The fuel of ML is data. The entire lifecycle we saw yesterday ultimately depends on "what data we have and how we handle it." So today we tackle data itself head-on. We'll organize what kinds of data exist (structured/unstructured), why we don't use data as one chunk but split it into three pieces (training/validation/test), and why "data quality" determines the fate of a model.
These three topics appear frequently in AIF-C01, both directly and in scenario form.
The most basic way to divide data is whether it has a predefined structure.