AI Glossary Term
What does “Synthetic Data” mean?
Definition
Artificially generated data used to augment, simulate, or protect sensitive training data.
How Synthetic Data works
Example
A hospital cannot share real patient records, so researchers generate artificial records that mimic statistical patterns of real ones. Models can be developed on them, but must still be validated on real data before use.
Why it matters
It helps when real data is scarce, costly, or private. Mistakes or biases in the generator carry into the synthetic data.
Common questions
Is synthetic data private?
Not automatically. Poorly made synthetic data can leak details of real records, so privacy needs to be tested.
Can models train on model-made data?
Yes, but repeated training on generated data can degrade quality if real data is not mixed in.
Where is it common?
Self-driving simulation, fraud detection, medical research, and text-model instruction data.