Databricks explains CPU vs GPU for AI workloads

Video outlines a simple mental model: CPU for data work (ETL, SQL, classic ML) and GPU for deep learning (fine-tuning LLMs, computer vision). Focus is on token economics rather than training from scratch.