@davidaugust Eventually, people will be paid to fill these things out based on their ability to prove they are actually human, which not every person will be able to do.

@davidaugust the treachery of capcha images

@davidaugust

This is the precise theme of one of the most famous Zen koans. ("The Moon in a Bucket".)

@davidaugust When you think about it, though, captchas never show you actual things. It's always just a picture on a screen. We automatically understand that "select all squares with traffic lights" means to select all squares containing pictures of traffic lights. And there *is* a picture of a traffic light there, on the picture of the sign.

@davidaugust There Are No Lights and I Must Scream

@davidaugust
did u need to select the entire sign, or just the painted picture?

@davidaugust my new game with these is to see how much i can get wrong and still get through

@davidaugust "Ceci n'est pas les feu de signalisation."

@davidaugust I believe these captchas are created by humans

@davidaugust with luck you are teaching their driverless cars to idle in front of this sign forever.

@davidaugust THERE! ARE! FOUR! LIGHTS!

@davidaugust On the plus side, when the Robot Uprising begins we now know we can stop their autonomous vehicles in their tracks by simply painting pretend traffic lights everywhere. The robots will patiently sit there for eternity waiting for the green to get brighter.

@davidaugust I can't find it now, but I once saw a Treachery of Images parody with an Airpod.

@davidaugust @stavvers

@davidaugust https://www.youtube.com/watch?v=nbRyH6fkee0

@davidaugust
I love when they use blurry photos so a bot can't decipher then but then I gotta guess is there supposed to a car in the far background there or not.

@davidaugust

@davidaugust I'm late with this reference but I'm sharing it anyway because I didn't see this version in the replies :P

AI, AI, AI! 😂 🤦🏻‍♂️
@davidaugust

@davidaugust Ce n'est pas un feu de circulation

@davidaugust These pictures are used for machine learning.

It amuses me this is helping to train the machines that pictures of traffic lights are traffic lights.

Somewhere some machine learning engineer is going "oooohhhh yeah pictures of the thing are not the thing... right..."

@davidaugust @notwhatwethink hahaha, wow.

@davidaugust @burgerdrome there are four lights

@davidaugust @beat Ceci n’est pas un feu tricolore.

@davidaugust Since their answer is essentially used to train machine learning somewhere else, I'd say it's perfectly reasonable to go along with what it "wants" in order to mess with them.

@davidaugust This thread gave me an idea: put a fake captcha with a countdown timer, and when it finishes, display a video of a smoldering car with the caption "you weren't fast enough."

@davidaugust both hilarious and depressing at the same time...

@davidaugust you can participate in this too https://adversarial-designs.shop/products/this-shirt-is-a-stop-sign

@davidaugust 🤣🤣🤣🤣

@davidaugust
This is not a pipe (it’s an image of a pipe).

@davidaugust
Captchas: "Prove you're human by clicking on the traffic lights and motorcycles, because robots are incapable of identifying traffic lights and motorcycles."

Tech billionaires: "Look, we've invented self-driving cars and – trust us here – they're PERFECTLY SAFE!"

@davidaugust My car used to provoke me with statements like, “The door is a jar”. Sadly, it was unresponsive to my counter-arguments.

@davidaugust lolololololol

@davidaugust I'm sure I once saw one which asked for bicycles, and the picture was a bicycle painted on the road - which looked nothing like a real bicycle would.

I do wonder whether I imagined this as I didn't take a screenshot.

@davidaugust THERE ARE! NO! LIGHTS!

@davidaugust 🙃

@davidaugust ceci n’est pas me fucking screaming

@davidaugust

Plato's Cave: Remastered

@davidaugust
So we know these captchas are training data for self driving cars but is that data itself being selected by a person?

@davidaugust @gus_shepherd@furry.engineer...did it want all squares with the sign in it, or only the squares with the iconography?

I'm not even sure which of those would be more ontologically correct...