AI & ML Fundamentals

I attended an insightful webinar hosted by REVA Academy for Corporate Excellence - RACE with Shriram Vasudevan (FIE, FIETE,SMIEEE) discussing the topic : Generative AI on the cloud.

This diagram I made below summarizes the idea of AI, its categories, and concepts from the session.

Machine Learning (ML) focuses on using algorithms like classification, clustering, and regression to enable machines to learn from data and make decisions based on it.

We focused on the following:

Types of Models in DL-

Discriminative Models: They classify or predict elements, for example, distinguishing between pictures of cats and dogs by learning their characteristics from labeled data.

Generative Models: They generate new content based on trained data. For example, we train the model with a series of cat pictures and now the model knows a cat's features. Now if we want to be creative and create a new picture of a cat with special features like, a longer tail, red eyes, pointy ears, etc., this type of model comes to play and will generate that new content for us.

Gen AI generates content, whether text, images, or other data, by learning patterns from existing data. These models are trained on massive datasets.

Within this domain, we have Large Language Models (LLMs). Take GPT-3.5 as an example, which is trained on vast amounts of text data and designed to understand and generate human language.

There are two basic types of LLMs:

Base LLMs: Foundational models trained on large amounts of data with minimal or no specific instructions, Like a freshman in college who has general knowledge.

Instructional LLMs: Enhanced base LLMs fine-tuned with instructions to perform tasks and to answer prompts more effectively, similar to a senior graduate who can answer questions and perform tasks efficiently.

Now, this is the interesting part:

Prompts are the questions or instructions we give to chatbots like ChatGPT. 

They can be:

Simple: Basic questions or commands with no detail (e.g., "What is 1+2?")

Detailed: More sophisticated prompts that ask for explanations (e.g., "What is 1+3x5? Explain how you calculated this.")

To get the correct response from the machine, it's important to use the right prompting strategy:

Zero-shot prompts: Directly prompting the model with no examples.

Few-shot prompts: Providing a few examples to guide the model.

Chain-of-thought prompting: Using complex reasoning to guide the model's response for better accuracy. For example, asking a chatbot to solve the math problem 8 * (5+3) step-by-step helps ensure the model demonstrates the correct reasoning process.

The more detailed your prompt, the better the result.

Lastly, we explored how the Intel Developer Cloud Service offers learning content, training, and workshops to enhance our skills. This platform is very resourceful and AI beginner-friendly where we saw real-time text-to-image generation through prompting.