K-Culture GlossaryㄷIndustry and policy
Data Labeling
The work of attaching correct answers to data used for training AI. It's the invisible human labor that raises AI.
In plain words
Data labeling is the process of attaching correct-answer tags to data used to train AI. Writing "cat" under a photo of a cat, marking a sentence's sentiment as "positive," or tracing the outline of a pedestrian in a self-driving car's video feed frame by frame — all of this is labeling.
Think of it like a teacher preparing an answer key for a workbook. AI learns by looking at this answer key, figuring out "ah, this kind of photo is a cat." So if the labels are wrong, the AI learns the wrong things too. That's why the quality of labels directly determines the quality of the AI.
Most of this work is still done by humans — the global data labeling industry has grown into a massive market employing hundreds of thousands of workers, and discussions about working conditions and fair pay are growing louder. In the news, "annotation" refers to the same thing.
How it shows up in the news
"Tens of thousands of labeling workers are attaching correct answers to AI training data" — an article shining a light on the human labor hidden behind AI.
See also
Stories using this term
No story has used this term yet. New ones attach here automatically.