Machine learning (ML) is the study and use of algorithms that improve with experience—typically by learning patterns from data rather than from explicit programming. Supervised learning infers mappings from labelled examples; unsupervised and self-supervised methods find structure in unlabelled data; reinforcement learning optimises behaviour through trial and feedback. ML underpins modern AI applications: recommendation systems (e.g. Netflix, Amazon), image and speech recognition, language models, fraud detection, predictive maintenance, and many decision-support tools. Frameworks and platforms from Google, Microsoft, Amazon, and open-source communities have made training and deployment more accessible.
ML addresses problems where rules are hard to specify but data is available: personalisation, forecasting, anomaly detection, and automation of perception or judgement. It is widely adopted in finance, healthcare, retail, manufacturing, and media. Performance depends on data quality, quantity, and relevance; biased or unrepresentative data can lead to biased or unreliable behaviour. Interpretability and explainability remain active research areas, especially for high-stakes applications. As models grow in size and scope, compute and data requirements have concentrated capability among well-resourced actors.
Deployment continues to expand into regulated and safety-critical domains, where assurance and accountability are increasingly demanded. Research continues into more data-efficient learning, robustness to distribution shift and adversarial inputs, and alignment with human intent. ML will remain a core enabler of AI-driven products and services and a focus of both investment and governance.