Machine learningorDeep learning
Machine learning is the set of methods that derive their behaviour from examples rather than from written rules. Deep learning is one family within it, the one that stacks many layers of computation, and generative AI is a narrower subset still.
Point by point
| Criterion | Floor 1Floor 1 · The Modela solid block on its own: the prediction machineMachine learning | Floor 1Floor 1 · The Modela solid block on its own: the prediction machineDeep learning |
|---|---|---|
| Relationship | Contains the second | Contained in the first |
| Features | Chosen by a person | Learnt by the layers |
| Data volume required | Moderate | Considerable |
| Hardware | An ordinary server | Accelerators |
| Explainability | Often readable | Structurally opaque |
| Home ground | Tabular data | Text, image, sound |
On the ground four situations
Predicting a risk of non-payment from thirty columns of a spreadsheet.
Floor 1 · The Modela solid block on its own: the prediction machineMachine learningOn tabular data, the classical methods are often more accurate, faster and readable, which counts when a refusal has to be justified.
Spotting an anomaly on an X-ray.
Floor 1 · The Modela solid block on its own: the prediction machineDeep learningNobody could say by hand which characteristics to measure on an image: that is exactly what the layers learn to do.
A company picks a deep network for a forecasting problem on a spreadsheet, because that is the word going around.
Floor 1 · The Modela solid block on its own: the prediction machineMachine learningThe useful counter-example: it will pay three times more, wait longer, lose the explanation, and often end up with a poorer result.
A language model, trained on trillions of tokens.
BothIt is deep learning, hence machine learning: the circles are nested, and both labels are true at the same time.
The test that settles it
Ask who chose the characteristics being measured. If it was a person, column by column, you are in classical learning. If it is the layers of the network, you are in deep learning.
The trap
Treating these words as successive generations, the most recent making the others obsolete. They are nested circles: most systems in production still belong to the outer circles, and it is those that carry the most sensitive decisions.
Check click your answer
Level 1 · Recognise
What is the relationship between machine learning and deep learning?
Level 2 · Distinguish
Compared with classic machine learning, what does deep learning make unnecessary?
Level 2 · Distinguish
Why is choosing depth by default a frequent mistake?