Generative AI Community Session - Lecture Notes

Introduction

• Date & Time: First session today, continuing for two weeks, 3:00 PM - 5:00 PM.
• Purpose: Discuss generative AI concepts, applications, and practical implementations.
• Format: Theoretical discussions followed by practical applications, quizzes, and assignments.
• Instructor: Sunny (Savita) - 3 years of experience in data science, specializing in ML, DL, and applications.

Session Structure

• Dashboard Overview:
  • Dashboard link shared in chat for enrollment (free).
  • Access to lectures, assignments, quizzes on the dashboard and YouTube channel.

Curriculum Overview

1. Generative AI Basics:
   • Introduction to generative AI
   • Types of applications
2. Large Language Models (LLM):
   • Overview and history of LLMs
3. OpenAI and LangChain:
   • OpenAI API usage and comparison with LangChain.
4. Application Development:
   • Building applications using generative AI
5. Vector Databases:
   • Importance in generative AI applications.
6. Open Source Models:
   • Discuss models like Llama, Falcon, Bloom.
7. End-to-End Project Development:
   • Use acquired knowledge to create and deploy projects.

Prerequisites for Participants

• Basic knowledge of Python.
• Familiarity with machine learning and deep learning concepts is beneficial.

Generative AI and LLM Concepts

• Definition of Generative AI:
  • Generates new data based on training samples (images, text, audio, video).
• Generative vs. Discriminative Models:
  • Generative models generate new data; discriminative models classify data.
• Types of Neural Networks:
  • Artificial Neural Networks (ANN)
  • Convolutional Neural Networks (CNN)
  • Recurrent Neural Networks (RNN)
  • Generative Adversarial Networks (GANs)
  • Long Short-Term Memory (LSTM)
  • Gate Recurrent Unit (GRU)

Applications of Generative AI

• Use Cases:
  • Text generation, summarization, chatbot development, language translation, and more.
• Transformers:
  • The architecture behind modern LLMs, which allows parallel processing of inputs.

Important Papers and Research

• Mentioned key research papers that shaped the concepts in this field:
  • Sequence to Sequence Learning
  • Attention is All You Need
  • ULMFiT (Universal Language Model Fine-tuning)

Tools and Technologies Discussed

• OpenAI Models:
  • ChatGPT, GPT-3, GPT-3.5, GPT-4
• Open Source Models:
  • Llama, Falcon, Bloom
• Hugging Face:
  • Model hub for various open-source models.
• AI21 Labs:
  • Alternative to OpenAI with free credit.

Next Steps

• Tomorrow’s Session:
  • Practical implementation of OpenAI API.
  • Application examples and prompt engineering.
• Assignments and Quizzes:
  • Available on the dashboard for practice.
• Feedback:
  • Participants encouraged to provide session feedback and questions.

Conclusion

• Encouragement:
  • Stay engaged, participate in discussions, and practice with assignments.
• Next Class Time:
  • Tomorrow at 3:00 PM.