unsupervised learning techniques.
In supervised learning, the learning of the machine (model) is done with the help of labelled data.
In unsupervised learning, the learning of the machine (model) is done with the help of unlabelled
data.
Generative AI models learn patterns and structures from large datasets during training and use
this knowledge to generate new data that is similar to the training data but not identical.
The content generated through generative AI resembles human-generated content.
Regular AI or conventional AI predicts the output based on predefined rules or patterns that they
learn from the data; whereas, generative AI is specially made to produce new and unique content.
Generative AI comes in a variety of forms or types, like Generative Adversarial Network (GAN),
Variational Autoencoders (VAE), Recurrent Neural Networks (RNN), and Autoencoder.
Generative AI is being used in many fields, like art, music, language, natural language processing,
etc.
Many generative AI tool, like Artbreeder, Runway ML, ChatGPT, Gemini, Copilot, etc, are available
today that enable users to create and experiment with generative models.
There are several ethical considerations with generative AI, like ownership, human autonomy,
biased outcome, misuse of tools, and privacy.

Kips | Artificial Intelligence 201

 AI Quiz
Objective Type Questions Solved

A. Select the correct option.

1. Generative AI focuses on _________________ .
a. Solving complex calculations b. Generating new content
c. Optimising software d. Performing human-like reasoning
2. Which type of machine learning model requires labelled data for training?
a. Unsupervised learning b. Self-learning
c. Supervised learning d. None of these
3. Which type of machine learning model requires unlabelled data for training?
a. Supervised learning b. Unsupervised learning
c. Self-learning d. None of these
4. Discriminative models are also called _________________ .
a. Predictive models b. Superactive models
c. Regession models d. Proactive models
5. Unsupervised learning is also called _________________ .
a. Independent learning b. Pattern learning
c. Self-supervised learning d. Individual learning
6. Which of the following processes or categorises already-existing data?
a. Conventional AI b. Generative AI
c. Both a and b d. None of these
7. Which of the following is not a common type of generative AI model?
a. Generative Adversarial Networks (GANs)
b. Variational Autoencoders (VAEs)
c. Autoencoders
d. Decision Tree
8. Generative Adversarial Network (GAN) is a combination of _________________ neural
networks.
a. One b. Two
c. Three d. Four

202 Artificial Intelligence | Kips

 9. Which of the following neural networks is specifically designed to handle sequential data,
such as music or text?
a. GAN b. VAE
c. RNN d. Autoencoder
10. Generative AI models are often trained on _________________ .
a. Small sets of handcrafted rules b. Huge amount of data
c. User-provided feedback loops d. Predefined outputs
11. What does VAE stand for?
a. Variety Autoencoder b. Variational AI Encoder
c. Variational Autoencoder d. Variational Autoencoding
12. In AI, algorithms learn _________________ from data.
a. Formula b. Patterns
c. Image d. Sound
13. Which of the following is not a generative AI tool?
a. ChatGPT b. Gemini AI
c. Microsoft Word d. Copilot
14. Which of the following is an example of a generative AI that generates images?
a. Artbreeder b. Google Photos
c. ChatGPT d. Both b and c
15. Which of the following is a potential benefit of generative AI?
a. It can replace humans in all fields.
b. It can only be used for entertainment purposes.
c. It can generate new ideas and content.
d. None of these

B. Fill in the blanks.

1. _________________ is an advanced form of AI that creates new content.
2. Generative AI models learn _________________ and _________________ from large datasets
during training.
3. _________________ learning is a method that helps you understand how a machine learning
model is trained.
4. _________________ models predict the category or label of a given input based on the learnt
patterns from the training data and use this knowledge to generate new content.
5. The content generated through generative AI resembles _________________ content.

Kips | Artificial Intelligence 203

 6. _________________ data is the initial dataset used to train an AI model.
7. _________________ data is used to evaluate the performance of the AI model.
8. In GAN, the task of the _________________ network is to analyse the generated data and
provide feedback.
9. _________________ are neural networks that are trained to learn a compressed representation
of data.
10. _________________, like Siri, Alexa, and Google Assistant, use generative AI to understand
user commands and generate responses.
11. If the data used for training is biased, then the results produced by the model may be
_________________ .

12. _________________ is an AI-powered platform that enables users to generate new images by
combining different Generative Adversarial Network (GAN) models.

C. Write T for True and F for False.

1. Generative AI is simpler than traditional AI.
2. Generative AI cannot be used for image generation.
3. Generative AI model requires huge amount of data and processing capacity.
4. Results may be unpredictable in generative AI.
5. Spam email detection is an example of generative AI.
6. Question answering systems are examples of generative AI.
7. The important challenge in generative AI is to have unbiased data.
8. The important ethical issue in generative AI is deciding who will be the owner of
the content generated by AI.
9. Recurrent neural networks are suitable for image data.
10. Artbreeder is an art generation tool.

Answers
A. Select the correct option.
1. b 2. c 3. b 4. a 5. c 6. a 7. d 8. b
9. c 10. b 11. c 12. b 13. c 14. a 15. c

B. Fill in the blanks.

1. Generative AI 2. Patterns, Structures 3. Supervised 4. Discriminative or Predictive
5. Human-generated 6. Training 7. Testing 8. Discriminator
9. Autoencoders 10. Virtual Assistants 11. Inaccurate 12. Artbreeder

C. Write T for True and F for False.

1. F 2. F 3. T 4. T 5. F 6. T 7. T 8. T
9. F 10. T

204 Artificial Intelligence | Kips