Select a course.
Artificial Intelligence and Machine Learning
infoWhy this? A critical model of rule-based and learning-based AI helps us understand how systems are trained, evaluated, and used, as well as why their outputs have limitations. We learn to interpret performance evidence and assess reliability, bias, fairness, and the need for human oversight.
scheduleWhy now? Prior learning about algorithms, programming, data processing, and testing provides the foundation for reasoning about features, labels, datasets, and model performance. We can now move beyond using digital systems to evaluating how data-driven systems make predictions and where they may fail.
neurologyYou need to know
- AI includes both rule-based systems and learning-based systems.
- Machine learning infers patterns from data rather than explicit rules.
- Generative AI is a type of learning-based AI.
- Supervised learning uses features as inputs and labels as target outputs.
- Training data are used to fit a model, validation data are used to compare or tune models, and test data are reserved to estimate final performance on unseen examples.
- Overfitting means strong training performance but weak generalisation.
- Underfitting means the model is too simple to capture useful patterns.
- A confusion matrix counts true and false positives and negatives.
- Accuracy can be misleading when class distributions are imbalanced.
- Precision and recall capture different error trade-offs.
- Bias can enter through sampling, labelling, and deployment context.
- Generative language models generate text by predicting probability distributions for the next token from patterns learned in training data; this does not guarantee factual accuracy.
- High-stakes AI use requires human oversight and verification.
rocket_launchYou must be able to
- Distinguish rule-based systems from learning-based systems, and identify generative uses within learning-based AI.
- Interpret simple model-performance evidence beyond accuracy alone.
- Evaluate AI outputs for reliability, fairness risk, and appropriate use.