Mostly Helpful Econometrics. Why machine learning researchers should… | by Mikey Shulman | Kensho Blog

When asked how he would spend an hour to save the world, Albert Einstein famously said that he would spend 55 minutes defining the problem and 5 minutes solving it. Einstein’s suggestion has profound lessons that apply to all fields, in particular to machine learning. Historically, machine learning and econometrics¹ have been two sides of the same coin. Both fields use similar techniques (e.g., logistic regression) to solve similar prediction problems, albeit with a completely different vernacular and jargon. The key difference between machine learning and econometrics is the approach to framing and understanding a problem. Recently, practitioners of these two schools of thought have drifted away from one another. Machine learning researchers favor exploiting the abundance of easily obtained data to train larger, more powerful, more complex, and less interpretable models, and econometricians favor cleverly crafted experiments and more interpretable models. Accordingly, we see big tech

When asked how he would spend an hour to save the world, Albert Einstein famously said that he would spend 55 minutes defining the problem and 5 minutes solving it. Einstein’s suggestion has profound lessons that apply to all fields, in particular to machine learning. Historically, machine learning and econometrics¹ have been two sides of the same coin. Both fields use similar techniques (e.g., logistic regression) to solve similar prediction problems, albeit with a completely different vernacular and jargon. The key difference between machine learning and econometrics is the approach to frami