Impact, Opportunity, and Challenges of Generative AI

IMPACT,
A MEITY, NEGD & NASSCOM INITIATIVE

OPPORTUNITY AND
CHALLENGES OF
GENERATIVE AI

1
This work is licensed under a Creative Commons Attribution 4.0 International License.

You are free to:

Share — copy and redistribute the material in any medium or format
Adapt — remix, transform, and build upon the material
for any purpose, even commercially.

Under the following terms:

Attribution — You must give appropriate credit, provide a link to the license, and
indicate if changes were made. You may do so in any reasonable manner, but not in
any way that suggests the licensor endorses you or your use.

No additional restrictions — You may not apply legal terms or technological measures
that legally restrict others from doing anything the license permits.

Authored by Humans
Lead Author: Jibu Elias
Co-Author: Anjali Raja
Research: Nibedita Saha & Dr Nivash Jeevanandam
Illustrations: Generated by Tapan Aslot
 Impact, Opportunity, and Challenges of Generative AI

Sri Narendra Modi

Honourable Prime Minister of India

“Artificial intelligence is a tribute

to human intellectual power,
the power to think enabled
humans to make tools and
technologies. Today, these tools
and technologies have also
acquired the power to learn
and think. In this, one key
technology is AI. The teamwork
of AI with humans can do
wonders for our planet.” 1

From the inaugural speech of RAISE

1.
3
2020 Summit
Impact, Opportunity, and Challenges of Generative AI

4
 Impact, Opportunity, and Challenges of Generative AI

CONTENTS

Introduction 6
INDIAai Generative AI Roundtables 7
What is Generative AI 10
Foundation models 11
Unimodal and Multimodal 11
Popular Generative AI Models 14
Large Language Models 18
Some of the popular Large Language Models 18
List of Large Language Models 20
Current state of Generative AI 21
Public perception of AI 21
Is Generative AI ‘creative’ or ‘generative’? 22
Future of jobs 22
Total job growth and job loss 24
Generative AI will transform work across industries 25
Opportunities 29
Generative AI tools from India 30
Bridging the digital divide 31
Generative AI’s impact on different sectors 32
Generative AI impact on different sectors in India 33
Next frontiers for Generative AI 35
Challenges 36
Ethics, rights, and intellectual property 37
How to ensure responsible use of Generative AI? 38
Regulation 38
AI regulation efforts across the globe 39
Recommendations 43
References 44

5
Impact, Opportunity, and Challenges of Generative AI

Introduction
Artificial Intelligence, both as an area of research and technology, has existed
for more than half a century with varying levels of progress and success. The
origin of machine learning and deep learning, fuelled by vast data pools and
advancements in semiconductors and the internet, brought the most significant
leap in the 2000s, giving us fascinating AI programs such as AlphaGo and Alpha
Fold. However, the advent of transformer architecture in 2018 and the subsequent
boom of Generative AI models, huge language models such as a GPT, has finally
brought AI into the mainstream in recent months.

With the growing popularity of Generative AI tools such as Chat GPT and Midjourney,
and tech companies such as Microsoft and Google in a race to incorporate these
models into their search and enterprise offerings, we have reached a point where
AI, especially Generative AI, is causing a significant influence in our day to day
lives and beyond.

Considering this growing influence of Generative

AI, INDIAai - the National AI Portal of the
Government of India, conducted numerous
research and held three roundtables that featured
some of the prominent voices in Generative AI, AI
Policy, AI Governance and Ethics, and academia
to analyze the impact, ethical and regulatory
questions, and opportunities it brings to India.
This report is the outcome of that process and is intended to be used by policymakers,
entrepreneurs, practitioners, and students.

6
 Impact, Opportunity, and Challenges of Generative AI

INDIAai Generative AI Roundtables

In the last few years, Generative AI technologies have On the other hand, the socio-economic impact that
exploded, ranging from sophisticated language models Generative AI models can create is broad and deep
such as GPT-3 and image generation models such as and can even threaten our country’s unity and national
DALLE-2. With the investment in AI to reach $422.37 fabric, leading to the question of how we can ensure the
billion by 2028, at a 39.4 % CAGR, experts believe ethical and responsible use of Generative AI models.
Generative AI will play a critical role in pushing AI Furthermore, the question of rights, attribution and IP of
innovation and investment in the coming decade.2
AI-generated works that mimic specific human creators
Generative AI tools, primarily focused on low-cost and opens more ethical questions.
higher-value solutions, are perceived as the future of
text, image and even code generation, sometimes With these questions in mind, INDIAai organised a series
indistinguishable from human creation. However, with of roundtables with stakeholders and ecosystem players
the pace at which these technologies advance, we to understand the present Generative AI landscape,
need to foresee the risks it brings. On the one hand, ethical and legal challenges, and solutions to mitigate
countries such as China have implemented regulatory
the harmful impact these models can have in our society.
measures to put a leash on Generative AI risks, such as
watermarking the final product, auditing the algorithms, The first roundtable on 31st Jan 2023 was a big success
and even ensuring the consent of users are taken before as it explored the current state of Generative AI and
using their data for training these models. analysed recent hype around the subject.

The first roundtable on 31st Jan 2023 was a big success as it explored the current state of Generative AI and
analysed recent hype around the subject.

The participants were:

Prof Chandra Sekhar Seelamantula Aman Taneja

Professor, Indian Institute of Science Principal Associate, IKIGAI Law

Dr. Jyoti Joshi Emmanuel R Goffi

CEO & Founder, Kroop AI Founder and Director, Global AI Ethics Institute

Kazim Rizvi Aravind Aggarwal

Founder, The Dialogue Senior Research Scientist, IBM Research Lab, India

Tapan Aslot Abhinav Aggarwal

Communication Designer & Generative AI Artist Co-Founder & CEO, Fluid AI

2.
Industry Trends, Share, Growth, SWOT Analysis, Forecast by Zion Market Research

7
Impact, Opportunity, and Challenges of Generative AI

The second roundtable was organised on 28th Feb 2023, which looked deeper into the ethical questions regarding
Generative AI and the need for regulatory frameworks and how to formulate a policy approach.

The participants were:

Shashank Reddy Divya Dwivedi

Managing Partner, Evam Law & Policy Lawyer, Supreme Court of India

Prateek Sibal Anna Danes

Technology Policy Researcher, UNESCO Data Ethicist, OdiseIA

Vibhav Mithal
Managing Associate, Anand & Anand / For
Humanity

The third roundtable was held on 24th Apr 2023, and it focused on the India opportunity in Generative AI, especially
the economic impact and how this technology could transform various sectors and parts of the economy.

The participants were:

Aveekshith Bushan Jaspreet Bindra

Vice President, GM - APJ - Aerospike, Inc Author, Founder of TechWhisperer

Deepak Visweswaraiah
Neethu Joy, Ph.D.
Vice President, Platform Engineering and Site
Founder & CEO of a stealth startup
Managing Director, Pegasystems

Harsha Mundhada S Pavankumar Dubagunta

Principal, Inflexor ventures, Speech Scientist, Uniphore Software Systems

This report contains essential learnings, findings, analysis, and opinions from these roundtables.

8
Impact, Opportunity, and Challenges of Generative AI

9
Impact, Opportunity, and Challenges of Generative AI

What is Generative AI?

Generative Artifical Intelligence (GAI) describes to generate sentences is to learn word distribution
algorithms (such as ChatGPT, Midjourney, Bard, using N-gram language modelling and then search
DALL E, etc.) that can be used to create new content, for the best sequence. This modelling can be used for
including audio, code, images, text, simulations, and disambiguating the input.
videos. Generative AI models use neural networks to
identify the patterns and structures within existing data In addition, they can be used for selecting a probable
to generate new and original content. solution. This modelling depends on the theory of
probability. Probability is to predict how likely something
In recent years, large-scale models have become will occur. Models that assign probabilities to sequences
increasingly important in the AI-generated content of words are called Language Models or LMs.
space, providing better intent extraction and, thus,
improved generation results. Furthermore, with the However, this method cannot effectively adapt to long
growth of data and the models’ size, the distribution that sentences, and this issue was tackled by introducing
these models can learn becomes more comprehensive Recurring Neural Networks for language modelling
and closer to reality, leading to more realistic and high- tasks.
quality content generation. The advent of generative AI models in various
Contrary to popular belief, Generative AI (GAI) models domains has followed different paths, but eventually,
have existed as a technology since the early days of AI. the intersection emerged due to the Transformer
The history of Generative AI models can be traced back Architecture. Introduced for NLP tasks in 2017, a
to the 1950s with the development of Hidden Markov Transformer Model is a neural network that learns
Models (HMMs) and Gaussian Mixture Models (GMMs), context and thus meaning by tracking relationships in
as they generated sequential data such as speech and sequential data like the words in this sentence.
time series. With the advent of Deep Learning, Generative Transformer models apply an evolving set of
Models saw significant performance improvements.
mathematical techniques, called attention or self-
The current boom of Generative AI has its origins rooted attention, to detect subtle ways even distant data
deeply in advance of Natural Language Processing elements in a series influence and depend on each
(NLP), which is a subfield within AI which focuses on other. Transformer was later applied in computer vision
how computers process and analyse large amounts of and then became the dominant backbone for many
natural language data. In NLP, the traditional method generative models in various domains.

10
 Impact, Opportunity, and Challenges of Generative AI

Unimodal – CV & NLP

DDPM
GAN Style- ViT
BiGAN GAN MoCo
VAE
RevNet BigBiGAN
Flow
GPT-3
ELMO OPT Sparrow
BERT BART chatGPT
GPT-2 T5
N-Gram LSTM/GRU Transformer

2014 2016 2018 2020

StyleNet CAVP Visual- CLIP

Show-Tell DMGAN BERT ALBEF DALL-E
StackGAN
VQ-VAE ViLBERT BLIP BLIP2
UNITER VQ-GAN DALL-E 2

Multimodal - Vision Language NLP CV VL

Source: A comprehensive survey of AI Generated Content (AIGC): A History of Generative AI from GAN to ChatGPT by Yihan
Cao & others, March 2023

Foundation models aligns with the user’s intent. To overcome this issue,
reinforcement learning from human feedback (RLHF)
Proposed to solve the limitations of traditional models has been applied to fine-tune models in various
such as RNNs, Transformer is the backbone architecture applications.
for many state-of-the-art models, such as GPT-3,
DALL-E-2, Codex, and Gopher. Since the introduction Developing computing with enhanced hardware,
of Transformer Architecture, pre-trained language distributed training, and cloud computing contributed
models have become the dominant choice in NLP due to the development of the foundation model.
to parallelism and learning capabilities. Generally,
these transformer-based pre-trained language models
can be commonly classified into two types based on Unimodal and Multimodal
their training tasks: Generally, GAI models can be categorised into
• Autoregressive language modelling unimodal models and multimodal models. Unimodal
• and masked language modelling models receive instructions from the same modality as
the generated content modality, whereas multimodal
Furthermore, despite being trained on large-scale models accept cross-modal instructions and produce
data, the AIGC may not always produce output that results of different modalities.3

3.
Yihan Cao, Siyu Li, Yixin Liu, Zhiling Yan, Yutong Dai, Philips S. Yu, Lichao Sun, March 2023. A Comprehensive Survey of AI-
generated Content (AIGC): A History of Generative AI from GAN to ChatGPT

11
Impact, Opportunity, and Challenges of Generative AI

A comprehensive Survey of AI-Generated Content (AIGC):

A History of Generative AI from GAN to ChatGPT

Unimodal
Data
Pre-train Once upon a
Please write a story Prompt Decode time there was
Instruction I1 Generative AI Models Result R1 a cat named
about a cat.
Jessy..
Multimodal
Describe this This is a cat.
Instruction I2 Pro Data Result R2
picture mp
t Pre-train
Draw a picture Instruction I3 Prompt Result R3
of a cat. t Generative AI Models
mp
Pro
Write a song Instruction I4 Result R4
about a cat.

Generative language models (GLMs) are unimodal Multimodal generations are relatively hard to learn
models trained to generate readable human language compared to unimodal. The generation of state-of-
based on patterns and structures in input data they have the-art multimodal in vision language generation, text
been exposed to. These models can be used for a wide audio generation, text graph generation and text code
range of NLP tasks, such as dialogue systems, translation generation aided in tackling this issue.
and question-answering. These include decoder models The application of these architectures can be seen in
and encoder-decoder models. Vision generative models Chatbots, AI art generation, music generation, coding
are other kinds of unimodal. for AI-based programming systems, and education.

12
Impact, Opportunity, and Challenges of Generative AI

13
Impact, Opportunity, and Challenges of Generative AI

Popular Generative AI Models

Name Description Function
ChatGPT It became one of the most famous chatbots ChatGPT works in the same way as automated
in just 2 months. chat services seen on customer support websites.
It can answer queries and write many types of
OpenAI has developed a language model written content, such as articles, social media
that is close to a conversation with humans. postings, essays, code, and emails.
It learns from interactions and processes
information based on its learnings.
Additionally, ChatGPT is trained on
Generative Pre-Trained Transformer
architecture — the neural network that
lends the ‘GPT’ to its name.
DALL- E DALL-E and DALL-E 2 are deep learning DALLE 2 can generate unique, realistic visuals
models developed by OpenAI to generate and art based on a text description. It is capable
digital images from natural language of combining concepts, traits, and styles. It can
descriptions, called “prompts”. be used with an image made in DALL-E 2 or
an image you’ve uploaded - it’s been used to
extend great works of art, such as the Mona Lisa,
by adding more backdrop to the image.
Midjourney Midjourney is a generative artificial Midjourney has introduced a new function called
intelligence program and service created /describe, which lets users upload an image and
and hosted by a San Francisco-based receive written prompts that attempt to explain
independent research lab Midjourney, Inc. it. These prompts can then be utilised to create
Midjourney generates images from natural one-of-a-kind and engaging AI-generated art
language descriptions, called “prompts” that is unlike anything else on the internet.
Chatsonic It is a revolutionary generative AI tool Chatsonic can generate high-quality text,
built to beat all the limitations of Open AI, graphics, content for multiple social networking
turning out to be the best alternative to networks, emails, and website content based
ChatGPT. It integrates with Google Search on the cues it receives. In addition to written
to create content with the latest information. commands, Chatsonic can be used with a voice
command.
Additionally, it can create digital images
and respond to voice commands.
Jasper Chat Jasper Chat is a new AI-powered generative Jasper Chat may produce ideas, modify text
way of interacting with generative AI.Instead passages, answer inquiries, and create creative
of having to think in commands or strict stuff like poems or stories. Jasper Chat is intended
prompts, you can converse with AI and for commercial use cases such as marketing and
refine responses with each engagement. sales, as well as social media posts.
Copy.ai Copy.ai is the new generative AI that is built Copy.ai assists the team with several types of
for sales and content marketing teams. copywriting and sales/copy, such as product
descriptions, ad copy, website copy, blogs, and
It can help you create articles, sales emails, emails. It can make bullet points out of LinkedIn
social media captions, ad copy, blog posts, profiles. From a single term, it may generate a
code and more all with real-time data. full content brief.
ChatFlash - You can quickly start using this chat feature Neuroflash creates high-quality artificial
Neuroflash by asking questions or giving instructions. intelligence material. It generates documents
AI The Chatflash AI will answer you and using multilingual, modern language models
provide results. (auto-regressive language transformers; GPT-3).
GrowthBar GrowthBar is an AI SEO content writing tool GrowthBar creates AI content using advanced
that specialises in AI content generation. algorithms that produce and optimise your blog
The main feature is the long-form content posts for you. Furthermore, GrowthBar performs
editor, which helps you to write blog posts in keyword research, rank tracking, and backlink
2 minutes flat. research in the same way as large SEO tools do.

14
 Impact, Opportunity, and Challenges of Generative AI

Name Description Function

Rytr Chat Rytr is an AI writing assistant that helps you Rytr Chat was created to give Rytr users a more
create high-quality content, in just a few natural and conversational experience. Rytr Chat
seconds, at a fraction of the cost. Powered allows you to create content with a single click. It
by state-of-the-art language AI to generate changes the content’s tone, duration, or style.
unique, original content for almost any vertical
Easy Peasy Easy-Peasy.AI is the AI Content Generator, Easy Peasy AI Chat allows users to easily
AI Chat which assists you and your team in breaking compose emails, social media messages,
past creative barriers to create fantastic, resumes, blog posts, and other documents. It
original content 10X faster. It is an artificial can also make AI visuals, summarise large text,
intelligence content tool that may assist and transcribe audio.
you with a number of writing jobs, such as
producing blog posts and creating better
resumes.
LaMDA LaMDA AI is a conversational Large It is a natural language processing tool
discourse Model (LLM) developed by Google developed by Google with 137 billion
to power dialogue-based systems that parameters. It was built by fine-tuning a group
generate natural-sounding human discourse. of Transformer-based neural language models.
For pre-training, the team created a dataset of
It has already been used for zero-shot 1.5 trillion words which is 40 times more than
learning, program synthesis, and BIG- previously developed models.
bench workshop.
GhostWryter Ghostwryter is an AI application that The app uses AI to help you create SEO
generates high-quality material for marketing optimised text, marketing content, and even blog
techniques such as SEO text and blog entries. posts for your business. You can simply write
It is easily available via Google Docs and blog posts, and other marketing content with
employs artificial intelligence technologies GhostWryter.
through the use of an OpenAI licence key.
Ellie AI Ellie AI is a platform that offers data teams Ellie is a writing assistant exclusively for drafting
with tools and frameworks to ensure that email replies. Ellie learns from your own
new data is constantly valuable to the individual writing style, and allows users to pick
business. It provides lightweight data a “mood” for her replies (i.e. her tone can be
modelling and information architecture casual, professional, irritated, etc.) Ellie is also
elements and is intended for enterprises multilingual and will automatically reply in the
of all sizes. Ellie AI also offers an AI email language to which she was initially written.
assistant that can respond to your emails in
your preferred style.
Murf AI Murf AI allows users to make studio quality Murf AI features a straightforward user interface
voice overs for things like explainer videos, that anyone can utilise. Simply upload your text
podcasts, advertisements and more using and select a voice from the library. You can vary
text. Just choose from the 120+ voice the voice’s tempo and pitch as you see fit. You may
styles available on the site, including 20+ also upload your own sound clip and utilise the
languages and accent options. voice changer to make it sound more professional.
Stockimg.AI Stockimg.AI aid in art generation. When Stockimg.AI is an AI image generating platform
generating an image, you let the AI know that makes it simple to create AI logos, book
if you are looking to make a book cover, covers, posters, and other graphics. Stock
logo, icon, wall paper or stock images, and pictures, book covers, posters, wallpapers, logos,
it creates visual content to suit your needs. illustrations, icons, and web and mobile UI are
just a few of the categories available.
Heyday Heyday calls itself an AI powered memory Heyday by Hootsuite is a conversational AI
assistant, but it’s actually more like a platform developed to assist retailers and
research assistant. The app works with eCommerce enterprises in capturing leads and
browsers, search engines, and apps like increasing buyer engagement. Customizable
Gmail and Google Docs to help “resurface” branding, multi-channel communication, wait time
content you may need to find once you’ve management, prioritisation, push notifications, and
already reviewed it. scripted responses are among the key features.

15
Impact, Opportunity, and Challenges of Generative AI

Name Description Function

Wordtune Wordtune does offer a generative writing Wordtune Read quickly and accurately
Read option, but what’s more interesting to summarises any document, making it possible
me is its reading option. Wordtune Read for anyone to read lengthy, complex texts in
summarises lengthy documents and helps a fraction of the time. Not only does it extract
highlight key points so that users can essential lines like other reading aids, but
understand more, more quickly. It can even thanks to our advanced AI technology, it also
provide different summaries based on a summarises the main themes for you.
specific point of view, for instance what
the text means for an overall industry or
decision-makers within the field.
Regie AI This app will help research potential Regie.ai is used to develop, test, and analyse
contacts and help sales agents create personalised prospecting sales campaigns.
personal emails that feel authentic to your This product package may be used to develop
company. Regie uses a large language interesting sales sequences, collaborate on
model to help create a voice specific to your content, and boost front-line efficiency. It can
brand. also be used to validate content so that people
can converse with confidence.
Resume The app was designed to provide instant Resume Worded is a career platform that assists
Worded feedback on your resume or LinkedIn you in adapting your resume to the position for
profile, including scores on key criteria which you are pursuing. This resume scanner will
that real recruiters use when hiring assess the job posting and highlight keywords
candidates. Users can also use the app to and skills that are missing from your resume in
identify keywords most relevant to the job 10 seconds after you upload your resume and
descriptions they’re searching for. the job description.
Looka Looka allows you to create a beautiful logo Looka provides high-resolution PNG and JPG
with the click of a button. Just input your logo files, SVG and EPS logo files for scaling,
company’s name, industry type, and color and colour variations for use online or offline.
palette, and you’ll have an endless range of Other features include business card designs,
logo possibilities at your fingertips. social media templates, brand information, and
lifetime logo support.
Google Bard Bard is Google’s AI-powered interactive chat Google’s Bard is a chatbot that uses
service. It is intended to function similarly to generative artificial intelligence to have natural
ChatGPT, with the main distinction being that conversations; it was built using the LaMDA
Google’s service will obtain its data from family of LLMs and then the PaLM LLM. In direct
the internet. Bard, like most AI chatbots, can competition with OpenAI’s ChatGPT, it was
code, solve math questions, and assist you created.
with your writing needs.
Adobe Adobe Marketo Engage’s Sensei GenAI Adobe Analytics can automatically analyse
Sensei aligns marketing and sales teams with customers and audience groups using Adobe
real-time conversational experiences that Sensei to uncover meaningful distinctions,
accelerate time to market, increase pipeline, attribute conversion drivers, and anticipate future
and increase calendared meetings. behaviour.
Adobe Firefly Adobe Firefly is a generative AI engine Adobe Firefly’s mission is to help people develop
for creativity. It’s just arrived in Adobe their innate creativity. Firefly, as an embedded
Photoshop, and it’ll change the way you model within Adobe products, will provide
create forever. generative AI capabilities tailored to creative
demands, use cases, and processes.
Amazon Amazon Bedrock is a fully managed service Amazon Bedrock’s serverless architecture makes
Bedrock that makes FMs from major AI companies it simple to get started, personalise FMs with
and Amazon available via an API, allowing your own data in private, and integrate and
you to select the model that is best suited deploy them into your applications with the AWS
for your use case from a diverse set of FMs. tools and capabilities.

16
Impact, Opportunity, and Challenges of Generative AI

17
Impact, Opportunity, and Challenges of Generative AI

Large Language Models Some of the popular Large

Large Language Models (LLM) are neural networks that
Language Models
analyse and comprehend natural language. Commonly
BERT
trained on large datasets, they can be utilised for
tasks including text generation, classification, question BERT (Bidirectional Encoder Representations from
response, and machine translation. Transformers) is a family of masked-language models
developed by Google researchers in 2018. According
Large language models process vast volumes of text data
to a 2020 literature review, “in little more than a year,
to approximate human speech. The LLM uses a deep
learning model, a network of interconnected neurons, to BERT has become a ubiquitous baseline in Natural
process, analyse, and forecast complex data to produce Language Processing (NLP) experiments, counting over
these natural language responses. 150 research publications analysing and improving
the model.”
Today, many tasks in NLP, such as speech-to-text and
sentiment analysis, rely on language models as their BERT was first implemented in English with two model
basis. These models can analyse a text and guess what sizes:
word will come next. For example, the parameters of an
LLM let it make predictions about the likelihood of word (1) B
 ERTBASE (12 encoders with 12 bidirectional self-
sequences by considering the text’s relationships. The attention heads totalling 110 million parameters)
model can capture complex relationships and handle
(2) B
 ERT LARGE (24 encoders with 16 bidirectional self-
unusual words with more parameters.
attention heads totalling 340 million parameters)
By 2030, a major blockbuster film will be BERT is a free and open-source NLP machine learning
released with 90% of the film generated by AI framework. The purpose of BERT is to provide context to
(from text to video), from 0% of such in 2022.4
help computers interpret ambiguous words in the text.

4.
Gartner Article, January 2023. Beyond ChatGPT: The Future of Generative AI for Enterprises

18
 Impact, Opportunity, and Challenges of Generative AI

GPT-3 has 176 billion parameters and was trained on about

GPT-3 (Generative Pre-trained Transformer 3) is a 366 billion tokens from March to July 2022. BLOOM
language model developed by OpenAI, a San Francisco- employs a modified Megatron-LM GPT-2 decoder-only
based artificial intelligence research laboratory. The transformer model design.
175-billion-parameter deep learning model can
produce text that resembles human language and was GPT-4
trained on large text datasets containing hundreds of OpenAI’s Generative Pre-trained Transformer 4 (GPT-
billions of words. 4) is a multimodal big language model and the fourth
in its “GPT-n” series of GPT foundation models. It was
GPT-3 employs deep learning to generate human-like published on March 14, 2023, and is now publicly
text. It will generate text that continues the prompt when available in a limited form through the chatbot product
provided a prompt. ChatGPT Plus (a premium version of ChatGPT), with
access to the GPT-4-based version of OpenAI’s API
LaMDA available through a waitlist.
Google’s LaMDA (Language Model for Dialogue
Applications) is a collection of conversational large Google - Bard
language models. Initially developed and released Google introduced Bard, a LaMDA-powered
as Meena in 2020, the first-generation LaMDA was conversational generative artificial intelligence
announced during the Google I/O keynote in 2021, chatbot. It is intended to function similarly to ChatGPT,
followed by the second generation the following year. with the main distinction being that Google’s service
LaMDA attracted significant attention in June 2022 will obtain its data from the internet.Google’s Bard
after Google employee Blake Lemoine claimed that the is a conversational generative artificial intelligence
chatbot had grown sentient. chatbot based on the LaMDA family of large language
models (LLMs) and later the PaLM LLM. It was created
BLOOM in direct response to the rise of OpenAI’s ChatGPT,
BigScience The huge Open-science, Open-access and it was initially published in a restricted capacity
Multilingual Language Model (BLOOM) is a large in March 2023 to lukewarm reviews before moving to
language model based on transformers. Over 1000 other nations.
AI researchers collaborated to construct a free big
language model for anyone who wants to explore it. It Furthermore, Bard, like most AI chatbots, can code, solve
is considered an alternative to OpenAI’s GPT-3, which maths questions, and assist you with your writing needs.

19
Impact, Opportunity, and Challenges of Generative AI

List of Large Language Models

Name Developer Year

BERT Google 2018
XLNet Google 2019
GPT-2 OpenAI 2019
GPT-3 OpenAI 2020
GPT-Neo EleutherAI March 2021
GPT-J EleutherAI June 2021
Megatron-Turing NLG Microsoft and Nvidia October 2021
Ernie 3.0 Titan Baidu December 2021
Claude Anthropic December 2021
GLaM (Generalist Language Model) Google December 2021
Gopher DeepMind December 2021
LaMDA (Language Models for Dialog Google January 2022
Applications)
GPT-NeoX EleutherAI February 2022
Chinchilla DeepMind March 2022
PaLM (Pathways Language Model) Google April 2022
OPT (Open Pretrained Transformer) Meta May 2022
YaLM 100B Yandex June 2022
Minerva Google June 2022
BLOOM Large collaboration led by Hugging Face July 2022
Galactica Meta November 2022
AlexaTM (Teacher Models) Amazon November 2022
LLaMA (Large Language Model Meta Meta February 2023
AI)
GPT-4 OpenAI March 2023
Cerebras-GPT Cerebras March 2023
Falcon Technology Innovation Institute March 2023
BloombergGPT Bloomberg L.P. March 2023
PanGu-Σ Huawei March 2023
OpenAssistant LAION March 2023
PaLM 2 Google May 2023

(Disclaimer: This is not complete list of LLMs, but a curated list by INDIAai team)

20
 Impact, Opportunity, and Challenges of Generative AI

Current state of Generative AI

With the launch of ChatGPT, Generative AI has taken gaining a clear understanding of what the tool can
over the world by storm. According to NVIDIA Founder accomplish.
and CEO Jensen Huang, Generative AI has finally
caused the inflexion point in AI. He calls it the iPhone “A significant number of people are still looking at AI as
moment for AI as a new computing platform has been a threat. There is generally low acceptance with regard
developed and emerged. to AI as people are still relating AI with Hollywood movies
they might have watched, in which it is often portrayed
“This is a new computing model that you program in a as how Ai is going to take over the world. There is a dire
new way, and this new way is using human language,” need to educate people,” said Generative AI artist Tapan
stated Huang. “The first computer you can program Aslot at the first INDIAai Generative AI Roundtable.
in any language you like; English, Chinese, French,
Japanese, however you like to do it,” he noted. Lack of basic understanding of AI is yet another major
issue. “The public is lost in the complexity of the field.
The open-to-public version of ChatGPT gained 100 explaining AI without hiding its complex nature is
million users in just under two months, while Stability essential, said Emmanuel Goffi, Founder and Director
AI’s Stable Diffusion, which can generate images based of Global AI Ethics Institute.”Some people do not even
on text descriptions, garnered more than 30,000 stars have access to 4G in countries like India and Africa.
on GitHub within 90 days of its release. Today, many Their knowledge of Chat GPT and Generative AI
believe that ChatGPT has the potential to change the is limited. This clarity will lead to inclusivity and the
entire business stream. maintenance of balance,” he added.

Public perception of AI By 2025, 30% of outbound marketing

messages from large organisations will be
While some opine generative AI as a threat rather than synthetically generated, up from less than
a boon because of the lack of awareness of the actual 2% in 2022.5
capability of the tool, others are extensively using it and

5.
Gartner Article, January 2023. Beyond ChatGPT: The Future of Generative AI for Enterprises

21
Impact, Opportunity, and Challenges of Generative AI

Is Generative AI ‘creative’ or When an AI system learns from data, it is dubbed as non-

creative, whereas if a similar thing is done by humans, it
‘generative’? is termed as creative. Hence, it is either that the definition
of creativity is blurred or we term one to be less creative
The emergence of creative tools such as Midjourney and than the other.
DALL- E has opened the debate regarding the “creativity”
of these tools. According to Prof Chandra Sekhar of IISc According to the first INDIAai Generative AI Roundtable,
Bangalore, these models are only ‘generative’ as they at a finer level, one needs to distinguish between creative
derive results from existing material and are not creative. and generative AI. While at present, the difference is
In his opinion, creativity is still unique to humans as blurred, and both are used interchangeably, a decade
natural intelligence would be hard to replicate. down the line, this difference will amplify. Defining
creativity in an era of ChatGPT is a task in itself.
“AI is not completely well founded. That is the actual
threat,” Prof Chandra Sekhar noted. Furthermore, self-
learning is still a dream for the models. But, for now, Future of jobs
AI lacks thought; the initial input must still be from the The fear of being replaced by machines in the workspace
human,” he stated in the first roundtable. has always been a major concern whenever a new
Abhinav Aggarwal, whose company Fluid AI recently technology arises. The recent boon in Generative AI
released a book authored by AI, pointed out that even has just amplified this concern. According to many
the “creative” humans are also dependent on other reports, Generative AI could impact up to 300 million
reference materials before creating a particular model. jobs globally. Generative AI could substitute up to 25%
of current work in the US and 24% in Europe while
“Why is it that when a human does it, it is creative and complementing most of the remaining work.
generative when an AI does it?” he asks.
Jyoti Joshi, Founder and CEO of Generative AI startup
To elaborate on this, he added that till now, AI has been Kroop AI, pointed out that, though there are concerns
able to generate something based on the information about AI replacing humans, in the present stage, human
being fed to it. But being creative is something which is intervention is a necessity. “Correct innovation is the key
unique, on which humans have a clear advantage. He to the correct output”, she added.
pointed out that one must distinguish between inference-
based and memory-based logic. At present, AI systems Despite AI bringing in a lot of automation in many
are working on inference-based logic, but a decade processes, the technology will still need to replace jobs.
later, it will be functioning on memory-based logic. It will add to new kinds of job profiles and lead to job

22
Impact, Opportunity, and Challenges of Generative AI

23
Impact, Opportunity, and Challenges of Generative AI

efficiency, concluding the first INDIAai Generative AI potential to replace as much as one-fourth of the current
Roundtable. workload. Alternatively, Generative AI has the potential
to automate around 18% of the total workforce.6 Also,
The insights from the roundtable further backed
an independent study7 on GPTs and their impact on
Goldman Sachs findings that should generative AI
reach its full potential and capabilities, the labour labour market points that 80% of the US workforce
market might experience quite an upheaval. By might have at least 10% of their tasks affected by such
analysing data on occupational tasks in both the large language models. With access to LLMs, 15% of all
United States and Europe, the report came to other work tasks could be expedited while maintaining
the conclusion that approximately two-thirds of the same quality level. Given below are representations
existing jobs are susceptible to various degrees of of job growth and job loss and how Generative AI will
AI automation. Not only that, Generative AI has the transform work across industries.

Total job growth and job loss

Future of Jobs

Total job growth and loss

23% of
today’s job will
change

One million Lost jobs Stable jobs New jobs

Source: World Economic Forum, Future of Jobs Report 2023

6.
 oldman Sachs Economics Research, March 2023. The Potentially Large Effects of Artificial Intelligence on Economic Growth
G
(Briggs/Kodnani)
7.
Tyna Eloundou, Sam Manning, Pamela Mishkin, Daniel Rock, March 2023. GPTs are GPTs: An Early Look at the Labour Market
Impact Potential of Large Language Models

24
 Impact, Opportunity, and Challenges of Generative AI

Generative AI will transform work across industries

Work time distribution by industry
Banking 54% 12% 24% 10% and potential AI impact
Insurance 48% 14% 26% 12% Based on their employment levels in the US
in 2021
Software & Platforms 36% 21% 28% 15% Higher potential for automation

Capital Markets 40% 14% 29% 18% Higher potential for augmentation

Lower potential for augmentation and automation

Energy 43% 9% 14% 34%
Non-language tasks
Communications & Media 33% 13% 21% 33%
Retail 34% 7% 12% 46% 40% of working hours across in-
Industry Average 31% 9% 22% 38%
dustries can be impacted by Large
Language Models (LLMs)
Health 28% 11% 33% 27%
Public Service 30% 9% 35% 26%
Aerospace & Defence 26% 13% 20% 41%
Automotive 30% 6% 13% 50%
High Tech 26% 8% 16% 50%
Trave 38% 6% 15% 50%
Utilities 27% 6% 15% 52%
Life Sciences 25% 8% 17% 50%
Industrial 26% 6% 14% 54%
Computer Goods & Services 24% 6% 13% 57%
Chemicals 24% 5% 14% 56%
Natural Resources 20% 5% 11% 64%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Source: World Economic Forum, May 2023. These are the jobs most likely to be lost - and created - because of AI

25
Impact, Opportunity, and Challenges of Generative AI

Occupations with high proportion of tasks that could be automated by Generative AI8
Occupation Proportion of task that can be automated by Gen AI
Office and administrative support 46%
Legal 44%
Architecture and engineering 37%
Life, physical and social science 36%
Business and financial operations occupations 35%
Occupations with least proportion of tasks that could be automated by Generative AI9
Occupation Proportion of task that can be automated by Gen AI
Building and grounds cleaning and maintenance 1%
Installation, maintenance and repair occupations 4%
Construction and extraction occupations 6%
Education, training and library occupations 7%
Health-care practitioners and technical occupations 8%

However, on the brighter side, AI has the potential 31% respectively. These developments are anticipated
to create jobs as well. The World Economic Forum to generate an additional 2.6 million jobs.10 Another
predicts by 2027, there will be a significant increase in skill that will gain much prominence in the future is
the number of AI and machine learning specialists by prompt engineering. It is a comprehensive process
40%. In other words, 1 million jobs are expected to be encompassing the entire interaction cycle between
created with the increase in usage of AI and machine humans and AI. Prompt engineering aids in smooth,
learning across all industry verticals. Additionally clear and efficient human-AI interaction. The field
roles such as data analysts, data scientists, big data is expected to flourish and continue to evolve and
specialists and information security analysts are also develop as new techniques and demand for advanced
expected to experience an increase of 30-35% and AI systems increase over time.

8.
 lenn Mossy, March 2023. The Impact of Generative AI on Labor Productivity, Employment, Wages and GDP
G
9.
Glenn Mossy, March 2023. The Impact of Generative AI on Labor Productivity, Employment, Wages and GDP
10.
World Economic Forum, Future of Jobs Report 2023

26
 Impact, Opportunity, and Challenges of Generative AI

Fastest growing vs. fastest declining jobs

Top 10 fastest growing jobs Top 10 fastest declining jobs

AI and Machine Learning Specialists Bank Tellers and Related Clerks

Sustainability Specialists Postal Service Clerks

Business Intelligence Analysts Cashiers and Ticket Clerks

Information Security Analysts Data Entry Clerks

Fintech Engineers Administrative and Executive Secretaries

Material-Recording and Stock-Keeping

Data Analysts and Scientists
Clerks

Robotics Engineers Accounting, Bookkeeping and Payroll

Clerks

Electrotechnology Engineers Legislators and Officials

Agricultural Equipment Operators Statistical, Finance and Insurance Clerks

Door-to-Door Sales Workers, News and

Digital Transformation Specialists
Street Vendors, and Related Workers

Source: World Economic Forum, Future of Jobs Report 2023

40% of working hours across the industries can be impacted by Large

Language Models11

11.
World Economic Forum, Future of Jobs Report 2023

27
Impact, Opportunity, and Challenges of Generative AI

28
 Impact, Opportunity, and Challenges of Generative AI

Opportunities

Generative AI has emerged as a powerful tool that has estimated USD 15.7 trillion of potential contribution to
significantly impacted economic opportunities across the global economy by 2030.14
various industries. The potential economic benefits of
generative AI are vast, offering increased productivity, According to Jaspreet Bindra, in technology as
automation, and enhanced decision-making capabilities, fundamental and transformative as generative AI, there
ultimately driving growth, efficiency, and competitiveness would be both sides.
in numerous sectors of the economy.
“In India as a country, I see five possibilities. Firstly, it
Generative AI is expected to increase from USD 11.3
has the potential to make everyone a 10x engineer,
billion in 2023 to USD 51.8 billion by 2028. The CAGR is
which will be a benefit in the technical space.
expected to touch 36% between the forecasted period.12
Secondly, with open APIs and the cloud, it provides
“There are endless opportunities for generative AI in
greater possibilities for entrepreneurship. Thirdly it will
India. Starting from big companies making chatbots to
benefit service companies. Fourthly if we do it right,
students using ChatGPT for writing academic papers.
AI provides huge possibilities for the creative industry,
Even the older generation is curious about ChatGPT,”
and India has a massive creative industry. And finally,
pointed out Pavankumar Dubagunta, Speech Scientist
India stat can be leveraged with the abundant amount
at Uniphore Software Systems.
of data. With AI in the market, we now have a tool to
According to Goldman Sachs estimates, Generative leverage this data”, he added.
AI could raise global GDP by 7% (USD 7 trillion) and
increase productivity by 1.5% in the next decade.13 PwC There are multiple opportunities across the generative
estimates that the openness to tap into the power of AI value chain, with much of the opportunity in its
Generative AI will likely only continue to grow with an applications and services.15

12.
 arkets and Markets Report. Generative AI Market by Offering, Application, Vertical and Region - Global Forecast to 2028
M
13.
Goldman Sachs Article, April 2023. Generative AI could raise global GDP by 7%
14.
PwC Report. Sizing the Price: What’s the real value of AI for your business and how can you capitalise?
15.
Quantum Black, AI by Mckinsey, April 2023. Exploring Opportunities in the Generative AI Value Chain

29
Impact, Opportunity, and Challenges of Generative AI

There are opportunities across the generative AI value chain, but the most significant is building
end-user applications.

Generative AI value chain Opportunity size for new entrants

in next 3-5 years, scale of 1-5
Services
Services around specialized knowledge on how to leverage generative
AI (e.g training, feedback, and reinforcement learning)

Applications
B2B or B2C products that use foundation models either largely as is of
fine-tuned to a particular use case

Model hubs and MLOps

Tools to curate, host, fine-tune, or manage the foundation models (eg,
storefronts between applications and foundation models)

Foundation models

Core models on which generative AI applications can be built

Cloud platforms

Platforms to provide access to computer hardware

Computer hardware

Accelerator chips optimized for training and running the models

Source: McKinsey article Exploring opportunities in the generative AI value chain

Generative AI tools from India

KissanGPT KissanGPT is an AI chatbot that leverages the power of GPT 3.5 and the Whisper
model exclusively for serving India’s underserved agricultural domain. Launched
on March 15th 2023, KissanGPT has already earned the hearts of farmers all
over the country with its remarkable ability to guide the farmers in irrigation,
pest control and crop cultivation.
PolicyGPT Plum, a Bengaluru-based startup, has introduced PolicyGPT to transform the
insurance industry. The purpose of the GPT-3-based chatbot is to educate
consumers on their health insurance policies. In addition, the chatbot is
designed to reduce the complexity of insurance policies by answering queries
about coverage and elucidating inclusions and exclusions.
GitaGPT A Google India software developer named Sukuru Sai Vineet developed the AI
chatbot called GitaGPT. It uses GPT-3 technology and the Bhagavad Gita to
offer answers to life’s problems. Users can ask questions on the GitaGPT app,
and a chatbot will respond by researching the Bhagavad Gita’s teachings.

30
 Impact, Opportunity, and Challenges of Generative AI

BharatGPT Bengaluru-based conversational AI platform CoRover recently released

BharatGPT. In comparison, OpenAI’s ChatGPT supports only 95 languages
and mainly interprets English instructions. In addition to text, CoRover’s
chatbot can also process rich data types, including photos, audio, video, and
maps, which is not the case with ChatGPT. While the accuracy of ChatGPT has
yet to be determined, CoRover asserts that BharatGPT is 90% accurate.
Lexi Lexi is an AI chatbot created by Velocity, a financial startup powered by
ChatGPT. Lexi seeks to help e-commerce entrepreneurs by simplifying business
analytics. Lexi, integrated with Velocity’s analytics product, Velocity Insights,
assists organisations in tracking market spending, sales, and more and
provides daily business reports over WhatsApp. In addition, customers may
now utilise Lexi, the AI chatbot, to get answers to their concerns, increasing
their company operations, thanks to the integration of ChatGPT in the same
WhatsApp interface. Velocity announced this in an official blog post.
Jugalbandi Jugalbandi is a free and open platform that combines the power of ChatGPT
and Indian language translation models under the Bhashini mission to power
conversational AI solutions in any domain.

When it comes to talent, presently, much of the AI talent even have to learn Python. You just have to speak your
works in the US, India, UK 60% of the Generative AI language. And by communicating to this computer
talent pool comes from the US, India, UK, Germany, what you need, what you want, what problems you
Canada and France. Yet, in India, there is a rush to find want to be solved, this computer will write the software
talent in AI despite around 416,000 people working in AI by itself.” He theorises that everyone is a computer
and Data Science. At present, India still faces a demand programmer now.
of approximately 629,000 AI talent.16 According to
NASSCOM, there is still an additional demand for “This is going to have the greatest opportunity for us to
One million workers in the space, which is expected to democratise this very powerful instrument we call the
increase exponentially in the coming years.17 computer for the very first time in history,” he said.

“I believe it will lift so many segments of society. It will

Bridging the digital divide bring great education to people who don’t have access
According to Jensen Huang, Generative AI can to education. It’s the most powerful democratisation
positively impact countries like India and is one of the force I’ve ever seen,” he concluded.18
most fantastic opportunities we have ever had to close
When it comes to bridging divides, linguistic barriers
and bring together the social and technology divide.
have plagued the country for many decades, and
For the last 30-40 years, only so many people know Generative AI tools built alongside the Government of
how to program a computer. So the number of people India’s Bhashini program can be a valuable asset in the
who know how to use this incredible instrument for the future.
benefit of themselves or their business or their country is
really quite limited. “Creating content in multiple languages is an issue for
India. This shortcoming can be tackled with Generative
“And yet, all of a sudden, there’s a new type of computer,
AI,” says Harsha Mundhada of Inflexor Venture.
this new type of computer, you don’t have to learn C,
C++, you don’t have to learn Pascal, you don’t have to The language translation and other Generative AI
learn Fortran, you don’t have to learn Java. You don’t capabilities will make India’s apex judicial bodies more

16.
 ASSCOM study, February 2023. State of Data Science & AI skills in India - Data and the Art of Smart Intelligence
N
17.
NASSCOM study, February 2023, State of Data Science & AI skills in India - Data and the Art of Smart Intelligence
18.
INDIAai article, March 2023. Generative AI can help India close the technology divide, NVIDIA CEO Jensen Huang

31
Impact, Opportunity, and Challenges of Generative AI

accessible to people across the country, she added. aptitude.

Furthermore, many believe that Generative AI tools are “AI is becoming the supportive tutor providing
personalised teaching plans for students. I see
a blessing for many people with disabilities. According
applications which are catering towards that, be it in
to Pavankumar Dubagunta, Speech Scientist, Uniphore one-on-one skill development training to something as
Software Systems,” people who have deficiencies in simple as fluency improvements,” said Neethu Mariam
speaking and autism can leverage technologies like AI, Joy, who is currently working on a new Generative AI
startup that focuses on EdTech.
and I think it will be a life-changing experience for them.”
“Not just in EdTech, in healthcare, it is simple to build
Generative AI’s impact on a query-based system where AI takes the role of
different sectors proxy doctors. Most of the applications are bringing
personalization into whichever sector you are applying
The economic opportunities brought about by generative it into,” she added.
AI are vast and diverse. It empowers industries to Generative AI could revolutionise healthcare by helping
leverage data-driven insights, enhance productivity, doctors analyse medical data, diagnose patients, and
and drive innovation. As technology advances, we customise treatment strategies. Recently, researchers
can expect further transformations across sectors, from Drexel University’s School of Biomedical
paving the way for a more efficient, personalised, and Engineering, Science and Health Systems recently
demonstrated that Open AI’s GPT-3 program could
technologically advanced future. Some of the significant
identify clues from spontaneous speech that are 80%
impact sectors for Generative AI include healthcare, accurate in predicting early stages of dementia.19
fintech and education, with the education sector already
being transformed into Generative AI. Since there is still no cure for the disease, diagnosing
it early can give patients more options for therapeutics
Generative AI can play the role of personalised teachers and support. The current practice for diagnosis involves
in the education sector. It can provide customised medical history review and a lengthy set of physical and
modules and classes for students based on their neurological evaluations and tests.

19.
Drexel University Research, December 2022. AI behind ChatGPT could help spot early signs of Alzheimer’s disease

32
 Impact, Opportunity, and Challenges of Generative AI

Generative AI impact on different sectors in India

Sector Applications Tools

Agriculture • Crop Management- Generative AI systems can find patterns and trends • Ama KrushAI
in satellite imagery, weather data, and soil composition to help farmers • Farmer.chat
decide when and where to sow, water, and harvest crops. • Apurva.ai
• Crop Optimisation - Generative AI can also help farmers choose the best • Kissan GPT
crops and production methods for their needs.
• jiva ag
• Resource Management- Generative AI can assist farmers in maximising
water, fertiliser, and pesticide use, boosting sustainable farming.
• Pest control - Generative AI can improve pest control by assessing pest
populations, weather, and crop health to predict outbreaks and offer
focused interventions.
• Supply Chain Optimisation - Generative AI helps farmers and
agribusinesses assess massive logistical data, detect bottlenecks and
inefficiencies, and enhance supply chains.
• Climate-Resilient Agriculture - Generative AI can analyse climatic trends,
soil health, and crop performance to build more resilient agricultural
practices.
• Crop Innovation - Generative AI can help researchers and plant breeders
find and develop novel crop types with improved performance, yield, and
pest and disease resistance.
• Farmer education - Generative AI could transform farmers’ information
availability and decision-making.
Education • Course Design - Generative AI techniques can organize syllabi, lesson • NOLEJ
plans, and exams. Practice problems and interactive exercises can be • TutorAI
tailored to students’ knowledge gaps, skills, and learning methods. • Snapxam
• Personalized Lessons - Personalized lesson plans help students get the • QANDA
best education based on their needs and interests.
• tabnine
• Content Creation for Courses - Generative AI can help create quizzes,
exercises, and concept explanations. It can help teachers who need to
develop a lot of stuff. AI can transform original content into new content.
• Restoring Old Learning Materials - Generative AI can enhance low-
quality learning materials, including historical documents, photos, and
films. AI can improve the resolution of these items, making them more
interesting for kids acclimated to high-quality media.
• Tutoring - Students can connect with a virtual tutor and receive real-time
feedback and guidance using generative AI. It can aid pupils without in-
person tutoring.
Healthcare • Diagnosis and Screening - Generative AI helps clinicians identify and • MedPaLM
treat patients faster and more accurately, improving patient outcomes. • Viz.ai
• Personalized Medicine - Generative AI algorithms can analyze enormous • Enlitic
medical datasets to find patterns, predict outcomes, and improve health. • Merative
These personalized medicine methods can help doctors create more
effective treatment regimens and follow-up care. • CloudMedX
• Drug Discovery - Generative AI algorithms can analyze clinical trials
and other data to find potential medication targets and predict the most
successful molecules. New medications could be developed faster and
cheaper.
• Creating Research Ideas - In healthcare, generative AI can also be
utilized to generate new concepts.
• Avoiding Medical Errors - Generative AI can automatically repair
spelling problems in the documentation, which is helpful for electronic
prescriptions, and populate the system with the relevant data.

33
Impact, Opportunity, and Challenges of Generative AI

Sector Applications Tools

Manufacturing • Product Development and Design - Generative AI can accelerate the • Autodesk
development of new design concepts, pharmaceuticals, materials, and • Fanuc
market research, reducing time to market. • Birlasoft
• Quality Control - Predictive AI systems monitor and report production • FortiSandbox
concerns in real-time, whereas generative AI systems find and alert
against faults. • Altexsoft
• Supply Chain Management - Generative AI may eliminate the necessity
for an integration platform for interoperability. As original equipment
tools evolve, continuous integration platform changes are unnecessary.
• Customer Interactions and Support - Generative AI systems can provide
24/7 customer service, reducing staff and increasing productivity.
Retail • Personalised Product Recommendations - Generative AI can analyze • vertex ai vision
customer data to create tailored journeys, customized discounts, and • Discovery AI
engaging content for specific shoppers. • Market360
• Manage Inventory Levels - Generative AI analyses sales data and • Scanunlimited
suggests inventory management. AI can help retailers optimize supply
chain and delivery by aRetailng historical data, consumer sentiment, • numerator
and competition data to predict trends and make smart purchasing and
production decisions.
• Monitor And Optimise Prices - Generative AI algorithms can quickly
optimise retail prices by analyzing rival prices, demand patterns, and
market trends.
• Build Customer Service Chatbots - Generative AI can help customers
solve problems with chatbots. As a result, it can help retailers improve
customer service and reduce customer support agent workload.
• Detecting Fraudulent Activities - Generative AI systems can prevent
fraudulent purchases and returns, saving companies money and building
customer confidence.
Media20 • Automated Content generation: LLM can be used to generate content, • Lately
such as articles, blog posts, or social media posts. This will be beneficial • Sprout Social
for those who want to create content on a regular basis. For example, • HubSpot
social media influencers,
• Copy.ai
• Variety and improved content: LLMs can create a variety of interesting
content that caters to different audience sections. The quality of content • Emplifi
will be much better as AI models learn from a large amount of data.
• Personalised content: AI models can generate personalized content based
on the preferences of individual users
• Customer management: Generative AI can help in customer
management, providing timely responses and engagement with a digital
audience. ChatGPT, a form of Generative AI, can be used to answer
more general questions. Off late, there are Google Chrome extensions
that allow brands and Twitter users to install Chat GPT bots that respond
to all Tweets
• Localization of content using deep fakes: The OTT platforms have
increased the need to create content in local/regional languages. For
example, the task of dubbing and content moderation for lip-syncing with
an on-screen character can be smoothly accomplished using deep fake
technology. The technology superimposes voice from a source to target
face with a perfect lip-sync. This is achieved through face synthesis.

20.
INDIAai Knowledge Asset, March 2023. How Generative AI will influence media industry

34
 Impact, Opportunity, and Challenges of Generative AI

Sector Applications Tools

• Upscaling content through super-resolution: Super-resolution is a method
of using ML algorithms to infer form an original image (with fixed pixels)
and create another version of the same image but with higher pixels.
Super resolution improves the display of images and video. Such super
resolution is mostly required by documentaries focusing on nature,
wildlife or even sports.
• Upscaling Visual Effects: The rise in demand for animation and VFX
content has powered immersive experiences, with exponential growth
in AR & VR technology that has Media the entertainment market. For
example, AI can help in superimposing/transposing a source face to a
target face with much ease.
Marketing21 • Cost-saving and scheduling – Voice cloning via synthetic media can cut • Jasper AI
down on the time needed to chase busy voice actors. To put it simply, if • MarketMuse
you have Awkwafina voicing a character in your animated film, you can • Acrolinx
capture her voice sample and generate the lines.
• Semrush
• Custom regional accents – Rather than selecting just one human
character, advertisers could generate hundreds or thousands of synthetic • Seventh Sense
characters to appeal to narrow demographic bases. This means,
instead of having one celebrity extol the virtues of a toothpaste, different
characters can appeal separately to college-going teens, stay-at-home
dads, and potential trendsetters.
• Reaching people in their language – Using AI synthesis, one can create
David Beckham’s video about malaria in Gujarati or Jackie Chan’s
video about wearing a mask in Tamil. This can create an opportunity for
advertisers to reach out to people in their own languages.
• Archiving people around – Advancements in synthetic media can also
let us preserve ourselves as well as our loved ones. You might be able to
ask questions to a 5-year-old you or listen to your grandmother sing long
after she’s gone.

Next frontiers for Generative into the security protocols followed currently. Unless
and until the FinTech sector finds the means to tackle
AI this possibility, the probability of sectoral disruption is
strong.
According to experts, one of the sectors expected to be
disrupted by Generative AI in coming years is Fintech.22 According to Jaspreet Bindra, “The sector which would
be disrupted is the one which created it- the tech (IT)
One of the sectors which will face a huge impact
sector,” “Generative AI has been writing codes involving
due to Generative AI will be FinTech, says Aveekshith
Bushan, Vice President and General Manager for APJ basic process automation and many more. I recently
at Aerospike. “ If there is fraud in the FinTech sector, it read a tweet by a programmer who used ChatGPT for
will affect thousands of users,” he added. He believes the first time for writing codes. He said the value of 90%
that at some point, someone will find a way to break of his skills had gone down to zero dollars”, he added.

21.
INDIAai article, November 2021. Generative AI & Synthetic Media: A New Era of Advertisements
22.
Insights from INDIAai Roundtable Discussion What is India’s Opportunity in Generative AI, May 2023

35
Impact, Opportunity, and Challenges of Generative AI

Challenges

While generative artificial intelligence (AI) holds “In markets like India, content can go out and spread
tremendous potential and has made significant quickly,” says Aveekshith Bushan. “There could be
advancements in various fields, it also presents unique biases and misinformation. We have already seen it on
challenges. The complexity and power of generative AI the other side of ChatGPT, where the output is irrelevant
algorithms can give rise to ethical, legal, and societal to the content searched. People will have to decipher
concerns. and say what looks relevant and what does not,” he
added.
According to Kazim Rizvi, Founding Director of a tech
policy think tank called The Dialogue, “Generative For many years, we have seen Generative AI
AI has inspired many to gain first-hand experience techniques, such as deep fakes, being used to bring
with AI in recent times. However, an opaqueness in social and political unrest across the globe. The recent
data collection, developers of the models, and biases developments in Generative AI technology can make
being fed into the system still exists. This ambiguity will these kinds of threats more prevalent.
continue to remain even if the technology is evolved.”
“It was challenging to analyze the impact of the
One of the major concerns that Generative AI possesses previous AI revolution, which was algorithms and
at present is its potential to amplify the spread of information shared. It is also complicated to foresee
misinformation that can have varying impacts, especially the implications of deep fakes. This is a threat to the
when these language models can often produce democracies of the world which is hard to tackle,” says
outputs that are not real. According to researchers, this
Anna Danes, Data Ethicist who works with organizations
phenomenon of AI generating confident responses that
to create frameworks and practical methods to ensure
do not seem to be real or justified by its training data is
responsible developments in AI.
called hallucination.23

23.
 iwei Ji, Nayeon Lee, Rita Frieske, Tiezheng Yu, Dan Su, Yan Xu, Etsuko Ishii, Ye Jin Bang, Andrea Madotto, Pascale Fung, March
Z
2023. Survey of Hallucination in Natural Language Generation

36
 Impact, Opportunity, and Challenges of Generative AI

Another area of concern regarding the widespread use leaders with a Western appearance, such as Angela
of Generative AI models is the different kinds of bias the Merkel, Emmanuel Macron, and Donald Trump. None
results produced by these models show. of the popular Asian leaders were included in the result.

According to Divya Dwivedi, a lawyer at the Supreme These demonstrations have only strengthened the belief
Court of India, who focuses on tech-law, “We really do that these models amplify human biases.
not know what kind of data it will generate and what kind
of bias it will come up with since we already have made Ethics, rights, and intellectual
these models very biassed as a result of our own nature.”
property
Bias in AI models has been a central point of debate
for many years, and Generative AI models have just The use of copyrighted materials for training AI models
amplified these biases. is still unclear, and hence is said to be in a legal grey
area. It can be said that there are no copyright laws so
“There is an inherent bias that is there when we are far that would safeguard any wholly AI generated model
training the dataset that comes out of how humans or creation. It becomes irrelevant whether that creation
have kind of embedded their prejudices in the system,” stemmed from a human-crafted text prompt. While fair
points out Prateek Sibal, Programme Specialist for AI, use laws permit the use of copyrighted material under
Emerging Tech and Internet Governance at UNESCO. certain conditions without the owner’s permission, the
ongoing legal disputes could disrupt this status quo and
Recently, India Today conducted a political experiment bring uncertainty in the future of AI model training.
involving leading platforms that use AI to generate
visual imagery, suggesting that these platforms could be Undoubtedly, the advent of generative AI has
biased in terms of their knowledge and understanding revolutionized our lifestyle, labor practices, and
of various nations. artistry output within a mere few months. In turn, the
inundation of AI-fabricated written works, pictures, and
The Generative AI tool Midjourney was asked to create tunes, alongside the mechanisms through which they
pictures of “most popular elected political leaders posing were created, has stimulated a plethora of intricate
in front of the Eiffel Tower in 2023.” The prompt was legal inquiries. These challenge our understanding
generic, without mentioning names to check the scope of ownership, fairness, and the core foundation of
of the results. The AI-generated image only included innovation, writes Anndy Lian.24 Anndy Lian is an all-

24
INDIAai article, May 2023. The Legal Complications of AI-generated Content in Copyright Law

37
Impact, Opportunity, and Challenges of Generative AI

rounded business strategist in Asia and the author of According to Jaspreet Bindra, Founder and MD of Tech
the book NFT From Zero to Hero. Whisperer Ltd. “The biggest thing we will have to figure
out is how to control this monster. The democratization of
“There are papers and research where Generative technology is the biggest challenge here. The closest I can
AI or ChatGPT has been mentioned as co-authors, think about something we came across like AI was nuclear
what does this mean for original scientific work? ” power. Countries shall come together to regulate it.”
asks Prateek Sibal. Prateek is Programme Specialist,
Digital Innovation and Transformation at UNESCO On the other hand, people like Deepak Visweswaraiah,
who is working to understand the impact of digital Vice President, Platform Engineering and Site Managing
technologies, specially artificial intelligence (AI), on Director at Pegasus Systems, point out that “Controlling
societies from a human rights, openness, inclusive the monster does not happen due to government
access and multi-stakeholder governance perspective. regulations. It should be a collective effort. How do
we teach developers to be mindful of AI ethics? The
Copyright laws play a fundamental role when it comes developers should be clear about the purpose of each
to protecting intellectual property and encouraging model and share awareness to avoid misuse.”
creativity. It allows creators to rightfully control the
usage, distribution, and adaptation of their work. The
law encourages creators to create more by offering Regulation25
them exclusive rights. Creative Commons licenses When it comes to regulation, countries will not have
enable the free distribution of an otherwise copyrighted uniform reasons to regulate Generative AI. For
“work”. They also provide more options for creators to example, China’s draft regulations are geared towards
choose the level of protection they want for their work. preventing further developments in AI technology
With advancements in AI technology, there is a that might undermine the government’s control over
simultaneous increase in its usage in the creative domestic internet and tech space. On the other hand,
process. With the increase in AI capabilities to generate the EU places the prevention of harm to individuals
fresh and original content there is a growing need front and center in its draft for the AI Act. Therefore,
for a legal framework that addresses the copyright the purpose of regulations must be defined to ascertain
protection of collaborative works involving AI.This legal what the regulations should contain.
framework is highly important to strike a balance At a foundational level, all generative AI regulations
between safeguarding the rights of creators while should attempt to protect individuals against potential
boosting innovation and originality. harm. Harms that could include violating an individual’s
This is difficult to predict how copyright law will shape privacy and data rights, discrimination in access
up around AI-generated work. However, it is clear that to services, or being subject to false or misleading
the legal framework governing copyright protection will news and information. Protection against these, and
undergo significant transformation. similar harms, has to be non-negotiable. There is an
international consensus on the necessity to ensure such
protections, though the granular details of practical
How to ensure responsible implementation still need to be clarified.
use of Generative AI? “When we are talking about regulation, what we are
With the rise in the usage of generative AI, there is a dire trying to do is regulate the impact of technology on
need for a holistic,and a comprehensive framework that society. And sometimes we just don’t know what it is
would ensure responsible usage of the technology to going to be, and you don’t want to make any regulation
harness the best benefits while mitigating its risks. This because probably it’s too soon”, says Prateek Sibal.
framework shall include ethical guidelines, regulatory
Of late, there has been a significant debate on whether
measures, transparency, accountability, and ongoing
regulations should also include protection from second-
collaboration among stakeholders. By boosting fair
practices we can nurture the technology, increase order harms, such as violation of intellectual property
people’s trust in it, protect it against potential harm, rights (IPR) and defamation. While this has not yet been
and ensure that generative AI serves humankind and settled in a definitive manner, generative AI systems will
society in more productive ways. likely be subject to at least IPR laws soon.

25
INDIAai article, April, 2023. Analyzing AI regulations based on dynamics in the AI ecosystem

38
 Impact, Opportunity, and Challenges of Generative AI

Apart from individual harms, regulations could also about the impact of what they are developing. I see
look at systemic harms, specifically the increased people from engineering teams saying I’m also partly
concentration of economic and market power in accountable for what I’m doing. This shows that people
the hands of Big Tech players. The fundamental are becoming more aware and vigilant.” says Anna
requirements for developing generative AI systems - Danes, a Digital Ethicist who works with organizations
access to substantial amounts of data and computing to create frameworks and practical methods to ensure
power to process this data - are readily available only responsible developments in AI.
to the Big Tech companies.
She also believes that, “we all start thinking a little bit
“The regulation of AI tends to focus on formal equality about what incentives we can give companies, private
rather than substantial equality, even though the end companies, and also public institutions to awaken their
goal is often equity,” points out Shashank Reddy, ethical knowledge or imagination so that they can
Managing Partner at Evam Law & Policy. develop ethical products”.

Algorithmic Auditing is one of the tools put forward to AI regulation efforts across
mitigate Generative AI risk.
the globe
“Auditing AI systems is a mechanism or criteria for the
system to comply with legal obligations. Such legal The capabilities of the generative AI models
obligations could be a hard law obligation, which may have been a matter of concern for governments,
be an existing law, or it could just be just a company researchers, and the common man worldwide. As
that is aware of the AI act or some other law and is a result, governments worldwide have imposed AI
looking to be compliant with it”, says Vibhav Mithal, of regulations. The European Union’s AI Act, Canada’s
Anand and Anand and Future of Humanity. Artificial Intelligence and Data Act (AIDA), the United
States’s AI Bill of Rights and State Initiatives, and
In order to narrow down to the exact entity that is causing China’s Algorithm Transparency and Promoting
the harm, it is very critical to examine the generative AI Industry Development have been a subject of
AI value chain closely. Commercial Generative AI discussion in several forums.
system has four pillars: Developers, who develop the
system; Deployers, who create and work on the base Each of these country’s AI regulations are eceptionally
mode for advanced functionalities for themselves or thorough and have great visions around the usage and
other third-party customers; Users, who are individuals, protection of AI. While the AI Act concentrated on a risk-
corporate organizations, or platforms who use the AI based approach to guide the use of AI in both the private
system themselves either internally or through product and public sectors, the AI Bill of Rights targets specific
offerings; and ultimately the Recipients, people who use cases. Though the Chinese government has yet to
receive and use the output of the AI system. pass rules on AI technology at large, recently, the country
introduced a law that regulates how private companies
“I see a lot of engineering teams being very skeptical use online algorithms for consumer marketing.

39
Impact, Opportunity, and Challenges of Generative AI

Name of the country Description

EU • The EU regulation places the prevention of harm to individuals front and
center in its draft AI Act.
• According to the proposal, these AI tools will be classified under their
perceived risk level, from minimal to limited, high, and unacceptable.
• According to the proposal, these AI tools will be classified under their
perceived risk level, from minimal to limited, high and unacceptable.
• The act focuses on biometric surveillance, spreading misinformation or
discriminatory languages.
• The high-risk tools are not expected to be banned but are expected to be
highly transparent in their operations.
US • The Biden-Harris Administration has announced new actions that will
further promote responsible American innovation in Artificial Intelligence,
considering the people’s rights and safety.
• prioritize people and communities by enhancing responsible innovation
that serves the public good, balancing it with protecting society, security
and the economy.
• companies have a fundamental responsibility to ensure their offered
products are safe to use before they get deployed.
• The administration released a Blueprint for an AI Bill of Right and related
executive actions, the AI Risk Management Framework and a roadmap for
standing up a National AI Research Resource made public earlier this year.
China26 • The Cyberspace Administration of China, the country’s top internet
watchdog, recently passed a regulation on “deep synthesis” technology,
which it defines as “technology that uses deep learning, virtual reality, and
other synthesis algorithms to generate text, images, audio, video, and
virtual scenes.”
• Users are prohibited from using generative Ai to engage in activities that
endanger national security, damage public interest, or are illegal
• Anonymity doesn’t exist in the Chinese internet
• Wants to censor what algorithms can generate
• Service providers must audit AI-generated content and user prompts
manually or through technical means
• Baidu, one of the first to launch a Chinese text-to-image model, already
filters politically sensitive content.
• The regulation bans people from using deep synthesis tech to generate
and disseminate fake news
• When the data used for AI training contains personal information,
technology providers should follow the country’s personal information
protection law.
• Platforms should also remind users to seek approval before they alter
others’ faces and voices using deep synthesis technology.
• If the result of generative AI may cause confusion or misidentification
among the public, the service provider should put a watermark in a
prominent place to inform the public that it is a work by the machine.

23.
 inistry of Industry & Information Technology, Government of China, Order No. 12, November 2022. Provisions on the
M
Administration of Deep Synthesis of Internet Information Services.

40
 Impact, Opportunity, and Challenges of Generative AI

Name of the country Description

UK27 • The UK Government published28 a White Paper entitled “A pro-innovation
approach to AI regulation”
• The White Paper elaborates on the approach to AI set out by the
Government in its 2022 AI Governance and Regulation Policy Statement
• The white paper identifies five major principles:
™ The AI systems should function in a Safe, Secure and Robust manner
™ Transparency and Explainability of organizations developing and
deploying AI to communicate about the purpose of AI systems.
™ The AI systems should be fair and devoid of any biases and
discrimination against the users
™ Effective oversight is necessary for the functioning of AI systems
™ Contestability and redress
• The white paper recognizes risks with a de-centralized regulatory
framework, including inconsistent enforcement or guidance across
regulators.
• Currently, the paper is waiting for feedback until 21st June 2023
Canada29 • The Government of Canada has issued The Artificial Intelligence and Data
Act (AIDA)
• AIDA is an important milestone in implementing the Digital Charter and
ensuring that Canadians can trust the digital technologies that they use
every day
• The framework proposed in the AIDA is the first step towards a new
regulatory system designed to guide AI innovation in a positive direction,
and to encourage the responsible adoption of AI technologies by
Canadians and Canadian businesses.
• AIDA would ensure that high-impact AI systems meet the same
expectations with respect to safety and human rights to which Canadians
are accustomed
• The Minister of Innovation, Science, and Industry would be empowered
to administer and enforce the Act, to ensure that policy and enforcement
move together as the technology evolves.
• Prohibit reckless and malicious uses of AI that cause serious harm to
Canadians and their interests
• The obligations for high-impact AI systems is guided by the principles of
human oversight and monitoring, transparency, fairness and equity, safety,
accountability, validity and robustness

27.
 overnment of China News, May 2023. Xi Jinping replied to the letter and encouraged the teachers and students of Macau
G
University of Science and Technology
28.
Marianna Drake, Jasmine Agyekum, Marty Hansen, Lisa Peets & Mark Young, April 2023. UK Government Adopts a “Pro-
Innovation” Approach to AI Regulation
29.
Government of Canada. The Artificial Intelligence and Data Act (AIDA) - Companion Document

41
Impact, Opportunity, and Challenges of Generative AI

42
 Impact, Opportunity, and Challenges of Generative AI

Recommendations from the roundtables

1. Foster Education and Public Awareness: It is AI as a tool to enhance human capabilities and
crucial to educate the public about generative improve productivity across various sectors.
AI to dispel fears of job replacement and 7. 
Foster Collaboration and Interdisciplinary
promote understanding of its capabilities. Approaches: The development of generative
This can be achieved through awareness AI models involves multiple stakeholders.
campaigns, workshops, and educational Encourage collaboration among researchers,
programs. developers, policymakers, and domain
2. Establish Data Sharing and Usage Regulations: experts to foster interdisciplinary approaches
Given the importance of data quality in building and ensure a comprehensive understanding
robust generative AI models, there is a need of the societal impacts and potential risks of
for regulations to govern data sharing and generative AI.
usage. This can help protect privacy, ensure 8. 
Address Ethical Concerns: Given the
ethical practices, and maintain transparency in intellectual capacity of generative AI, it is
AI development. crucial to address ethical concerns and
3. 
Develop Global Standards and Regulatory potential misuse. Developers should consider
Frameworks: The lack of global standards for the ethical implications of their models and
AI poses a significant challenge. Governments proactively work towards preventing any harm
and international organizations should or unintended consequences.
collaborate to establish common frameworks 9. 
Support Accessibility and Inclusivity:
that address the regulatory aspects of Generative AI can have a positive impact on
generative AI, including accountability, bias individuals with communication difficulties or
mitigation, and safety measures. disabilities. Encourage the development of
4. 
Encourage Self-regulation: In addition inclusive AI solutions that cater to the needs of
to governmental regulations, individuals speech-impaired individuals, autistic people,
and corporations involved in generative AI and others who may benefit from improved
development should adopt self-regulatory communication tools.
practices. This includes being clear about the 10. Emphasize Safety and Control: As generative
purpose of each model, promoting ethical AI continues to advance, safety and control
guidelines, and fostering responsible use of AI measures must be prioritized. Governments,
technologies. organizations, and developers should
5. 
Prioritize Bias Mitigation: While it may be collaborate to establish safeguards that prevent
difficult to completely eliminate bias, it is malicious use or unintended consequences,
important to prioritize efforts to mitigate bias ensuring that the technology is harnessed for
in generative AI models. Developers should the benefit of society.
invest in research and development of bias 11. 
Monitor Future Enhancements: Stay vigilant
detection and mitigation techniques to ensure about future enhancements to deep learning
fair and unbiased outcomes. models that require less training. It is important
6. Explore Augmentation of Human Intelligence: to carefully assess the implications of these
Generative AI tools have the potential to advancements, including potential risks
augment human intelligence. Encourage and unintended consequences, to ensure
research and development in using generative responsible and safe deployment of generative
AI technologies.

43
Impact, Opportunity, and Challenges of Generative AI

References

Arianna, Johnson. 2023. “Which Jobs Will AI Replace? These 4 Industries Will Be Heavily Impacted.” Forbes.
https://www.forbes.com/sites/ariannajohnson/2023/03/30/which-jobs-will-ai-replace-these-4-industries-will-be-
heavily-impacted/?sh=344667415957.

Cao, Yihan. 2023. “A Comprehensive Survey of AI-Generated Content (AIGC): A History of Generative AI from
GAN to ChatGPT.” arxiv 1, no. 2023 (March): 8.

Digital, Mckinsey. 2023. “What is generative AI?” McKinsey and company. https://www.mckinsey.com/featured-
insights/mckinsey-explainers/what-is-generative-ai.

Digital, Mckinsey. 2023. “What every CEO should know about generative AI.” Mckinsey Digital. https://www.
mckinsey.com/capabilities/mckinsey-digital/our-insights/what-every-ceo-should-know-about-generative-ai.

Drake, Marianna. 2023. “UK Government Adopts a “Pro-Innovation” Approach to AI Regulation.” Covington.
https://www.insideprivacy.com/artificial-intelligence/uk-government-adopts-a-pro-innovation-approach-to-ai-
regulation/.

Elias, Jibu. 2023. “Generative AI can help India close the technology divide, NVIDIA CEO Jensen Huang.” INDIAai.
https://indiaai.gov.in/article/generative-ai-can-help-india-close-the-technology-divide-nvidia-ceo-jensen-huang.

Eloundou, Tyna. 2023. “GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language
Models.” arxiv 4, no. 20 (March).

Enterprise, IT. 2023. “AI will drive 83 million “structural” job cuts in 5 years says WEF.” The Stack. https://thestack.
technology/ai-job-losses-wef-future-of-jobs-report/.

Gazala, Nawal. 2021. “Generative AI & Synthetic Media: A new era of advertisements.” INDIAai. https://indiaai.
gov.in/article/generative-ai-synthetic-media-a-new-era-of-advertisements.

Government, Canada. 2022. “The Artificial Intelligence and Data Act (AIDA) – Companion document.” Government
of Canada. https://ised-isde.canada.ca/site/innovation-better-canada/en/artificial-intelligence-and-data-act-
aida-companion-document.

JALAN, AYUSH. 2023. “7 Ways Generative AI Will Change the Job Market.” Make Use Of. https://www.makeuseof.
com/ways-ai-change-job-market/.

Jeevanandam, Nivash. 2022. “Dall-E - the AI that creates images of everything.” INDIAai. https://indiaai.gov.in/
article/dall-e-the-ai-that-creates-images-of-everything.

Jeevanandam, Nivash. 2023. “Exploring Amazon Bedrock language model.” INDIAai. https://indiaai.gov.in/
article/exploring-amazon-bedrock-language-model.

Jeevanandam, Nivash. 2023. “Exploring Firefly - a generative AI model from Adobe.” INDIAai. https://indiaai.gov.
in/article/exploring-firefly-a-generative-ai-model-from-adobe.

Ji, Ziwei. 2023. “Survey of Hallucination in Natural Language Generation.” ACM Computing Surveys 55, no. 12
(March): 38.

Justice, Cliff. 2023. “Is AI going to slash jobs?” KPMG. https://info.kpmg.us/news-perspectives/technology-

innovation/fact-or-fiction-episode-six.html.

44
 Impact, Opportunity, and Challenges of Generative AI

KPMG, team. 2023. “https://info.kpmg.us/news-perspectives/technology-innovation/fact-or-fiction-episode-six.

html.” KPMG. https://assets.kpmg.com/content/dam/kpmg/tr/pdf/2021/02/living-in-ai-world.pdf.

Lian, Anndy. 2023. “The Legal Implications of AI-Generated Content in Copyright Law.” INDIAai. https://indiaai.
gov.in/article/the-legal-implications-of-ai-generated-content-in-copyright-law.

Mdaka, Yamkela. 2023. “Jobs that are most exposed to generative AI, according to new study.” News24. https://
www.news24.com/news24/tech-and-trends/jobs-that-are-most-exposed-to-generative-ai-according-to-new-
study-20230329.

Milin, Stanley. 2023. “Everything you need to know about Prompt Engineering, one of the hottest jobs in Generative
AI.” INDIAai. https://indiaai.gov.in/article/everything-you-need-to-know-about-prompt-engineering-one-of-the-
hottest-jobs-in-generative-ai.

Mossy, Glenn. 2023. “The Impact of Generative AI on Labor Productivity, Employment, Wages and GDP.” Linkedin.
https://www.linkedin.com/pulse/impact-generative-ai-labor-productivity-employment-wages-glenn-mossy/.

nasscom, insights. 2023. “10 BEST AND HIGH-PAYING AI JOBS FOR 2023.” nasscom insights. https://community.
nasscom.in/communities/data-science-ai-community/10-best-and-high-paying-ai-jobs-2023.

nasscom, team. 2023. “State Of Data Science & AI Skills In India – Data And The Art Of Smart Intelligence.”
nasscom. https://nasscom.in/knowledge-center/publications/state-data-science-ai-skills-india-data-and-art-
smart-intelligence.

news, nvidia. 2023. “What is Generative AI?” nvidia. https://www.nvidia.com/en-us/glossary/data-science/

generative-ai/.

Raja, Anjali. 2023. “Analyzing AI regulations based on dynamics in the AI ecosystem.” INDIAai. https://indiaai.gov.
in/article/analyzing-ai-regulations-based-on-dynamics-in-the-ai-ecosystem.

research, INDIAai. 2023. “How Generative AI will influence media industry.” INDIAai. https://indiaai.gov.in/
research-reports/how-generative-ai-will-influence-media-industry.

Secretary, Parliament. 2023. “A pro-innovation approach to AI regulation.” Open Government Licence 1, no. 12
(March): 8.

Shanthi, S. 2023. “Can India’s Startups Tap Into Generative AI Opportunity?” Entrepreneur India. https://www.
entrepreneur.com/en-in/technology/can-indias-startups-tap-into-generative-ai-opportunity/451004.

Singh, Manish. 2023. “Where is India in the generative AI race?” Tech Crunch. https://techcrunch.com/2023/05/03/
where-is-india-in-the-generative-ai-race/.

team, Drexel. 2022. “Study: AI Behind ChatGPT Could Help Spot Early Signs of Alzheimer’s Disease.” Drexel News.
https://drexel.edu/news/archive/2022/December/GPT-3-alzheimers-disease.

Team, Insights. 2023. “The state of generative AI in 7 charts.” CBI Insights. https://www.cbinsights.com/research/
generative-ai-funding-top-startups-investors/.

Vaswani, Ashish. 2017. “Attention Is All You Need.” arxiv 1, no. 12 (December): 23.

Warnke, Arne. 2023. “The Impact of ChatGPT and Generative AI on Jobs.” Data camp. https://www.datacamp.
com/blog/the-impact-of-chatgpt-and-generative-ai-on-jobs.

WHO, Growth. 2023. “These are the jobs most likely to be lost – and created – because of AI.” World Economic
Forum. https://www.weforum.org/agenda/2023/05/jobs-lost-created-ai-gpt/.

Wiles, Jackie. 2023. “Beyond ChatGPT: The Future of Generative AI for Enterprises.” Gartner. https://www.gartner.
com/en/articles/beyond-chatgpt-the-future-of-generative-ai-for-enterprises.

45
Impact, Opportunity, and Challenges of Generative AI

Notes

46
 Impact, Opportunity, and Challenges of Generative AI

The National AI Portal of India (INDIAai) is a central repository of AI information and

resources, serving the Indian AI ecosystem and providing global insight into AI in India.
Launched in May 2020, it has gained recognition and features interviews and articles
by industry leaders. The portal houses information on AI initiatives by the government,
academia, and corporates, while maintaining a database of Indian AI startups. INDIAai
is respected as an authoritative source on AI in India and worldwide. With the potential
to add $957 billion to India’s economy by 2035, INDIAai aims to foster a unified AI
ecosystem, drive knowledge creation, develop an AI-ready workforce, and utilize AI for
economic growth.

Team INDIAai
Kavita Bhatia, Senior Director, MeitY
Prashant Kumar Mittal, Director, NeGD, MeitY
Sushil Kumar Jangid, Scientist, MeitY
Bramhanand Jha, Senior Consultant - PM, MeitY
Sangeeta Gupta, Senior Vice President, NASSCOM
Asna Siddiqui, Project Lead, INDIAai
Jibu Elias, Content and Research Lead, INDIAai
Anjali Pathak, Product & Social Media Lead, INDIAai
Nibedita Saha, Senior Research Associate, INDIAai
Dr. Nivash Jeevanandam, Senior Research Writer, INDIAai
Anjali Raja, Content & Research Associate, INDIAai
Milin Stanly, Content & Research Associate, INDIAai

47
Impact, Opportunity, and Challenges of Generative AI