Using GPT-3 for Traditional Machine Learning: A Tutorial
Abstract
Large Language Models (LLM) like ChatGPT, Bard and Claude have clearly demonstrated that Artificial Intelligence (AI) has not just come of age but is fueling speculation that AI programs understand language like humans do and are even conscious. Machine Learning (ML) has been in the limelight longer. This tutorial looks at the intersection of these two exciting areas. It takes an early, high-performing LLM (GPT-3) outside the zone of its mainstream application and shows the steps to use it for more conventional ML tasks. We provide a quick and simple introduction to GPT-3 and ML and then show all the required steps from obtaining a GPT-3 account, all the way to running ML with GPT-3 on an example dataset. We also provide all the resources for people to try these steps on their own datasets. Some familiarity with running Python programs on a command line will help, but is not essential. Both Windows and Mac users can follow the tutorial.