Fine-tuning

Fine-tuning means taking an AI ✕ Artificial intelligence is a computer system’s ability to perform tasks we associate with human intelligence, such as understanding language, recognizing images or making decisions. More in the glossary → model that has already been generally trained and continuing to train it on a smaller, specialized dataset, to adapt it to a specific task, industry or style. Instead of building a model from scratch —which is extremely expensive— you build on what it already knows and teach it, for example, a brand’s tone, the jargon of a medical field, or how to classify a support team’s tickets. This yields more accurate and consistent answers for that use case. It is an alternative to techniques like detailed prompts or RAG ✕ RAG connects a language model to an external data source, so it answers based on real, up-to-date information instead of only its memory. More in the glossary → , and the approaches are sometimes combined. It requires quality data and some technical know-how, but it allows models much better tuned to a real need.