Through the Policy Innovation Lab, Stellenbosch University’s School for Data Science and Computational Thinking is offering a new short course, Fundamentals of Large Language Models, designed to give students from non-STEM backgrounds a practical understanding of how Large Language Models (LLMs) are built, deployed and evaluated.
The course responds to the growing role of AI across fields beyond Computer science, including but not limited to the Humanities, Commerce, Law, Education and Theology fields. It is aimed particularly at students who may encounter AI systems in their studies and future careers but do not have a background in programming, mathematics, statistics or computer science.
The course aims to introduce the participants to concepts including tokens, transformer processing, training data, model deployment, retrieval and AI agents, hallucinations, verification, benchmarks and evaluation. This will help them understand how to interpret AI system documentation and assess claims about the capabilities and reliability of AI systems. A key focus of the course is developing the ability to distinguish between what the evidence about an AI system establishes and what it does not. Participants will work with system cards and evaluation evidence to develop a more informed understanding of the strengths, limitations and appropriate use of LLMs.
The course will be presented by Prof Willem Fourie, Chair of the Policy Innovation Lab, and Dr Gray Manicom, both from the School for Data Science and Computational Thinking at Stellenbosch University. Through four sessions, participants will explore how language models work, how they learn from data, how models become deployed AI systems, and how their capabilities and limitations can be evaluated. The course is delivered fully online and synchronously, with recordings made available afterwards. No prior quantitative or technical background is required, and participants do not need to purchase any prescribed texts. The first presentation will take place from 28 September to 1 October 2026, with sessions running from 17:00 to 19:00 each evening. The course carries a proposed two credits and 20 notional learning hours for internal course-design purposes; these credits do not accumulate towards a formal qualification.
Participants who meet the assessment requirements will receive an SU Certificate of Competence in Fundamentals of Large Language Models. Those who complete the course but do not meet the assessment requirements will receive a certificate of attendance. With no prior technical background required, the course provides students across disciplines with an opportunity to build foundational AI literacy and develop the skills to critically interpret and interrogate claims about large language models.