One thing I think many people miss in such privacy debates is that Google knows things about you that you don't know about yourself. (Clearly the sense of "know" in each case has to be somewhat different.)

For example, they probably know an enormous amount about your sleep habits (not just when, but where! ← from IP address data) and your travel (← from IP address data, not just Maps). There's plenty of noise in that data, but it's been collected over a long time.

They remember searches and interactions from many years ago that you don't. They might know about a fight or conflict with someone that you had in the past that you've forgotten. Thanks to Maps, they might know about excursions that you've taken that you no longer remember.

They may be able to determine things about the state or evolution of your relationships with other people when you're not consciously aware of those things yourself, in terms of patterns in how and when and how often you contact each other, and perhaps with what degree of affection or intimacy. (Obviously people often know a lot about the state of their own relationships, but they might not consciously notice certain kinds of gradual changes.)

If there are inferences that can be made from your searches, they might be able to make those inferences when you haven't made them yourself.

One example would be a disease that someone is trying unsuccessfully to self-diagnose (so they search for the symptoms), or a set of symptoms over a long period of time whose combined significance the sufferer never recognizes.

Another example would be a person who has same-sex sexual attractions that they haven't consciously admitted to themselves, but that are evidenced in some of their searches or search patterns or clicks on search results. (Or just other online activity; there was research about figuring out if people are gay from their Facebook likes, and there must be lots and lots of correlates.)

Clearly Google hasn't productized most of these inferences in the sense of making them available to advertisers to use for targeting. There are many possible reasons for that. Some inferences would require a manual analysis of data; some are too speculative and have too high a probability of error; some are too creepy or alarming to the user.

This actually makes me want to start a separate thread: What does Google know about you that you don't? (What questions about you can Google answer that you can't readily answer yourself?)