Department of Information and Computing Sciences

Departement Informatica Onderwijs
Bachelor Informatica Informatiekunde Kunstmatige intelligentie Master Computing Science Game&Media Technology Artifical Intelligence Human Computer Interaction Business Informatics

Onderwijs Informatica en Informatiekunde

Vak-informatie Informatica en Informatiekunde

Logics for safe AI

Course code:INFOMLSAI
Credits:7.5 ECTS
Period:period 4 (week 17 through 26, i.e., 26-4-2021 through 2-7-2021; retake week 28)
Participants:up till now 20 subscriptions
Schedule:Official schedule representation can be found in MyTimetable
innovatie          Natasha Alechina
Brian Logan
lecture          Natasha Alechina
Brian Logan
Contents:This course is about ensuring the safety and reliability of autonomous AI agents and multi-agent systems. In order to guarantee that the behaviour of a system achieves its objectives, we use formal proofs rather than empirical studies, and either formally verify that the system behaves in accordance with the specified objectives, or automatically synthesise provably correct behaviours from specifications of the system objectives.

The formal techniques for doing this include epistemic and temporal logics and their combinations, and constitute the main technical content of the course. The emphasis is on mastering these techniques, their computational aspects, and the use of tools implementing them to verify and synthesise AI agents and multi-agent systems. In particular, applications of temporal logic in automated planning and specifying reward functions in reinforcement learning will be discussed.

The course also prepares students for undertaking research on formal aspects of artificial intelligence, and provides the foundation for undertaking Masters projects on developing safe and reliable AI systems. Lab sessions will introduce students to relevant specification and modelling techniques and the use of tools such as MCMAS and SynKit for the verification and synthesis of AI agents and multi-agent systems.

Literature:We will use the following free textbook:

It will be complemented by papers related to particular topics of the course, all available online.

Course form:Two lectures a week and a lab/practical session for mastering techniques and tools for the specification, verification and synthesis of AI agents and multi-agent systems.

In the current 2020-2021 academic year, lectures are on Monday and Wednesday mornings, and practical sessions are after the Wednesday lecture. The first lecture will be live online on Monday at 11:00, after that we will provide lecture recordings and live Q&A sessions on Monday at 11:00 and on Wednesdays at 10:00.

Exam form:One written open-book exam at the end of the course, taken online.

We will provide sample questions and rehearse the exam procedure in a mock exam half way through the course.

Minimum effort to qualify for 2nd chance exam:The original exam grade must be at least 4, or an AANV; and satisfactory completion of at least 50% of the coursework assignments.