This page is the practical orientation: the rhythm of each week, what the assessments look like, how to navigate the materials, and what to expect if you're a UCT student versus a self-learner reading along. For the bigger picture — who built this, why, and the licensing — see the About page.
New Here? Start in Three Steps
- Read this page. It tells you the weekly rhythm, the time each week takes, how the work is assessed, and how to find your way around.
- Open the contents page and start at Week 1. The weeks are meant to be read in order — Weeks 1–4 lay the foundations everything else builds on.
- End each week with the activities. Every week finishes with a hands-on "Activities and Assessment" page and a "Discussion Questions" page. Doing those — not just reading — is where the learning actually happens.
The Weekly Rhythm
Each of the twelve weeks follows the same three-phase structure. The phases are designed to be done in order, but each one stands alone if you only have time for part of the week.
Pre-Class
Readings, short videos, and a few framing questions. This is where you build the conceptual ground for the week. Most weeks have 2–4 core readings plus optional supplementary material. All readings are openly accessible.
In-Class
Discussion, hands-on activities, worked examples, and group work. For UCT students this happens in the scheduled session; for self-learners, the activity prompts are written so you can run them yourself or with a study group.
Post-Class
Deeper reading, an assignment, or a reflective task. This is where the week's ideas get applied to your own research questions and where most of the assessment work happens.
Time and Workload
Honest estimates, so you can plan. These cover reading plus the week's hands-on activity and discussion preparation — not the open-ended project work, which is on top.
- Per week: roughly 3–4 hours for most weeks. The lighter, more conceptual weeks (2 and 3) are nearer 2–2.5 hours; Week 1 runs a little longer because it sets the foundations and includes several videos. The estimate for each week is shown next to it on the contents page.
- Whole course: roughly 35–45 hours across the twelve weeks, plus the Week 12 capstone (about 3 hours) and the research-enhancement project, which is deliberately open-ended and scales with how far you take it.
- If you only have part of a week: the three phases (pre-class, in-class, post-class) each stand alone. Reading the pre-class material on its own is still worthwhile.
Assessment
The course is graded across three components. The weighting is deliberately balanced so that no single deliverable dominates.
| Component | Weight | What it covers |
|---|---|---|
| Weekly practical exercises | 40% | Short, hands-on tasks tied to each week's content — verifying citations, running an analysis, auditing AI-generated text. Submitted progressively across the term. |
| Research enhancement project | 40% | An applied project using AI tools to support a piece of your own research, with a final presentation. Combines the practical skills built across the course. |
| Personal ethical framework | 20% | A written framework for your own use of AI in research, developed iteratively from Week 4 onwards and grounded in the ethical lenses introduced there. |
How to Navigate the Materials
The landing page lists every week and every sub-lesson in order. A few things worth knowing as you work through them:
- Core vs supplementary pages: sub-lessons marked (supplementary) in the contents list are deeper dives or adjacent topics. They are useful but not required for the assessments.
- Linear vs cherry-picked: the weeks build on each other — particularly Weeks 1–4 (foundations and ethics) which underpin everything that follows. Weeks 5–8 (literature, writing, code, multimodal) can be read in any order once you've done the foundations.
- Hands-on activities: every week ends with an "Activities and Assessment" page. Even if you're not formally enrolled, working through these is by far the most effective way to internalise the material.
- External links: all open in a new tab. Where readings are paywalled, we link to an open-access version (preprint, postprint, or institutional copy) wherever one exists.
Who This Page Is For
UCT students enrolled in MAM5020F
The authoritative versions of these materials, plus all assignments and grading, live on Amathuba. This site mirrors the lesson content for ease of reference and revision — treat Amathuba as the source of truth for deadlines and submissions.
Self-learners and other educators
Everything you need is on this site. The materials are released under CC BY 4.0, so you are welcome to read, adapt, and reuse them in your own teaching with attribution. See the About page for the suggested citation.
Prerequisites
None. The course is pitched at NQF Level 9 (postgraduate) but assumes no prior background in machine learning, computer science, or programming. What it does assume is that you are an active researcher or graduate student with a research project, problem, or question of your own to bring to the material.
The hands-on weeks (especially Weeks 5, 7 and 8) involve using AI tools directly. You will need a laptop, a free or paid account with at least one frontier AI assistant (Claude, ChatGPT, or Gemini), and a willingness to try things and break them.
Frequently Asked Questions
Do I need to pay for an AI tool?
No. The whole twelve-week core is designed to run on free tiers — this is a deliberate choice the course argues for in Weeks 10 and 11, because access is not a detail. The only exception is the optional Advanced Track, which uses a paid tool and says so plainly up front.
Which AI assistant should I use?
Any current frontier assistant — Claude, ChatGPT, or Gemini — is fine for almost everything here. The course is deliberately tool-agnostic and talks about model families rather than chasing version numbers, because the specifics change monthly. Where a particular tool matters, the lesson says so.
Do I have to do the weeks in order?
Do Weeks 1–4 first — the foundations and the ethical framework underpin everything else. After that, Weeks 5–8 (literature, writing, code, multimodal) can be taken in any order, and Weeks 9–11 build toward the capstone.
Is it safe to use my own research data in the exercises?
Treat that as a real question, not a formality. Do not put confidential, identifiable, restricted, or embargoed data into a third-party AI tool during the exercises — use fictional or non-sensitive material instead. Weeks 4 and 11 cover the ethics and disclosure side of this in depth.
A model or tool mentioned here looks out of date.
It probably is — and that is itself a lesson. The capability claims are dated snapshots, deliberately calibrated, and Week 9 teaches the habit of checking the primary source rather than trusting a figure a page wrote months ago. If a specific link is broken or a claim is wrong, please tell us (below).
Getting Help and Reporting Problems
Found a broken link, a wrong citation, an out-of-date claim, or any other error? Email jonathan.shock@uct.ac.za or open an issue on the course GitHub repository. Corrections are genuinely welcomed and made promptly — a course about AI hallucination would be a poor one to leave errors in.
UCT students: for anything to do with deadlines, submissions, or grades, Amathuba is the source of truth — use the channels there rather than email for those.