Torchnet – Lighting the way to deep machine learning

I don't use Facebook products because I don't want to give Zuckerberg my money. I don't use Apple because I don't want to give money to a company whose original figurehead thought that wearing Mom-jeans to press conferences was in any way professional.

The point is that these companies, however monopolistic they might seem to you (and me), are driving innovation several orders of magnitude faster and more efficiently than organizations like Mozilla or GNU. A company exists to make money. That's an ethos that I personally find more consistent and more comfortable to think about than "we should make software that EVERYONE can use" and it's a safer bet as well, since these companies will go out of business if they don't make money whereas GNU/Mozilla/et al will vomit out software that sorta works for a million years.

Also, machine learning is not about predicting the future anyways, it's almost exactly the opposite. Understanding what people have done is much more actionable than trying to figure out what people are GOING to do, because we are very fickle, complex, imaginative creatures. Could you have predicted that I'd write this comment? No. Could a program that has read every previous comment that I've ever made predict this comment? No. Could it predict that I might write a comment of a similar nature? Sure. If we were playing poker and I thought that you were going to fold, that's much different from some algorithm that was CERTAIN you were going to fold.