Artificial Intelligence & Machine Learning (AI/ML)

1. What AI and machine learning really are

Artificial intelligence is the field; machine learning is the technique that dominates it today; and neither is magic, consciousness or a search engine with ambition. This lesson separates the vocabulary that marketing has muddled — AI, ML, deep learning, generative AI, agents — traces how the field arrived at large language models, and sets honest expectations about what statistical learning can and cannot do.

2. The machine learning workflow

Most failed AI projects fail before any model is trained. This lesson covers the workflow that prevents it: framing the problem as a measurable prediction with a decision attached, preparing and splitting data honestly, choosing a baseline model before a complex one, evaluating with the metrics that match the cost of errors, watching for overfitting and data leakage, and monitoring a deployed model as the world moves away from its training data.

1. Neural networks and generative AI

This lesson opens the box far enough to reason about behaviour: what a neuron computes, why layers build representations, what training actually adjusts, what a transformer's attention and context window mean, and how diffusion models make images. Then the practical half — prompting, retrieval-augmented generation with citations, fine-tuning versus prompting, and the cost, latency and evaluation realities of building products on top of models you did not train.

2. Ethics, risk and deploying AI responsibly

An AI system that works is not yet an AI system that is acceptable. This lesson covers the harms that are specific to machine learning — bias against groups, privacy loss, opaque decisions, confident errors and environmental cost — the practices that reduce them (data governance, documentation, human review, explainability, evaluation per subgroup), and how the regulatory picture in Ghana and internationally is shaping what teams must be able to show about a model that affects people.