How AI Works
Training Data
SHARE THIS TERM
In Plain English
Training data is the information used to teach an AI model to recognise patterns and develop its capabilities. Depending on the model, this may include text, images, audio, code or other types of data.
A Real-World Example
A company develops a model to identify different types of customer enquiries. It trains the model on a carefully prepared collection of example enquiries and their correct categories so it can learn patterns useful for classifying new messages.
Why It Matters
The quality, relevance and composition of training data can affect what a model does well, where it struggles and which biases it may reproduce. Data rights and privacy also matter when training or adapting models.
A Common Misconception
Training data is not a reliable, searchable library of facts inside the model. A model may learn patterns from its training without accurately recalling individual documents or knowing whether information is current.
Practical Takeaway
Ask whether training data is appropriate, representative and responsibly sourced; use current, traceable references when factual accuracy matters.
When You’ll Encounter It
When discussing how models learn patterns, their capabilities and possible biases.
What to Look Out For
A model's training material is not a reliable searchable catalogue of original sources.
How to Handle It
Use current, traceable sources for facts instead of assuming the model remembers its training accurately.