Duration
25h Th, 20h Pr, 45h Proj.
Number of credits
Lecturer
Language(s) of instruction
English language
Organisation and examination
Teaching in the first semester, review in January
Schedule
Units courses prerequisite and corequisite
Prerequisite or corequisite units are presented within each program
Learning unit contents
This course is an introduction to artificial intelligence: how to build an agent that perceives an environment, reasons about what it cannot directly see, and decides what to do next.
It is not a course about using ChatGPT or Claude. It is about understanding and building such systems yourself, and it is the first step of a sequence (INFO8006, ELEN0062, INFO8010, INFO9023) that leads to the full stack, from a simple neural network to a working AI agent.
You will learn how a program searches for a plan, plays against an opponent trying to beat it, represents and updates beliefs under uncertainty, tracks a target it can only observe noisily, and learns a policy from experience. These ideas sit underneath most working AI systems, including the recent ones, and they do not go out of date every eighteen months.
Lectures follow several chapters of "Artificial Intelligence: A Modern Approach" (S. Russell and P. Norvig), the standard reference since 1995, with recent developments added where the book has aged.
Topics covered:
- Foundations of artificial intelligence
- Intelligent agents
- Solving problems by searching
- Games and adversarial search
- Quantifying uncertainty
- Probabilistic reasoning
- Reasoning over time
- Machine learning and neural networks
- Making decisions
- Reinforcement learning
This is a third-year computer science course. It is mathematical in places, particularly around probability, and the projects take real time.
Learning outcomes of the learning unit
At the end of the course, you will have a working map of the field, and you will have implemented its central algorithms yourself. You will be able to formulate a problem so that a machine can solve it, choose an algorithm and justify that choice, and judge what current AI systems can and cannot do.
Much of that happens through Pacman. Over the semester you build agents that play the game: searching for a path, playing against ghosts, tracking ghosts they cannot see, and learning to play by imitation. It is a toy world on purpose, small enough that you can see exactly why your agent behaves as it does.
Prerequisite knowledge and skills
Programming experience in Python. Probability and statistics.
Planned learning activities and teaching methods
- Theoretical lectures
- Exercise sessions
- Programming projects on the Pacman game: search and adversarial search as a tutorial, then a Bayes filter project and an imitation learning project
Mode of delivery (face to face, distance learning, hybrid learning)
Face-to-face course
Course materials and recommended or required readings
Other site(s) used for course materials
- Github (https://github.com/glouppe/info8006-introduction-to-ai)
Further information:
Slides are made publicly available on GitHub during the semester.
The course will be based on "Artificial Intelligence: A modern approach", Stuart Russell, Peter Norvig, Third Edition, 2010. This book is highly recommended.
Exam(s) in session
Any session
- In-person
written exam
Written work / report
Further information:
The evaluation has two components:
- Programming projects: 30%
- Written exam: 70%
The projects are not graded on the submitted code alone. They also include an oral defense, where you answer direct questions on your own code, justify your design decisions, and may be asked to modify or extend your implementation. Submitting the projects is mandatory to sit the exam.
A minimum of 8/20 is required for the written exam. If you reach it, the final grade is the weighted average of the two components. If you do not, the exam grade becomes the final grade.
Use of generative AI tools
Using AI assistants is allowed, without restriction. The rule is simple: everything you submit, you must be able to explain. Submitting a solution you do not understand, because an AI assistant produced it, will be strongly penalized. This is why the evaluation is built the way it is: a written exam you sit alone, and questions on your project asked with you in the room. A model can write your Pacman agent, but it cannot sit the exam, explain why your Bayes filter updates the way it does, or modify your implementation on request.
Work placement(s)
Organisational remarks and main changes to the course
The website for the course is https://github.com/glouppe/info8006-introduction-to-ai
Contacts
- Teacher: Prof. Gilles Louppe (g.louppe@uliege.be)
- Assistants: info8006@montefiore.ulg.ac.be