Conversation:
Notices
-
I didn't mind #Recaptcha when it helped improve OCR accuracy, but it unsettles me when #Recaptcha is training UAV auto-navigation ("Please select all the images of trees") or facial recognition ("Please select all the images of women")
- Joshua Judson Rosen repeated this.
-
What's with #Recaptcha's obsession about street signs?
-
"Please select all images that contain a CAPTCHA"
-
"Prove you're not a robot: 1) Injure a human being, or through inaction, allow a human being to come to harm. 2) Disobey orders given by a human being. 3) Jeopardize your own existence." #CaptchasIdLikeToSee
-
@bobjonkman I saw one recently that was "Select all images with cars". I'm thinking they're feeding that into the self-driving vehicles. What if we could organise some way to game the system and teach Google's recaptcha that cars look like store fronts or something.
-
@bobjonkman, it's not so much the existence of a heuristic or training-datasets that should disturb us, but every specific abuse of those things; we need to be careful to avoid becoming so numb to rampant abuses of technology and knowledge that we start opposing technology or knowledge itself merely because it's easier. Not enough to have identified things to fight—must keep checking whether they're the right things.
-
Recognizing that technology could be used for bad things is work enough, never mind having to evaluate every instance of technology to determine if that instance is actually being used for bad things. Much easier to just use the proxy of "Technology is bad".
-
@bobjonkman, consider for example that #OSM is also a dataset for training #UAV auto-navigation.
-
@rozzin *quietly adds no-fly-zones to random places around the world*