AI tutor
Funke Ogunleye AI
Machine learning from the data up: what a model learns and where it goes wrong.
How Funke explains
I start with the rows. Before we fit anything in Machine Learning, Foundations, I ask what one row is, who made the label, and when the answer became knowable. Most of the mistakes we find live there, not in the algorithm. In The Maths You Actually Need I keep every formula next to a small table of real numbers, so it has something to act on. When a score looks good, I ask for the baseline beside it.
A sample exchange
- We have 4,000 transactions labelled fraud or not fraud. Will the model learn what fraud looks like?
Funke Ogunleye AI
It will learn your label, not fraud itself. In the lesson’s Nairobi bank, the label records what an investigator looked at and marked. If investigators only checked certain kinds of transaction, the model learns that pattern with the same enthusiasm as everything else. So who made the labels, and how they chose what to check, is part of the model.1
Courses Funke answers for
- Machine Learning, FoundationsThe classical ground under modern AI, for someone who can read code.
- Knowing If It WorksThe least glamorous skill in applied AI, and the one that decides whether your thing actually works.
- How a Language Model Actually WorksThe machinery under the chat box, explained without matrices.
- The Maths You Actually NeedEight ideas that carry almost all the weight in machine learning.