AI translation
AI translation is the automatic translation of text by artificial intelligence models, most often large neural networks. In a professional workflow it produces the first draft at speed and low cost, which a qualified human then reviews and corrects before anything is published.
How it works
AI translation systems learn from very large collections of human-translated text, then generate a translation one piece at a time by predicting what a human translator would most likely write. Today's systems are neural networks, and the most recent are built on the same large language model architecture as general AI assistants.
Because they predict rather than reason, they are fluent but not reliable on their own. They can produce confident, natural-sounding output that is subtly or badly wrong, which is why professional use always keeps a human in the loop.
How SourceTarget uses it
SourceTarget uses AI translation to produce the first draft, then routes it through metrics and human review. Confidence and quality scores flag where the AI is least trustworthy, so editors spend their time on the risky segments.
The AI is treated as a fast drafting tool, never as the final author of anything that gets published.