I had one of the old fashioned distorted text ones the other day, but instead of something like “please enter the text above” it just said “are you human?” next to the text box. Naturally, I typed “yes” but that turned out to be the wrong answer.
The correct answer is “or are you dancer?”
I still don’t know whether you’re supposed to hit those and I also don’t know if it’s normal to get two challenges or if that just means I did the first one wrong.
It doesn’t really matter, they don’t expect you to get everything right on these. While most of the time you need to get mostly right (Google is using these to train their AI so often they are not sure themselves), they are also looking at other things, like how you move your mouse, and the cookies that they use to spy on people to determine the probability of you being a human. If you pass a certain threshold they let you through, and you can do it even if you miss a square.
I’m convinced the person that put it together has never touched a faster-paced mouse+keyboard game in their life, because clicking all the correct squares too fast also causes the test to fail
That’s what a bot would say /s
But you’re right, the UX sucks, and there are other ways to detect and limit bots that don’t impact legitimate users as much - but Google needs to train their AI, and developers need to cargo cult stuff.
and the cookies that they use to spy on people to determine the probability of you being a human
which is why I assume, as a VPN user who rejects as many cookies as possible, I constantly have to do 5-6 fucking captchas in a row, sometimes more, before it’ll let me through… I can’t be that bad at doing them lol
Is it frustrating? Fuck yeah. Will it get me to change my behaviour and drop those measures so that the companies getting in my way can collect more of my data? Fuck no.
Have you tried using an automatic CAPTCHA solver (e.g. Buster)?
I will have to look into this as well
Isn’t it normal to get something like 6 challenges?
And suddenly one of them has new slow loading images which you won’t notice before clicking continue, thus failing
AFAIK, the first one is the real check, the second one is too train their image recognition AI.
It has to be more sophisticated than that. Otherwise users could easily taint the datasets by giving wrong answers on purpose.
It probably checks your answer against the current model’s best guess and if it’s close enough, you get a pass and your input is added to the training data for the next iteration. The more wrong you are, the more challenges you get.
This is how I approach these: that square only has a single traffic light, not multiple traffic lights like the prompt is asking for
Try the audio captcha, those seem to have actual valid answers to them.
Funny enough, there’s an extension that solves captchas by feeding that audio through a speech recognition algorithm. If anything it’s more reliable than solving them manually
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.
If google want me to do work for them, they can pay me.
They kinda do. This is the way the “free” model of internet services works. One of the reasons I think we should probably switch to expecting services to either be paid or non-profit, rather than ad/data-supported.
Yeah, but the whole point of offering free services was just a ploy to crush competition with shorter runways to profit. Google could just sustain "free"services longer than their competitors could remain solvent.
Now that they’ve run most of their competitors into the ground, and now that people and businesses have become dependent on these services. They can bank off advertising and monetizing services with subscriptions.
Google business accounts used to be free, now you have to pay 9 bucks a month per employee, and you are subjected to even more advertising. Neither advertising nor subscriptions are going anywhere, especially now that subscription plans are so normalized.
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.
The ones that get me are captchas saying select all squares with motorcycles when it is clearly a bicycle.
There was an inflection point where captcha went from “demonstrate human vision” to “guess what the robot sees.”
I got one asking for mountain ranges where one was plainly the tops of nearby trees. Which I got scolded for not clicking on.
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”
