Generative AI Foundations - Course - Module - 1

Generative AI Foundations:

AI is transforming creation, learning, and problem-solving. You'll gain the knowledge and practical skills to use Generative AI for tasks like content creation, data analysis, and tackling complex challenges.

 

Objectives:

• Understand the fundamentals of Generative AI and how it works. 
• Demonstrate effective prompt engineering techniques to optimize AI outputs. 
• Utilize AI tools to conduct research, summarize information, and generate content. 
• Analyze and interpret AI-generated data to extract meaningful insights. 
• Develop AI-driven solutions through brainstorming and prototyping. 
• Assess ethical implications and best practices for responsible AI usage.

 

We will learn Generative AI Foundations course in 4 modules:

 

Module -1:

In this module we will learn Generative AI works and introduce you to some powerful tools that reshape creativity, learning, and problem-solving. 

 

What You Will Learn in This Module

• Fundamentals: Understand the concept of Generative AI and learn how it differs from other types of AI.
• How It Works: Ever wondered what happens "behind the scenes" when AI creates something new? We will use a straightforward analogy to help you understand.
• Tools: Get practical experience with ChatGPT and Microsoft Copilot, two powerful AI tools you can start using immediately.
• Art of the Prompt: Learn the secrets of interacting with AI effectively using the helpful PROMPT framework.
• Advanced Techniques: Improve your prompting skills with Zero-shot, Few-shot, and Chain-of-Thought prompting.

 

What You Will Achieve

• Gain a Strong Foundation in Generative AI Concepts
• Understand how AI generates new content 
• Learn to use AI tools like ChatGPT and Copilot effectively
• Develop better AI interactions using Structured prompts
• Apply advanced prompting techniques to get high-quality AI outputs

 

Understanding the Generative AI: The Layers

Generative AI: It is a type of AI. AI works on predictions.

For example: Forecasting tomorrow’s weather?

Here Generative AI generates new contents from text & images to sounds & videos by recognizing patterns.

 

There are 4 layers like.

1.   Artificial intelligence (AI): Its broadest concepts. Machines perform tasks that typically require human intelligence.  Like Problem Solving, understanding languages or recognizing images

2.   Machine Learning (ML): It is a subset of AI. Machines learn how to identify patterns and make predictions from data.

3.   Deep Learning (DL): It is a subset of ML. It uses networks to process vast amounts of data. Handles such as Image recognition or Natural languages processing

4.   Generative AI (Gen AI): It is a subset of DL. It creates new content like images, text, audio & video based on learned patterns.

 

Let’s understand how Generative AI works:

 

Flow: Prompt -> Tokens -> Self Attention -> Transformers -> LLM

 

Prompt -> When u ask any question or request to AI then it is called Prompt.

For Example: You asked what is the capital of France?

Now AI breakdowns these questions into Tokens.

 

Tokens -> Split the sentence into individual words.

Now AI understands each word and links together.

In such cases Neural Network comes into picture.

Neural Network -> Branching routes to find answers.

 

Self-Attentions -> Each part of the questions relates to others. Understanding the relationship.  

In our question – Capital related to France

 

Transformers -> its traffic controller. They experts to manage all self-attentions signals. They highlight the connection between capital and France.

 

LLM -> Large Language Model: LLM navigates the neural network map from question to answer.

 

AI calculates the probability of each possible next word, then selects the most probable choice from ranges of possibilities and provides the final answers.

 

So, in shorts – Starting by breaking the prompts into tokens tagging them into embedding and passing them through neural networks. Along the way transformers analyzed the relationship between words, then generated the accurate answer using transformers and LLM.

 

ChatGPT & Microsoft Copilot:

ChatGPT: Its separate AI chatbot. You need to log in and use it.

Copilot: Its Microsoft application integrated AI assistant.

 

Prompt Engineering:

Here we need to write exactly what you want, ask your question more clearly, specifically then AI provide a clearer answer.

 

What is Prompt:

Prompt is Input given to AI.

 

PROMPT:

P: Purpose: More clear, specific Ask

R: Role: Choose AI Role as teacher, scientist

O: Output: Define specific output format like script, poem, coding etc.

M: Markers: Define essential criteria, set AI guidelines like set words or paragraph limit

P: Patterns: Provide AI format, offer examples, report templates   

T: Tone: Define AI voice, specify desired style, set tone like formal or professional, informal, serious etc.  

 

 

Prompt Framework:

For example, you plan one trip to Paris & want to AI prepare this whole plan.

Here,

Purpose: Generate travel itinerary. 3 day’s trips.

Role: Travel Agents

Output: If you mention day by day itinerary then provide specific output.

Markers: Specify budget, spots

Pattern: Morning & Evening activities.

Tone:   provide user friendly, informative & helpful guide.

 

Main 3 components is Purpose, Role & Tone.

So whole sentence like:

Assume travel agent and generate travel itinerary 3 days trip to Paris. Need detailed day by day itinerary including hotels within mid-range of budget and focus popular places. Suggest best morning & evening activities. Provide the output in informative and user friendly.

If you provide an input like this then AI provides detailed 3 days plan with activities.

Here, we see how prompt framework provides relevant results.

Advanced techniques:

We need to learn 3 advanced techniques to get outstanding results from AI.

 

1. Zero-Shot Prompting

2. Few-Shot Prompting

3. Chian-of-Thought Prompting

 

1.Zero-Shot: Its shine when you need creative suggestions, quick solutions and general knowledge answers. Fast and example free results.

For e.g. ask: Name 3 famous landmarks in Landon

 

2.Few-Shot: Slightly fantastic for more complex or specialized tasks. Give 2-3 examples to clearly demonstrate the task and exact output.

For e.g. Ask: Create new LinkedIn posts & provide 2-3 sample post examples.

 

3.Chian-of-Thought: It is used for problem solving. AI builds logical arguments towards conclusion. Provide Step-by-step justifications.

 

Copilot in Outlook:

In copilot multiple options are displayed which are helpful to automate the outlook of activities. You can schedule the meeting, correct the sentence, you can just draft, it’s converted into professional tone etc.