Inference is the work a trained AI model performs when it receives a new input and produces an output. It is distinct from training, and it is where the size of a request, the desired quality and the available capacity begin to shape the experience of using an AI system.
A large language model generates text from patterns in its training and from the context it receives in a request. It can produce useful drafts, summaries and transformations of text, but it does not independently establish that an answer is true, current or appropriate.
Neural networks are machine-learning models that learn patterns by adjusting many numerical parameters. Their output can be useful in complex tasks, but it is the result of learned calculation, not human-like understanding.
Machine learning is a way of building systems that learn patterns from examples. The result can be a prediction, classification or recommendation, but its usefulness depends on the examples, the objective and the setting in which it is used.
Artificial intelligence is an umbrella term for systems that recognise patterns, make predictions, work with language or generate new material. Those activities use different methods, which is why ‘AI’ alone rarely tells us what a system can do or how it should be judged.
