Original title: Generation of Explanations for Logic Reasoning
Authors: Yanyi Pu
In this thesis, the focus is on using a powerful AI, GPT-3.5-turbo, to explore a specific type of deductive reasoning called a fortiori arguments. These arguments pop up in law, philosophy, and AI, and the goal is to teach the AI to understand and explain them. The research involves putting the AI through various tasks to see how well it can handle these complex arguments. It’s a bit of a challenge for the AI to accurately spot and sort these arguments, but it shows some impressive skills in picking out the most important bits and understanding the underlying ideas. By adding more information, the AI gets even better at explaining these arguments. Despite some limitations, this thesis brings fresh ways to study AI and logic. It creates new ways to test AI’s reasoning, offers cool insights, and sets the stage for smarter AI in the future. It’s not just about what AI can do now but hints at where it could go next.
Original article: https://arxiv.org/abs/2311.13455