In order to perform deep learning, a machine has to have a formalized mathematical representation of what "correct" looks like. This is what the model optimizes for in order to minimize. This means that there has to be a codifable loss function in order to point machine learning at a problem. **** # More ## Source - Huberman Lab's podcast episode with Lex Friedman ## Related - [[Artificial Intelligence]] - [[Mean Squared Error]] - [[Math & Fairness]]