Some sort of principal component analysis to determine your "cohort".
An advertiser isn't that interested that you are you. They care that you are, for example, 55 with an interest in gardening and have been looking at lawnmower review sites.
Edit: Actually, come to think about it, I wonder how people would have felt in the 80's if they got one of several editions of a newspaper based on magazines that they also subscribed to. Households that subscribed to fitness magazines would get a newspaper with more gym membership and weightlifting gear ads, and households that received woodworking magazines would have more Craftsman ads in their newspaper. Would people appreciate the customization of the paper, or would they see it as an invasion of privacy?
The WSJ seems to be describing https://github.com/WICG/floc, but the kind of remarketing you're describing is what https://github.com/WICG/turtledove is intended to support. Advertisers would still be able to run that kind of personalized ad, but the browser API would not allow them to learn your browsing history in the process.
(Disclosure: I work on ads and Google, speaking only for myself)
The bit you're quoting sounds like the WSJ trying to describe https://github.com/WICG/floc. That depends on the browser choosing to support the API, yes.
(Disclosure: I work on ads at Google, speaking only for myself)
For example, if they aren't collecting information on you than how do they have enough information to create cohorts? Is it that they are deeming information processed on your local computer running their software something that isn't them collecting? This seems like a form of misdirection.
Without knowing the technical details (which one never knows with Google), it leave me the impression that they have moved some of their categorization software client side (Chrome, web workers, etc) and are saying that they aren't collecting the data in that case.
Is it Google using remote devices as edge devices to do a bunch of work they'd been doing server side?
[1] https://www.cookiebot.com/en/google-third-party-cookies/
Edit: I did not have time to read it, but here is the paper and probably something like the algorithm they want to use: https://github.com/google/ads-privacy/blob/master/proposals/...
I'm having a hard time parsing this out, and seeing what's actually changed. How do they determine an individual's "cohort(s)", without collecting information about that individual across multiple websites?
Is it simply that the data is collected and processed client-side, rather than server-side? Would using a non-Chrome browser effectively opt-out altogether, then? I find this difficult to believe.