It's pretty easy to use some ML-as-a-Service platform, where you simply hit an endpoint in Azure, AWS, etc. without any prerequisite knowledge of the underlying

It's pretty easy to use some ML-as-a-Service platform, where you simply hit an endpoint in Azure, AWS, etc. without any prerequisite knowledge of the underlying models powering it. Although asking someone to understand all the underlying math is a bit much, I think there's a middle ground here.