Answer wrong. The more of us humans that answer wrong, the less accurate we need to be to get past these stupid things. If google want me to do work for them, they can pay me.
The worst for me is the motorcycles one; half of the pictures are of motor scooters. Does it count those as motorcycles or is it counting on the user to know the difference because they’re not technically the same thing?
I was about to make a similar comment but you beat me to it.
It recently showed me a bicycle as part of “select all motorcycles” so I didn’t pick it. And I failed. Twice. Finally picked the bicycle and it let me through. Guess the computer knows best.
It also does it the other way around. I failed recently for not calling a motorcycle a bicycle.
Beatmeattoit
Whenever I get a capcha of anyone on a vehicle, I always make it a point to highlight the entirety of the driver too because I’m not going to just let Google train its self-driving vehicles to just ignore that every motorcycle has a rider on it.
This scenario pisses me off as I debate how pedantic to be.
Its worst when you clearly see small traffic lights far back in the same pucture
Hate this. Every time, I seem to guess wrong.
Guess I’m a robot.
I stress about the whole damn pole. If you showed me a picture of a traffic 🚦 on pole, and asked me what it was, I would say “a traffic light” not “a traffic loght and a traffic light pole”
It’s funny that captchas are in a never ending arms race with bots trained on the same datasets capturing humans answering these stupid puzzles.
Pretty soon we’re going to be drinking verification cans
Well of course we are. A bot can’t drink!
I started half-assing these a long time ago because they never fail anyway.
The test isn’t the panels you click on, it tracks how your mouse pointer moves. Bots or robots tend to move in straight lines, whereas with humans, the pointer moves in a more random fashion. That’s how you pass.
!??
This is what “AI training” looks like, folks. The companies developing AI constantly tells us how awesome it is, but it still needs the help of humans to recognize basic sh*t like cars, buses, crosswalks and traffic lights. They didn’t choose those images by accident.
Those fire hydrants will haunt me in nightmares
Especially because the answers to these captchas are used to train self driving car AI, so if you fuck it up the self driving cars will crash and you will have blood on your hands :/
Yeah sometimes I’m like, well that’s a moped, does it count as a motorcycle?
My big interest right now is motorcycles yet I still recently failed a motorcycle captcha
Most of these were trained/tested using third world labor on mechanical turk. It actually helps to approach the problem from that understanding:
“Would a person who speaks English as a second language and who is getting paid less than a penny to decide, call this a motorcycle?”
Dude needs to answer about a million of those to make money, so he’s not overthinking it, your knee-jerk answer is probably right.
I find that whatever your snap judgment is will be the correct answer, so basically yes the moped usually counts as a motorcycle
This was me in the early days of captchas. Now I’m all like 360 no-scope BOOM HEADSHOT let me in motherfuckers!
deleted by creator
I just think to myself: What would a robot do?
They do that on purpose, as an AI/Bot wouldn’t hesitate.
Sometimes I’ll be stuck on a captcha because I’m answering too fast, so I’ll wait a delay then hit the answer and suddenly it stops going in circles
I thought the reason is they want to see the limits of what a human considers qualify as answer to their question in order to better train their AI?
Edit: Although I guess nothing is stopping the answer from being Porque No los Dos?
I mean it could absolutely be both. There’s assumedly a lot that goes into these algorithms and most of it we will never know for sure, just make educated speculation lol