🪧

Generative AI course

1. What is Generative AI?

2. Roadmap of Generative AI?

3. What things do we need to learn in a Generative AI course?

4. What are the foundational models in Generative AI?

5. What is the future scope of Generative AI?

6. What is the eligibility for a Generative AI course?

7. Why should you learn Generative AI?

8. What are the types of Generative AI compared to other AI types?

9. What are the advantages and disadvantages of Generative AI?

Generative AI course 1
 what is Generative AI
Generative AI refers to AI technologies that can create new content, ideas, or
data that are coherent and plausible, often resembling human-generated
outputs. It has a plethora of practical applications in different domains such as
computer vision, natural language processing, and music generation.

Roadmap for Generative AI

1. Foundation in Machine Learning and Deep Learning

Ensure proficiency in probability, statistics, linear algebra, and calculus.

Gain hands-on experience with programming languages like Python/R.

Familiarize yourself with supervised and unsupervised learning algorithms.

Build machine learning models on tabular datasets.

2. Deep Learning Mastery

Develop a solid understanding of deep learning architectures such as
MLPs, RNNs, LSTMs, GRUs, and CNNs.

Gain proficiency in at least one deep learning framework like Keras,

TensorFlow, PyTorch, or FastAPI.

Train MLPs on tabular datasets.

Construct RNNs and CNNs for unstructured data (text and image).

3. Advanced Deep Learning Techniques

Learn about pretrained models for image data and their types.

Understand language models and build them using LSTMs/GRUs.

Explore attention mechanisms and their applications.

Study autoencoders and GANs architectures and train these models on

datasets.

4. Generative Models for NLP:

Discover Large Language Models (LLMs) like Transformers, BERT, GPT,
PaLM, etc.

Generative AI course 2
 Understand how to use LLMs for downstream tasks: finetuning, zero-shot,
one-shot, and few-shot learning.

Learn best practices for training LLMs, including scaling laws and efficient
training mechanisms.

Explore techniques to pre-train LLMs on domain-specific data.

Implement different techniques to finetune LLMs for downstream tasks.

Study optimization techniques like Adapters, LoRA, QLoRA, etc., to

accelerate finetuning.

Understand deployment considerations (LLMops) for deploying LLMs in

production.

Explore cutting-edge models like ChatGPT and BARD and understand their
training process, including reinforcement learning from human feedback,
supervised fine-tuning, and prompt engineering.

Gain hands-on experience with LLM frameworks like LangChain, AutoGPT,

Vector DB etc.

Disadvantages :

1. Source Ambiguity: Generative AI may not always attribute content to its

original source, leading to ambiguity in authorship and ownership.

2. Bias Assessment Difficulty: Assessing bias in original sources can pose

challenges, complicating efforts to ensure fairness and accuracy in
generated content.

3. Realism Challenges: The lifelike quality of generated content can obscure

inaccuracies, making it more difficult to distinguish between genuine and

Generative AI course 3
 fabricated information.

4. Tuning Complexity: Adapting generative AI for new contexts and scenarios

can be intricate, requiring nuanced adjustments to ensure optimal
performance.

5. Potential for Oversights: Despite advancements, there remains a risk that

generative AI results may inadvertently perpetuate biases, prejudices, and
hateful ideologies, warranting careful scrutiny and mitigation measures .

What are the benefits of generative AI?

Automating the manual process of writing content.

Reducing the effort of responding to emails.

Improving the response to specific technical queries.

Creating realistic representations of people.

Summarizing complex information into a coherent narrative.

Simplifying the process of creating content in a particular style.

What are the concerns surrounding

generative AI?

It can provide inaccurate and misleading information.

It is more difficult to trust without knowing the source and provenance of

information.

It can promote new kinds of plagiarism that ignore the rights of content
creators and artists of original content.

It might disrupt existing business models built around search engine

optimization and advertising.

Generative AI course 4
 It makes it easier to generate fake news.

It makes it easier to claim that real photographic evidence of a wrongdoing

was just an AI-generated fake.

It could impersonate people for more effective social engineering cyber

attacks.

What are some examples of generative AI

tools?

Text generation tools include GPT, Jasper, AI-Writer, and Lex.

Image generation tools include DALL-E 2, Midjourney, and Stable

Diffusion.

Music generation tools include Amper, Dadabots, and MuseNet.

Code generation tools include CodeStarter, Codex, GitHub Copilot, and

Tabnine.

Voice synthesis tools include Descript, Listnr, and [Link].

AI chip design tool companies include Synopsys, Cadence, Google, and

Nvidia.

What are use cases for generative AI?

Implementing chatbots for customer service and technical support.

Deploying deepfakes for mimicking people or even specific individuals.

Generative AI course 5
 Improving dubbing for movies and educational content in different
languages.

Writing email responses, dating profiles, resumes and term papers.

Creating photorealistic art in a particular style.

Improving product demonstration videos.

Suggesting new drug compounds to test.

Designing physical products and buildings.

Optimizing new chip designs.

Writing music in a specific style or tone.

Some of the very commonly known LLMs

are:

Open AI’s GPT 3, 3.5, and 4

Google’s LaMDA and PaLM

Hugging Face’s BLOOM

Meta’s LLaMA

NVidia’s NeMO LLM

Types of Models for Generative AI ?

1. Generative Adversarial Networks (GANs): GANs consist of two neural

networks, a generator and a discriminator, trained simultaneously in a
game-like scenario. The generator tries to create data (e.g., images) that
are indistinguishable from real data, while the discriminator tries to

Generative AI course 6
 differentiate between real and generated data. This adversarial training
process helps improve the quality of generated samples.

2.

1. Variational Autoencoders (VAEs): VAEs are generative models based on

the principles of autoencoders. They consist of an encoder network that
compresses input data into a latent space representation and a decoder
network that reconstructs the original data from the latent space. VAEs are
trained to generate new data samples by sampling from the learned latent
space distribution.

2.

1. Transformers: Transformers are a type of deep learning model primarily

used in natural language processing (NLP). They are based on a self-
attention mechanism that allows the model to weigh the importance of
different words in a sequence when processing each word. Transformers
have achieved state-of-the-art performance in various NLP tasks and have
also been adapted for other tasks beyond NLP, such as image generation
and reinforcement learning.

2.

Generative AI course 7
 1. Autoencoders: Autoencoders are neural networks trained to copy their
input to their output. They consist of an encoder network that compresses

Generative AI course 8
 the input data into a latent space representation and a decoder network that
reconstructs the original input from the latent space representation.
Variants like denoising autoencoders, sparse autoencoders, and contractive
autoencoders serve different purposes.

2.

Future capabilities of generative AI:

Continued Model Scaling and Efficiency Gains
Larger and more parameter-dense models promise to further expand
generative capabilities and general competencies.

At the same time, more efficient training techniques will reduce data and
computing needs to improve sustainability.

Also, new hardware and software architectures could be developed to

more effectively handle resource intensive workloads.

Generative AI course 9
 Steady Improvements in Output Quality and
Accuracy
As this technology advances, its responses may become indistinguishable
from, and possibly superior to human-generated material.

Improvements will be achieved through advances in responsible

development, model architecture, training techniques, data quality and
diversity, and evaluation metrics and tools.

Also, end users and developers contribute to improvements by providing

feedback on responses, labeling data, generating ideas, developing
models, and building tools.

Multiple Layered Agent Frameworks:

Multiple layered agent frameworks are a complex and hierarchical
structure of AI agents designed to interact and collaborate effectively in a
dynamic environment.

This type of framework allows numerous agents to communicate with one

another in order to solve problems that are too complicated for a single
agent to accurately accomplish on its own.

In addition, multiple AI agents working together can dramatically increase

the productivity of task completion.

For instance, picture a transportation network management system where

sensor agents gather real-time data, data processing agents analyze it,
routing agents determine optimal paths, coordination agents ensure
harmonious cooperation, and supervisory agents oversee the entire
operation.

Autonomous AI Agents
One or more AI agents with the ability to learn and adapt on their own,
without the need for human intervention.

This will allow them to automate tasks and make decisions in real time.

For example, an autonomous AI agent could be used to monitor current

events and automatically generate content on flagged topic categories.

Generative AI course 10
 Increased Customization and Specialization :
Generative AI will become more customizable, allowing for more
parameters to be fine-tuned on custom data.

This contextual fine-tuning improves relevance, quality and safety for

specific professional use cases while retaining general advantages.

Which will serve as springboards for developing specialized

systems tuned on custom datasets for focused tasks, verticals and
applications.

To illustrate, a healthcare company could fine-tune a generative AI model

on medical data to generate new drug candidates, predict the effectiveness
of different drugs, and design new drug compounds.

Accessibility :
Generative AI tools will become more accessible to people with less data
and expertise.

Making it easier for businesses of all sizes to use this technology by

lowering barriers to access, enabling broader adoption across languages,
domains, geographies and less resourced organizations.

Specifically, the development of user-friendly apps and AI-as-a-service

(ALaaS) platforms will enable access without having to build and maintain
their own infrastructure.

Text to Everything
The ability to generate text, images, audio, videos, and 3D environments
from a single platform.

This means users will provide a single text prompt and the model will
generate every content format simultaneously.

For example, a user could provide a prompt like “A beautiful sunset over a
beach” and the model would generate a text description of the sunset, a
video of the sunset, and an audio file of the sound of waves crashing on the
shore.

Generative AI course 11
 Ultra Realistic Talking Avatars :
Generative AI models will be able to create videos of talking avatars that
are so realistic that they are indistinguishable from real people.

These avatars will be able to lip sync to speech, have natural facial
expressions, subtle head movements and even generate their own
dialogue.

This will have a wide range of applications, including virtual assistants,

customer service representatives, and even educational characters

Impact of Generative AI on industry and

jobs :

Changing Job Roles and Required Skills :

Though AI won’t replace human roles outright, it will alter the associated
required skills.

workers and job requirements will need to transition into more strategic
roles.

This may involve upskilling to perform higher-value work like data

analysis, creative direction and strategy.

Automating Tasks to Increase Efficiency

Conducting process analysis can be used to identify repetitive and low
strategic value tasks that are ideal for AI automation.

Task categories such as document review, contract analysis, customer

service routines and content generation contain repetitiveness that humans
find tedious.

Generative AI course 12
 free up employees for more impactful and creative responsibilities better
suited to human strengths.

Renewed Focus on Data Governance

Generating high-quality outputs relies on extensive training data that is
accurate, comprehensive and aligned to use cases.

This places greater emphasis on responsible data governance, including

ethics, integrity, monitoring, security and compliance.

Where data collection practices will require more comprehensive

auditing to avoid perpetuating biases.

Tight Integration into Workflows and Software

To maximize user adoption and impact, AI systems will integrate
seamlessly into existing enterprise software, workflows and processes.

This integration will emphasize on providing a high-quality user

experience by actively seeking continuous feedback to drive
enhancements.

Ultimately, the main objective of these AI systems is to seamlessly

integrate into workflows, to feel like a natural system component rather
than function as isolated tools.

Lower Barriers of Entry

Generative AI will make it easier for new competitors to enter the market.

As pre-trained models and AI-as-a-service (ALaaS) platforms become more

accessible, smaller players will leverage these technologies without
massive in-house investments, somewhat leveling the playing field.

As a result, this will increase market competition and force existing

enterprises to adapt quickly and differentiate themselves.

Enabling New Niche Capabilities

With the help of generative AI, one can scan the market for untapped
areas to uncover potential niche opportunities.

Generative AI course 13
 Specifically, to identify new market opportunities from various sources by
analyzing customer data, social media, demographics, economic trends,
etc.

By feeding all these data points together, the AI model will be able to
identify new customer needs and wants that are not currently being met
by existing products and services.
• Then
integrate such findings into the process to generate new ideas for
products and services.

Applications Across Industries:

Marketing, Advertising, and Entertainment Industry:

1. Content Creation: Generative AI powers content creation in various

forms like art, music, and literature, allowing artists and musicians to
explore innovative creative directions.

2. Video Game Development: AI-driven systems create game

environments, characters, and dialogues, reducing development time
and resources.

3. Scriptwriting: Screenwriters use Generative AI to assist in scriptwriting

by generating dialogues, plotlines, and character interactions.

Education Sector:

1. Personalized Learning: Generative AI adapts educational content to

individual student needs by generating tailored assignments, quizzes,
and study materials.

2. Knowledge Base: AI creates exhaustive knowledge bases for students

to access information in a conversational style.

3. Virtual Labs: Generative AI powers virtual laboratories for simulating

experiments and scenarios in science and engineering disciplines.

Healthcare Industry:

1. Medical Image Generation: Generative AI creates synthetic medical

images for training models, enhancing diagnostic accuracy, and
simulating rare medical conditions.

Generative AI course 14
 2. Drug Discovery: Pharmaceutical companies use Generative AI to
discover new drug compounds by generating molecular structures.

3. Personalized Medicine: AI analyzes patient data to generate

personalized treatment plans, accounting for genetic factors and
medical history.

Manufacturing Industry:

1. Product Design: Generative design creates optimized product designs

considering materials, weight, and structural integrity.

2. Quality Control: AI generates synthetic data for quality control testing,

ensuring adherence to quality standards.

3. Supply Chain Optimization: AI-generated forecasts help make informed

decisions about production and distribution.

Software & Tech Industry:

1. Code Generation: Generative AI assists developers by generating code

snippets and templates, speeding up development.

2. Bug Detection: AI generates synthetic test cases to identify and fix

software bugs efficiently.

3. IT Security: Generative AI models simulate cyberattack scenarios to

identify vulnerabilities and enhance cybersecurity measures.

What is prompt engineering

Generative AI course 15
 Prompt engineering is the practice of designing inputs for AI tools that will
produce optimal output

why we need to learn prompt enginerring ?

Enhanced accuracy and relevance

Improved decision making

Personalized customer experiences
Efficient resource utilization
Ethical considerations and bias mitigation

There are 8 prompt engineering methods:

(1) Zero-Shot Learning, (2) One-Shot Learning, (3) Few-Shot Learning,

(4) Chain-of-Thought Prompting, (5) Iterative Prompting,
(6) Negative Prompting, (7) Hybrid Prompting, and
(8) Prompt Chaining .

1.
Zero-Shot Learning: This involves giving the AI a task without any prior
examples. You describe what you want in detail, assuming the AI has no prior
knowledge of the task.

Generative AI course 16
 . One-Shot Learning: You provide one example along with your prompt. This
helps the AI understand the context or format you’re expecting.
3. Few-Shot Learning: This involves providing a few examples (usually 2–5) to
help the AI understand the pattern or style of the response you’re looking for.
4. Chain-of-Thought Prompting: Here, you ask the AI to detail its thought
process step-by-step. This is particularly useful for complex reasoning tasks.

5. Iterative Prompting: This is a process where you refine your prompt based
on the outputs you get, slowly guiding the AI to the desired answer or style of
answer.
6. Negative Prompting: In this method, you tell the AI what not to do. For
instance, you might specify that you don’t want a certain type of content in the
response.

7. Hybrid Prompting: Combining different methods, like few-shot with chain-of-

thought, to get more precise or creative outputs.
8. Prompt Chaining: Breaking down a complex task into smaller prompts and
then chaining the outputs together to form a final response.

Key elements of a prompt :

Instruction. This is the core directive of the prompt. It tells the model what
you want it to do. For example, "Summarize the following text" provides a
clear action for the model.

Context. Context provides additional information that helps the model

understand the broader scenario or background. For instance, "Considering
the economic downturn, provide investment advice" gives the model a
backdrop against which to frame its response.

Input data. This is the specific information or data you want the model to
process. It could be a paragraph, a set of numbers, or even a single word.

Generative AI course 17
 Output indicator. Especially useful in role-playing scenarios, this element
guides the model on the format or type of response desired. For instance,
"In the style of Shakespeare, rewrite the following sentence" gives the
model a stylistic direction.

Technical skills for prompt engineering :

Understanding of NLP. A deep knowledge of Natural Language Processing

techniques and algorithms is essential.

Familiarity with LLMs. Experience with models like GPT, PaLM2, and other
emerging models their underlying architectures.

Experimentation and iteration. Ability to test, refine, and optimize prompts

based on model outputs.

Non-technical skills for prompt engineering

Communication. The ability to convey ideas, collaborate with teams, and

understand user needs.

Subject Matter Expertise. Depending on the application, domain-specific

knowledge can be invaluable.

Language Proficiency. Mastery over language, grammar, and semantics to

craft effective prompts.

Critical Thinking. Evaluating model outputs, identifying biases, and

ensuring ethical AI practices.

Creativity. Thinking outside the box, experimenting with new prompt styles,
and innovating solutions.

Generative AI course 18
Generative AI course 19