alt Hacker NewsThe state of the art has advanced so far in doing this. I remember way back in 2017, 9 years ago now, at the Scaled ML conference, Claudia Perlich gave a presentation about using RTB data to target ads. When she got to slide 23 [1] my jaw hit the floor. This was a small ad targeting company, and again, 9 years ago. Here's what they publicly said they had:
Consumer Events:
• 100B DailyEvents
• 20+ data integrations
• Clickstream
• App usage
• Ecommerce sales
• Cash register sales
• Precise Location
Context Data:
• User
• Device
• Location
• URL
• IP
• 200 Million Devices Daily
Universal DataStore
• 50 Trillion Record Consumer History
That's about 150,000 datapoints on everyone in the U.S. For a small company. In 2017.
[1] https://cdn2.hubspot.net/hubfs/6212008/ScaledML%20Media%20Ar...
Replies
See also Dark Data from DEFCON 25: https://youtube.com/watch?v=yWqdTVQsnPg
But it doesn't need to be marketed in such a sinister fashion. In 2012 when Google Maps informed me of delays along my usual commute, complete with a GPS trace of my route home, completely unprompted, I started turning off location history (lol, yeah right). I didn't even know they were collecting that data, much less analysing it that hard.
These days, that would be considered a feature - not a dystopian hellhole, and you would be a Luddite for turning off this new smartphone augmented brain. The product will make you happy. [0]
Welcome, to City 17. You have chosen, or been chosen, to relocate to one of our finest remaining urban centers. It's safer here.
The personalization stuff is why I avoided ML like the plague - all these folks making huge money but all of it to build a surveillance state for advertisers. Already having value my privacy enough to never work for an ad revenue company, they all seemed beyond the pale.