As far as I can tell, the idea has always been the same: more relevant ads means less ads needed to make the same revenue and cost businesses less in wasted ad budget.
The "old method" was to just make your ads more obnoxious and abundant since you could only compete through volume (both in the sense of "loudness" and "quantity"). Now if I open a shop in my city, I can advertise on Google. I tell Google that I want to pay for X ads and to show them to people in my city (potential customers) who have searched for similar items (more likely customers). This way I can minimize the amount of money I spend showing ads to people in other cities/countries or who aren't likely to be interested. Google can charge more for that space because it's more likely to result in a customer. Google makes more money from fewer/less obnoxious ads (no popups, flashy shit, etc). And as a web user, I don't see ads for diapers or restaurants in Wyoming because Google's algorithms have determined that I live in an east coast US city and have never searched for baby stuff.
Honestly the only real issue I have with their "big data" is that it might be captured by another organization (business, "hacker", or government) and used in less benign ways than tweaking the ads in my sidebar.
They most certainly do when there are national security letters involved.
You may argue that this was not Google's fault - it is uncertain how much choice they had with regards to providing access for PRISM, and they're welcome to the tiny compensation they received. The point is that they had the data at all, thus allowing it to be copied y others, regardless of google's intentions
