Also known as: ML
In plain English
Instead of programming every rule, you show the computer thousands of examples and it works out the patterns itself. Show it enough spam and normal emails, and it learns to tell them apart.
In practice
Machine learning powers churn prediction, demand forecasting, fraud detection and recommendations. Its quality depends heavily on the data: a model trained on last year's customers may perform poorly when customer behaviour changes.
Under the hood
ML algorithms fit a model's parameters to minimise a loss function over training data, then generalise to unseen inputs. The main paradigms are supervised learning (labelled examples), unsupervised learning (finding structure) and reinforcement learning (learning from rewards).
Example
"The bank's machine learning model flagged the transaction as likely fraud."