Instead of dictating to a programmer every single rule for spotting spam in e-mail (“if it contains the word ‘prize’ it is spam; if it contains ‘invoice’ it is not”), you hand the computer an enormous pile of e-mails and merely show it which of them were spam in the past and which were not. That is machine learning.
Machine learning is the process by which a computer works out for itself, from these examples, what the typical marks of spam are. It does not memorise the individual e-mails; it looks for patterns in them. In time it becomes so good that it can identify spam with high accuracy even in a completely new message it has never seen before. In essence it learns from experience, much as a person does.
Machine learning is a branch of artificial intelligence focused on developing algorithms and statistical models that allow computer systems to “learn” from data. Instead of being explicitly programmed to carry out a specific task, machine learning algorithms are trained on large datasets in order to recognise patterns, correlations and hidden structures within them. The aim is to teach the model to generalise these findings so that it can make accurate predictions or informed decisions on the basis of new, previously unseen data. The basic paradigms include supervised learning, unsupervised learning and reinforcement learning, each addressing a different type of problem.