Edit: sometimes there two pictures that are the same but the colour of the wall change. And you have to compare that to a black and white top down where you can’t see walls lmao
Edit: sometimes there two pictures that are the same but the colour of the wall change. And you have to compare that to a black and white top down where you can’t see walls lmao
These things are ableist. We are reaching the point where AI can solve these much more reliably than a human. As a result the difficulty has to rise and will exclude more and more people which might have problems with “basic” tasks from a neurotypical perspective. Not to speak sometimes there might be multiple solutions depending on language and cultural interpretations.
These captchas are also collecting data to train AI models
Yeah it’s almost like the whole thing was fucking stupid from the get-go.
“Prove you’re not a machine by training this machine to pass this exact test.”
And we know what the response will be, here. More and more of the internet gated by incredibly invasive verification methods.
If you’re not willing to let the OS/web browser fist fuck your computer’s most intimate areas, down to the hardware, and work it like a puppet just to make sure you’re not a bot, then you’ll just be hard blocked from every site.
Never mind how that will incidentally allow them to report that you’re using a VPN or an ad blocker to websites, or some other unapproved software, and even take action against all that while harvesting data…no, truly the biggest concern is the bots.
There is nothing stupid about this unless you believe that the people behind it had no plan to change out the challenges over time.
These have been used to train ai for decades. Of course they can solve them, it’s what they are being trained to do by them.
WCAG AAA actually mentions that. Which includes things like OTP. It’s going to be tricky to balance security with accessibility in situations like this for those with cognitive and physical limitations
https://www.w3.org/WAI/WCAG22/Understanding/accessible-authentication-enhanced
OTP is a success criteria instead of captchas but then you have issues with accessibility when you require somebody to have a device like a phone for the code or require them to app/context switch. So somebody using a device using their eyes or tongue as an input trigger would have a hard, if not impossible, time logging in with that arrangement.
The future wasn’t supposed to be like this.
Passkeys ftw
Eventually shouldn’t we theoretically reach a point where AIs can solve any possible practical to use captcha just as well as a human? I kinda wonder what the answer to replace them will eventually be
Let’s ask the AI to come up with something
“Prove you’re human by chopping off a finger.”
“Make yourself bleed on camera to prove you’re human.”
It’s already beatable right now, there are services in third world countries where people get paid fractions of a penny to solve captchas for machines.
Well, this isn’t a problem for smaller, less centralized services, so that might be an answer. Obviously not an answer big corporations will bring to the table, but ultimately, it might simply be among the reasons why users do still prefer smaller services.
Oh yeah, it’s ableist as fuck. The worst offender I’ve run into was a CAPTCHA I had to solve to make a neuropsychologist appointment, of all things. Sure, they only required the use of basic skills to solve, but those were the very skills I was seeking treatment for.
Gave up in tears, never got that appointment.
I feel your pain. I have rare disorders which affect my executive functioning, so simple tasks are insurmountable at times for me while my intellectual ability is unaffected and likely above average. Which leads to two things: I rarely get help making appointments and what to do and just “do it yourself” without guidance and get kicked out. And the other one is that I could not be possibly correct interpreting the test results, because how could I have the ability without being an official researcher to research things. I am deeply frustrated, the more tests I run and see my thesis is correct, the more push back by doctors it is more likely to have randomly a dozen conditions instead of a single one uniting them. Welcome to Ehlers Danlos.
I ran into this recently. Trying to get access to a credit union’s system as a vendor, they had a captcha that was the old style image of distorted text, with a text box labeled “are you a robot?”. Having the tendency to take things literally, I initially typed “no” into the box. That was not the right answer.
Don’t they normally offer other options for people who can’t do the visual challenges?
As if all people can understand English speech well.
Do audio captchas require English? I’ve never actually done one but I sorta assumed they were language agnostic, or at least would adapt their language to the system language.
On rockstar it’s listening for birds in music and stuff, but still it’s 20 of „pick which number song out of those 3 has bird sounds”
Yep, Blizzard’s is “pick which audio clip has only one person talking” and you literally have to do it 20 times.
It’s absolutely fucking awful.
Try them. Last one I tried was absolutely utterly impossible. Two times ago, it was easier or just as easy as the visual CAPTCHA. Unfortunately can’t recall either provider/site.
I’ve had to do a lot of Jira captchas over time. They were so horribly ambiguous that I had a failure rate of about one in two. So I tried the audio captcha and was met with the sound of a demon being murdered and nothing else.
Well at least you know there’s one less demon out there 👍
I stumbled on this recently
https://friendlycaptcha.com/
I can hardly claim to know enough about captchas to weigh up the cost / benefit, but I was delighted to come across a captcha that didn’t try to force me to train an AI