Is There Anything Inherently Inefficient About Machine Learning?

There's an attitude about machine learning that if i'm not too worried about my resources, just add as many features as i can. Well, from the point of view of your model, what are they doing inefficiently that you're able to capitalize on? The thing is, basically, those models are not customizing the size of the tables to their data. We want to ship a small enough packet. Do i still need to worry up front about having a good, well structured feature set that's somewhat minimized? Or can i just rely on your approach, like a post processing technique, to compress me down where i need to be? I think absolutely. Our approach helps but it's