Machine Learning is the new Algorithms
nlpers.blogspot.com
nlpers.blogspot.com
GIGO still applies, as it always has in engineering problem-solving. But if you take care to do the basics, like clean your data and find good features, you can use some ML methods as components with somewhat known characteristics (performance, flexibility, runtime).
It used to be that you needed a skilled person to guide you to a decent FFT or matrix library, and then core numerical software was commoditized. Same for data structures. It's fun to observe that some of this is happening to ML as well.