The three terms describe nested circles, each a subset of the one before.
Artificial Intelligence
The broadest term: any technique that lets computers perform tasks that seem to need intelligence. It includes hand-written rule systems, search algorithms and planning as well as learning systems.
Machine Learning
A subset of AI in which systems learn from data instead of following explicit rules. You provide examples — emails labelled spam or not, houses with their sale prices — and an algorithm fits a model that generalises to new cases. Classic machine learning includes linear regression, decision trees, random forests and gradient boosting, which remain the workhorses for tabular business data.
Deep Learning
A subset of machine learning that uses neural networks with many layers. Deep learning shines on unstructured data such as images, audio and text, where it can learn useful features directly from raw inputs. It usually needs more data and computing power than classic methods.
Where Generative AI Fits
Generative AI — systems that create text, images, audio or code — is built with deep learning, most often transformer models. Large language models are the best-known example.
Choosing Between Them
For a spreadsheet of customer records, classic machine learning is often best. For photos, speech or free text, deep learning or a pretrained model usually wins. Start with the simplest approach that could work.