Hacker News Reader

Comment by aeternus | Hacker News Reader

I'm not sure that images should be your largest concern.

Messaging apps and the sites you visit online are much richer sources of data for adtech and ML.

I agree that messaging apps (including gmail) are used to enrich the targeting cloud for a given user, but many use a secure-ish app by default (iMessage). Also, if you use an adblocker (like ublock origin), those should mostly muzzle pixel beacons.

The metadata you can pull from a series of images is remarkable. Did you know that most smartphones from the last 7 years will encode not just GPS coordinates, but the compass heading of the lens? Just from that stream, you know (most likely) where the person lives, works, and spends time. Throw in face detection and you know who else lives at your home, and who your coworkers are. It's a metadata treasure chest.

Plus, with as high resolution as cameras are now, you could even identify some of the products and brands in their house, estimate taste and income based on interior finishes and visible products. Is that birthday cake sitting on a granite countertop? Expensive furniture or cheap? KitchenAid mixer in the background? Oh look, a book shelf in the background, lookup all those titles. Bill O'Reilly book, must be conservative. Rachel Maddow books, liberal.... There really is a ton that can be gleaned from just random photos.

Advance image processing AI by 1000x (which could happen in 10 years) and the images are much more dangerous.

I think we're there already. A generic getting-started-with-AI online course will do a reasonable enough job at object detection and identification to be dangerous.

Facebook already knows who your friends are and what you've been up to even if you have never used the site, because it's seen you in the photos they uploaded.

Reply on news.ycombinator.com