Well, one of the reasons is that the technological stuff behind a database is already highly efficient and makes use of those fancy data structures. How else could you retreive big data sets fast with a simple SELECT statement? Indexing makes use of Balanced Trees (or other search algorithms) and understanding these concepts helps in getting what you need to feed your models fast.

Additionally, database technology evolves and there are powerful database systems out there that suppport complex data structures and machine learning algorithms (Snowflake for instance). So saying that RDBMS are 'dusted' falls a bit short.