short courseGenerative AI and ChatGPT

Session 1: Introduction to Generative AI

This session will introduce you to the history and evolution of artificial intelligence (AI), overall evolution and tools for Generative AI, brief about large language models and ChatGPT and cover some practical applications of generative AI

• Understand the history and evolution of artificial intelligence (AI) and Generative AI
• Understand essential terminology and tools used in Generative AI including ChatGPT, Stable Diffusion, Midjourney, Google Bard, ....
• Learn fundamental concepts of large language models (LLMs) like ChatGPT
• Learn about open-source and commercially available LLMs
• Real-world applications and use-cases of generative AI with live demos

Session 2: ChatGPT in Everyday Life and Business

This session will introduce you to the workflows of large language models like ChatGPT, fundamentals of prompt engineering, cover advantages and limitations of ChatGPT, showcase some real-world applications of ChatGPT and key considerations about Bias, Ethics and Moderation policies in Generative AI. You will learn how to use ChatGPT in different real-world scenarios like creating a travel itinerary, acting as a program manager, building meal and workout plans and more!

• Learn fundamental workflows of LLMs like ChatGPT
• Learn Prompt Engineering Fundamentals
• Learn about how to create intuitive prompts and prompting patterns for ChatGPT
• Advantages and limitations of ChatGPT
• Real-world applications and use-cases of prompt engineering with ChatGPT with live demos
• Bias, Ethics and Moderation policies for Generative AI

Session 3: Anatomy of Generative AI Models

This session will introduce you to the key model architectures of language models and large language models used in generative AI. We will also briefly cover other generative models used in domains like computer vision and structured data. Hands-on applications of such models will also be showcased on practical problems like summarization, translation, question-answering and more.

• Architecture of Language Models vs. Large Language Models
• Traditional Language Model architectures:
  • RNNs, LSTMs, GRUs
  • Sequence-to-Sequence Models
• Large Language Model architectures:
  • Transformers - Autoencoding, Autoregressive and Sequence-to-Sequence
  • Multi-modal Transformers
• How LLMs like ChatGPT evolved from transformers
• Real-world applications of Transformers

Session 4: - Unleash the potential of ChatGPT

This session will introduce you to ChatGPT and deep-dive into LLM model architectures and training workflows used to build ChatGPT. We will also cover various input and output types in ChatGPT and deep dive into prompt engineering patterns. You will learn various prompt patterns which can truly unleash the potential and creativity of ChatGPT to solve diverse real-world problems. This includes asking and answering questions, learning a custom language, generating detailed plans and workflows, reasoning and solving complex logical and business problems.

• Architecture and training workflow of ChatGPT
• Transformer models used to build ChatGPT - GPT-3.5, GPT-4
• Persona, input and output types in ChatGPT
• Prompt Engineering Deep Dive to unlock the potential of ChatGPT
• Learn and apply prompt engineering patterns including:
  • Question Refinement, Alternative Approaches, Ask for questions patterns’ to generate better prompts
  • Outline Expansion, Menu Actions patterns’ to organize and structure information better
  • Persona, Template, Recipe and Meta Language Patterns’ to reason and build complex workflows
  • Cognitive Verifier, Chain-of-thought and Few Shot Prompting Patterns’ to think, reason and solve complex cognitive problems
• Real-world applications of ChatGPT using a combination of prompting patterns

Session 5: Building Applications with the ChatGPT API - I

This session will introduce the audience to the OpenAI developer API ecosystem. You will learn how to setup your OpenAI account, converse with ChatGPT models via APIs and perform iterative prompt engineering. We will go over the essential operations including inference, summarizing, transformation, expansion and chain-of-thought reasoning as we explore various prompt engineering patterns learnt in the previous week using OpenAI and Python.

• Learn how to setup an OpenAI account and get your API key
• Converse with ChatGPT using Python and the ChatGPT API
• Prompt Engineering Development using ChatGPT API
  • Learn how to do iterative prompt engineering to build better prompts
  • Learn essential operations using ChatGPT - inference, summarizing, transformation, expansion and chain-of-thought reasoning
  • Chain a combination of prompting patterns and build your own applications

Session 6: - Building Applications with the ChatGPT API - II

This session will expand upon the building blocks learnt in the last session using OpenAI’s APIs to build real-world applications using ChatGPT. We will build a variety of real-world applications here including QA Search and Chatbot Assistants. Alternative approaches using open-source LLMs will also be explored. Finally we will bring everything together by showing you how to build an end-to-end application using ChatGPT, Streamlit \ Chainlit.

• Real-world application development with the ChatGPT API
  • Build a Question-Answering Chatbot
  • Build a Marketing Manager Assistant
  • Build a Store Chatbot Assistant
• Alternative approaches using open-source LLMs
• Build an end-to-end application with ChatGPT API and Streamlit, Chainlit

Session 7: LLM Development with Langchain - I

This session will introduce the audience to the Langchain framework which is extremely powerful to build LLM workflows and applications. You will learn the fundamental building blocks of Langchain, how to abstract boilerplate code, reuse prompting templates, leverage LLM conversation memory and build real-world application prototypes using LLM models.

• Learn the building blocks of Langchain - Models, Parsers, Prompts, Indexes, Chains and Agents
• Learn the art and science of abstracting boilerplate code, reusing prompt templates, using conversation memory for rapid prototyping of LLM applications using the Langchain API
• Build a Customer Support Assistant using Langchain and ChatGPT

Session 8: LLM Development with Langchain - II

This session will deep dive into some of the advanced components of Langchain where you will learn to leverage tools and agents to solve complex problems with LLMs. You will also learn how to evaluate performance of LLMs and build real-world applications like a QA Search Engine Assistant and a Financial Analyst assistant.

• Learn how to use pre-built tools and agents to perform complex tasks with LLMs
• Learn how to evaluate LLM prediction performance
• Real-world application development with ChatGPT and Langchain
  • Have a conversation with your data by building a Question-Answering Search Engine Assistant
  • Build your own financial analyst assistant

Session 9: Unleash the creative potential of Generative AI

• AI in Music Generation (muBert)
• AI in Art Creation (Stable Diffusion, Midjourney, presentations generation)
• AI in Content Creation (Text, Image)

Session 10: Commercializing Generative AI Projects

• Understanding Business Model Canvas
• Translating AI Capabilities into Business Solutions
• AI in Various Industries: Health, Finance, Retail, etc.
• Case Studies of Successful AI Startups