It is incredibly obvious that CAPTCHAs are at the very least a way of exploiting distributed labor to train AI.
They had been used to help with text recognition for book scanning for more than a decade. It has never been secret, it was explained on them time ago.
This is the logical progression, regardless of your feelings with “AI”
That was their selling point.
deleted by creator
Okay, this “$1 trillion” metric is a bit of a reach, and seems to be based on an arbitrary value assigned to an estimated amount of data Google has collected, and not actually $1,000,000,000,000 in revenue. It does not appear that Google has actually made a trillion dollars from CAPTCHA data.
A reminder that recaptcha is no longer free, but since a few months ago now the website owner needs to pay $0.001 each time a verification is performed
https://cloud.google.com/recaptcha/docs/compare-tiers
Free tier is only 10k verifications per month and must link a valid credit card so they can charge you immediately when you reach higher level.
Hopefully this kills the product in the long term as bots solve recaptcha faster than humans, so it’s just for slowing down humans than actual security. I personally use a browser extension that solves them with a click in a second.
Do you pay for successful verification only, or even for failed ones?
Lol so now site admins pay more than the bot farm companies pay to solve each one
This sounds like a conspiracy theory but I’d like to know more.
It’s true. They make us work to identify data, we are checking for them not confirming, then they also track us.
Since Cloudflare published Turnstile I’ve hated Captchas even more, because Turnstile does it so much better. Captchas are such a hassle. One website I occasionally visit does not keep me logged in and then presents one of the worst captcha puzzle systems. Shitty captchas are a huge barrier.
Turnstile is, in almost all cases, one checkbox to click (I’ve never been challenged beyond that). All captcha puzzles should be replaced with Turnstile or similar simple (for the user to solve) tech.
Cloudflare turnstile is also the only captcha system that works ok with most browsers and adblockers.
Especially Google recaptcha freaks out if you use Firefox or an adblocker or anything and asks you the hardest possible questions.
That’s a feature, not a bug.
Anticompetitive feature!
Better than not asking you any questions and just going in an infinite loop!
If your getting looped by turnstile maybe you need to stop being a bot?
Literally never had an issue with it.
Its a false positive. That’s the point.
Turnstile is worse. Its just an infinite loop. Literally clicking for hours
It’s a lot easier to determine the intent of this hed with the quote being closed somewhere. Just after “service” would have been my guess, but it’s a disservice to remove that and leave people dangling.
My larger issue is that when I’m faced with traffic lights – or, god forbid, motorcycles – this is performative nonsense wherein I’m supposed to guess percentage coverage on a given square without having been provided parameters.
At this point, CAPTCHAs feel designed to make sure you can never get through the first time, thus needing to continue training image models several times before I can just fucking do what I originally came to the site for.
I already hate them for access gating based on unnecessary labour, and deliberately making access more cumbersome for people not using chrome and using VPNs
But what really peeves me off, even though it’s much less important, is that they don’t localise them.
Where are the crosswalks? What the hell is a crosswalk. How many trolleys in this picture? None, that’s a picture of a tram!
“I see no trucks, only lorries.” Being on a VPN has been getting worse and worse with CAPTCHAs, almost like I’m being punished for telling my ISP they have no right to sell the details of my internet use since I’m paying them.
At this point, CAPTCHAs feel designed […] training image models
It was never a secret:
The reCAPTCHA program originated with Guatemalan computer scientist Luis von Ahn, and was aided by a MacArthur Fellowship. An early CAPTCHA developer, he realized “he had unwittingly created a system that was frittering away, in ten-second increments, millions of hours of a most precious resource: human brain cycles”
I was fine with it when it was wavy text to digitise old works. This shit is just asinine and a time sink.
Yeah… only OCR and AI have advanced to the point where a spammer/bot can easily bypass them.
20+ years ago, Microsoft proposed a [Penny Black project](https://en.m.wikipedia.org/wiki/Penny_Black_(research_project)), which was superseded by reCAPTCHA. Nowadays, we might have to go back to that… maybe by mining crypto as a proof of effort.
Proof of work. See mCaptcha and Friendly Captcha.
Does this apply to hCaptcha?
