Training Data
Etymology
A plain compound: "training" from the process of teaching a model by exposure, and "data" for the examples it studies. The phrase carries real weight because a model can only learn what its training data contains: gaps, skews, and errors in the data become gaps, skews, and errors in the system. This is the origin of the maxim "garbage in, garbage out."
noun phrase · /TRAY-ning DAY-tuh/
The collection of examples an AI system learns from, which shapes everything it can and can’t do.
What it names
The raw material of AI — and the honest source of its biases, blind spots, and copyright fights.
Example
The hiring tool discriminated because its training data reflected years of biased past decisions.
Related idea
Training data is what a model learns from during construction, distinct from the prompt or input it receives later in use — you can change a prompt anytime, but the training data is baked in.