I wonder if this has potential for solving practical optimization problems. Why glue together some ML model when you already have a premade solution with millio

I wonder if this has potential for solving practical optimization problems. Why glue together some ML model when you already have a premade solution with millions of years of training time by mother nature