“In machine learning, computers apply statistical learning techniques to automatically identify patterns in data. These techniques can be used to make highly accurate predictions.”
Principal Component Analysis is an example of a very low-level sort of machine learning algorithm. By identifying things that are probably (within some statistical certainty) in a training dataset, you can identify the same things in a real-world dataset. This is why Google’s Deep Dream needs a training image — it needs to know what to look for.