Generative AI
This course is an introduction to generative AI, especially large language models such as ChatGPT.
Description
Generative AI – Course Outline
Generative AI is artificial intelligence capable of generating brand new content - typically text and images. It is a new technology that anyone can use now and use well with a little training. We chat with a generative AI in a natural language like English. There's no need for people to learn a language like Python or SQL to accomplish something. You state what you want in English (other languages are available) and you will get a useful response. Examples of Generative AI models are OpenAI’s ChatGPT, Microsoft’s Copilot and Google’s Gemini. People use generative AI to write reports, to brainstorm ideas, to make a plan, to provide step-by-step instructions, to check and improve their work and so much more.
Duration: one day
Pre-requisites: None. This is a course for beginners. No prior knowledge is assumed.
Learning Objectives
Students gain an appreciation of this very important technology so they can make good decisions about how and when to use it. It focuses on practical advice on how to get the best responses from ChatGPT and other LLMs.
Content
The course will be taught by guided tutorials where students work in teams to explore, experiment and interact with examples of Generative AI services such as LLMs.
Introduction
- A very short history of AI, generative AI and LLMs
- Different types of LLM (chat or image generation, open-source or proprietary, model or service).
- Using LLMs responsibly. LLMs can make mistakes, and possibly exhibit bias. How we can reduce the risk of these.
- How to get the best results and response from large lange models
- This is known as prompt engineering and is the main content of the course and covers:
- Why is prompt engineering important? (because a good set of prompts drive a much better, specific helpful set of responses)
- Types of prompts (basic, complex)
- Use examples (explain zero-shot, one-shot, few-shot example types)
- Prompt Cues and Template
•
Best practices and frameworks such as RICE (Role, Instructions, Constraints, Examples)
How Generative AI and LLMs work
This will be a very brief explanation of how Generative AI and LLMs work under the covers e.g. supervised and unsupervised learning, how generative AI differs from “traditional” AI, deep learning and neural nets, tokenisation, and the transformer architecture.