Comment

Comment on It turns out that Juggalo makeup blocks facial recognition technology

The clothing they wear solves most of those. For the physical side (gait, height, etc) it’s a little harder but shoe’s have an effect on most of those (i.e. the round bottom shoes meant to help people work out just by walking wildly change a normal gait and posture).

For the devices though it could get fun. You could have a device mounted in the helmet that will pretend to be people you’ve passed, essentially just replaying the beacons (SSID broadcasts, etc) for the sake of a digital camouflage.

source

Sort:hotnew top

AHemlocksLie@lemmy.zip ⁨1⁩ ⁨day⁩ ago

The clothing they wear solves most of those

For now. This cat and mouse game will continue on and on. We’ll develop evasion techniques, they’ll learn how to recognize and see through them. We’ll develop new ones again, they’ll learn them again. What about if you speak near a camera? It’ll learn to analyze voice and diction. Voice scrambler? AI is learning to descramble video, can probably learn it for voice, too. Your clothing style will become a data point and an expensive one to consistently falsify. The locations you’re seen at is suggestive. If you walk a dog, good fucking luck convincing it to help you falsify data for the AI monitors.

Still, this is predicated on the assumption that you can recognize and falsify enough of the data points. My point is that they will collect however many data points it takes to make it nigh impossible to get a failure to identify you or a false positive. And if it’s a false positive, we have to question the ethics of pinning your trail on some other random dude.

source
- ITGuyLevi@programming.dev ⁨1⁩ ⁨day⁩ ago
  I agree for the most part, but if we are all walking around in jumpsuits and helmets (Daft Punk style) and repeating the digital beacons of everyone else it seems like false positives are a skill issue for AI. Not too long ago I was watching a video about a guy that was a 100% match in the eyes of AI as someone that was trespassed by the casino. When the cops showed up and he presented his documents, the cops brought him to the station as they thought he must have given false ID when he was originally trespassed. He was eventually able to prove his innocence but the fact he was taken into custody because AI messed up makes me have no issue with people doing stuff to intentionally poison the data.
  
  None of this matters in the present context though because just by wearing that you would be easily identifiable.
  
  source
  - AHemlocksLie@lemmy.zip ⁨1⁩ ⁨day⁩ ago
    
    repeating the digital beacons of everyone else
    
    You can’t perfectly mimic everyone else and accomplish something unique at the same time, even if it’s something as simple as pulling up a webpage nobody else around is requesting. Your device must in some way identify itself to the network so it can actually receive everything they request, and that’s an avenue for identification and tracking.
    
    Not too long ago I was watching a video about a guy that was a 100% match in the eyes of AI as someone that was trespassed by the casino. When the cops showed up and he presented his documents, the cops brought him to the station as they thought he must have given false ID when he was originally trespassed.
    
    Sure, modern AI can’t push the limits like I’m talking about, but I’m not talking about doing all this with modern AI as it is now, and things are advancing extremely rapidly. Processing power available is, too, as companies churn out as many new data centers as they can. It might not be as long as we hope before the things I suggest become feasible.
    
    He was eventually able to prove his innocence but the fact he was taken into custody because AI messed up makes me have no issue with people doing stuff to intentionally poison the data.
    
    Yeah, modern AI is trained unethically at just about every step of the process, so poison away.
    
    source