Detailed Description

The module is partly based on the first three sections of “Artificial Intelligence: A Modern Approach” (4th edition) by Russell and Norvig. The module covers the history of AI and traditional AI approaches to solving problems as well as covering the technologies behind more recent advances such as Large Language Models. The module introduces you to the Transformer architecture behind Large Language Models (LLMs). You will learn how to use LLMs well by use of advanced prompting techniques, and how to fine-tune LLM’s for particular tasks. You will also learn how to evaluate LLMs and consider the ethical issues in their use. The use of the “human in the loop” to improve these kinds if AI systems through reinforcement learning is also considered. Looking at more tradition approaches to AI, the concept of an intelligent agent is explored along with the architectures that might support intelligent agent systems.Traditional search techniques such as those used within pathfinding software such as Google Maps are explored and compared. Techniques for searching where there is an opponent or “adversary” – for example when playing chess, are also explained. We also look at ways in which these search techniques can be optimised. Other traditional approaches such as knowledge representation, propositional logic, and reasoning are covered providing insights into AI approaches that are “explainable”. You will also learn about simple ways to solve optimisation problems that have many constraints – for example automated timetable production, and some more advanced methods that are inspired by nature.