Well, some of the proposals flip the relationship on ads. For example, rather than getting access to fine-grained personal data about readers, a marketer gets some coarse data like general location (eg. midwest US) and language. Using that, they propose a set of ads to the browser, and the browser decides which might be relevant based on what it knows about you. The browser keeps the personal data to itself, though. Marketers get some analytics on the response to their ads, but not full-on tracking.