Training Data

Fundamentals

The examples an AI system learns from - text, images, audio, or numbers collected into large datasets.

AI models learn by finding patterns in training data. A translation model learns from millions of translated sentence pairs. An image generator learns from billions of captioned pictures.The quality of training data decides the quality of the model. If the data is wrong, outdated, or unbalanced, the model inherits those flaws - which is where many AI mistakes and biases come from. It also explains why languages with little online text, including several Balkan languages, are often supported poorly.