Training Data

Training data is the collection of examples an artificial intelligence (AI) model learns from, and it shapes almost everything the model can and cannot do. A model taught only on formal writing will struggle with slang, because it simply never saw it.

It is like the ingredients and recipes an apprentice cook learns from: the range of what they practise decides the range of what they can make. Training data matters because gaps or bias in what goes in show up as gaps or bias in what the model does, so the examples deserve as much care as the technology itself.