Natural Language Processing (NLP)

Overview

Natural language processing, or NLP, is the long-running effort to let computers work with human language, a notoriously messy, ambiguous, context-dependent thing. It covers a wide span of tasks: understanding the meaning of text, translating between languages, summarizing documents, answering questions, and generating fluent prose. The field predates modern AI by decades but was transformed utterly by deep learning and the transformer.

How it works

Where NLP once relied on hand-written rules and brittle statistical methods, it now rests overwhelmingly on neural networks trained on vast quantities of text. Models learn the patterns of language, grammar, meaning, and even reasoning, directly from data, by being trained to predict and generate text. The transformer architecture, with its attention mechanism, is what allowed these models to capture long-range context and scale to their current capability.

Why it matters

NLP is where much of the recent AI revolution has played out, because language is how humans encode so much of their knowledge and intent. Large language models are NLP systems, and they drive the single largest share of demand for AI compute today. For regions building local capability, the ability to train and adapt language models for their own languages and needs is especially consequential.