Fundamentals of Generative AI

Instructor: Dr Agha Ali Raza

The Fundamentals of Generative AI course provides an in-depth exploration of AI's latest capabilities across
various industries. Participants will gain expertise in working with text and images in Generative AI, prompt
engineering and AI-driven data analysis, enabling them to integrate these powerful tools into their professional
practices. Join us to uncover the potential of generative AI and learn how to create innovative solutions
tailored to your specific industry needs.

Audience: Professionals belonging to Marketing, Media and Creative Industries, Lawyers and Policy Makers,
Doctors and Healthcare professionals, Finance, Banking, Business strategists and business owners, Data
Scientists and Analysts, Product Managers, HR managers, Communications specialists, Authors, Artists,
Technology and IT, Teachers, Education Research and Academia.

By the end of the workshop, you will be able to:

1. Understand the foundations of AI, generative AI, and key terminology.

2. Master prompt engineering, multilingual and multimodal prompting techniques.
3. Comprehend how language models and large language models (LLMs) work.
4. Learn advanced data analysis and visualization using AI tools.
5. Image generation (text to image) and image analysis (automatically describing an image) using AI.
6. Learn Retrieval Augmented Generation (RAG) with hands on exercises.
7. Create your own custom GPT models.
8. Learn about the limitations and risks of AI integration and the importance of safety guardrails.
9. Learn practical steps towards transforming your company into an AI-enabled organization.

You will also be able to apply the above for:

1. Integrating AI tools into professional workflows for improved efficiency.

2. Developing customized AI bots tailored to specific industry use cases.
3. Creating innovative artwork, presentations, and creative content using AI.
4. Deploying chatbots as personal assistants, creative companions, and research aides.
5. Applying ethical considerations and best practices for responsible AI use.

Week / Day Topic Details

1/1 Introduction to Artificial ● Explore the historical milestones of AI

Intelligence and Generative development.
AI ● Understand the basic concepts and types of AI.
● Define generative AI and its distinction from other
types of AI.
● Review significant achievements and
breakthroughs in generative AI.

Acquaintance with
ChatGPT3.5, 4 turbo (4T), 4 ● How to login to each tool, get familiar with tools and
omni (4o), Gemini, Bing their capabilities
● Important terminology to be familiar with (neural
image creator, MS copilot
networks, transformers, RLHF, and LLMs)
 ○ Simple explanations of all of these concepts

1/2 1. Language Models and

LLMs - how do they work? ● Breakdown of Explanation of training processes,
2. The science of prompt pipelines and data requirements.
engineering and multilingual 1. Principles of Prompt Engineering:
prompt engineering
3. Chain of thought ● Crafting Effective Prompts: Techniques for
prompting, few shot creating accurate and relevant prompts.
prompting ● Context and Continuity: Ensuring coherence in
multi-turn interactions.
● Prompt Variations and Experiments: Exploring
different prompt variations for desired outputs.

2. Playing with Language and Creativity:

● Creative Applications of Prompts: Using prompts

for creative writing and storytelling.
● Multilingual Prompt Engineering: Designing
prompts for multilingual applications and translation
tasks.
● Interactive and Adaptive Prompts: Developing
dynamic prompts for interactive AI systems.

2/3 Gen AI: A Thought Concept and Mechanism:

Companion,
Assistant and Personal ● Understand the principle of RAG and how it
enhances AI capabilities.
Trainer - Retrieval
● Explore different RAG architectures and their use
Augmented Generation cases.
(RAG)
Implementation and Examples:

● Practical implementation of RAG models.

● Case studies demonstrating the effectiveness of
RAG in various applications.

2/4 Data analysis and Data Analysis with AI:

visualisation in Data
● Techniques for using AI tools to analyze complex
datasets.
● Visualization strategies to represent data insights
effectively.

Interactive Sessions:

● Guided exercises on advanced data analysis.

● Use of real-world data sets to practice visualization.
3/5 & 6 Custom GPTs Customization Techniques:
Plugins and how they work
Semantic Spaces ● Steps to customize GPT models for specific
applications.
Reinforcement Learning and
● Tools and libraries for fine-tuning GPTs.
RLHF
Practical Customization:
This week will also be the
week to discuss participant ● Hands-on sessions to create and deploy custom
queries regarding the final GPT models.
● Review and refine custom models based on
project (details shared
performance.
below) ● Demonstrations of semantic space applications in
chatbots.

Reinforcement Learning and RLHF

Understanding RLHF (Reinforcement Learning from

Human Feedback)

4/7 & 8 ●Chatbots as creative

1. Understanding Roles:
thought companions,
personal trainers/assistants, ● Creative Thought Companions: Using chatbots
and research assistants for brainstorming, generating ideas, and creative
●Creating artwork images feedback.
presentations music text ● Personal Trainers/Assistants: Custom workout
plans, health advice, and motivation etc through
chatbots.
● Research Assistants: Summarizing articles,
finding literature, and organizing research
information.

2. Creating artwork with Generative AI

● Introduction to tools like DALL-E, bing image
generator and their functionalities.
● Step-by-step guide on using these tools to create
images.
● Hands-on sessions to create unique artwork using
generative AI.
● Analyze and critique generated artwork for
improvement.
3. Textual applications with generative AI
● Write emails, grants, reports etc ensuring clarity
and precision while saving time
● Discover how generative AI can be used to quickly
create dynamic visuals, customized templates that
elevate the quality and engagement of your
presentations.

5/9 Open Source Models and ● Overview of open source models like Meta Llama,
Fine-Tuning LLMs MS Phi, and Google Gemma: how they work and
their key features.
● Steps to deploying open source models effectively.
 ● Techniques for fine-tuning language models
(LLMs).
● Practical demonstration and hands-on session.

5/10 ●Situational awareness in AI Capabilities:

the Era of AI
● Discussion on the realistic potential of AI in various
●Challenges, Risks, Ethics sectors.
● current limitations and ongoing challenges.
● Identification of key factors contributing to AI
success and failure.

Ethical Considerations:

● Exploration of ethical issues surrounding AI usage.

● Strategies to mitigate risks and ensure responsible
AI deployment.

Regulatory Frameworks:

● Overview of existing regulations and guidelines for

AI.
● Discussion on the future direction of AI policy and
governance.

6/11 Final project presentations

and discussions *

6/12 -Final project presentations

and discussions *
-Summary, wrap up and Q
and A

*Final Project Guidelines: Creating Custom GPTs on GPT-4

Objective: In groups of 3-4 people, Develop a custom GPT-4 tailored to a specific and creative use case,
showcasing your understanding and practical skills from the course.

Project Requirements:

1. Choose an Exciting Application:

○ Think outside the box! Select a fun, innovative use case for your custom GPT-4 (e.g., a virtual
travel guide, a personalized story generator, a creative recipe creator).
○ Clearly outline the goals and intended use of your custom GPT-4.
2. Data Collection and Preprocessing:
○ Gather unique and engaging datasets relevant to your chosen application.
○ Ensure your data is clean, diverse, and rich in content.
3. Model Customization:
○ Fine-tune the GPT-4 model using your collected data.
○ Add specific tweaks to make your GPT-4 perform optimally for its creative purpose.
4. Implementation and Testing:
○ Develop a fun and interactive prototype of your custom GPT-4.
 ○ Test the model thoroughly to ensure it’s both functional and entertaining.
5. Evaluation and Metrics:
○ Define creative metrics to evaluate your model’s performance (e.g., user enjoyment, creativity
score).
○ Compare results with baseline models to highlight your improvements.
6. Documentation and Presentation:
○ Create a lively report detailing your project process, including data prep, model customization,
testing, and evaluation.
○ Prepare a 10-minute presentation to showcase your custom GPT-4, including a fun demo of its
capabilities.

Submission:

● Be ready to present and demo your project in an engaging and entertaining way on day 11 and 12 of
the course

Deadline:

● All project components must be submitted by the 11th session of the course.

Bonus Challenge:

● Include a feature in your custom GPT-4 that allows it to adapt to user feedback in real-time, making it
more interactive and dynamic.