Machine translation

Automatically converting text from one language to another — historically dominated by phrase-based and neural systems, now overwhelmingly handled by LLMs.

Machine translation (MT) is the task of automatically converting text from a source language to a target language. Modern frontier LLMs handle MT extremely well — often comparable to or better than dedicated translation systems for general use, especially on idiomatic phrasing and context-aware translation. It matters because translation is one of the highest-volume real-world AI tasks. Google Translate, DeepL, and the translation features inside Microsoft Office, Apple Translate, and almost every consumer app rely on neural translation under the hood. For Chinese-English work specifically, LLM-based translation is now the default for serious use — the quality often exceeds older specialized neural MT systems. A concrete example: pasting a paragraph of formal English business writing into Claude and asking for translation into Traditional Chinese (繁體中文) yields output appropriate for Taiwan business audiences with proper register. Asking for Simplified Chinese yields a different version with mainland conventions. The LLM handles register, idiomatic equivalents, and context that older MT systems struggled with. For specific languages or domains, dedicated MT systems (DeepL, Google's Neural Machine Translation, NLLB) sometimes still win on edge cases. But for most production use, calling Claude or GPT-4 with a clear prompt is now the simplest, highest-quality path. Related: BLEU, text generation, multilingual.