- cross-posted to:
- [email protected]
- cross-posted to:
- [email protected]
Anyone who has been surfing the web for a while is probably used to clicking through a CAPTCHA grid of street images, identifying everyday objects to prove that they’re a human and not an automated bot. Now, though, new research claims that locally run bots using specially trained image-recognition models can match human-level performance in this style of CAPTCHA, achieving a 100 percent success rate despite being decidedly not human.
ETH Zurich PhD student Andreas Plesner and his colleagues’ new research, available as a pre-print paper, focuses on Google’s ReCAPTCHA v2, which challenges users to identify which street images in a grid contain items like bicycles, crosswalks, mountains, stairs, or traffic lights. Google began phasing that system out years ago in favor of an “invisible” reCAPTCHA v3 that analyzes user interactions rather than offering an explicit challenge.
Despite this, the older reCAPTCHA v2 is still used by millions of websites. And even sites that use the updated reCAPTCHA v3 will sometimes use reCAPTCHA v2 as a fallback when the updated system gives a user a low “human” confidence rating.
You must log in or # to comment.
This is actually a good sign for self driving. Google was using this data as a training set for Waymo. If AI is accurately identifying vehicles and traffic markings, it should be able to process interactions with them easier.
As I understand it, the point of those captchas was never really “bots can’t identify these things” (though you’re right on that it was used to train). They use cursor movement, clicks, and other behaviours while you’re solving it to detect if you are a bot or not.
The image choosing was always just to train their own bots
Since I started getting good at yosu and that fishing mini game in farmrpg I’ve been failing more captchas. I wonder if they’re related knowing this
It’s a combination.
Most captchas goals generally aren’t 100% prevention, it’s to put a workload in front, this makes spamming the site cost money, a bankrolled attempt could just as easily outsource the captchas to real humans.
The annoying thing is that they held us hostage for our free labor, but the results are proprietary for Google’s benefit only.
That training data ought to be forced to be made freely available to the public, since we’re the ones who actually created it.
Afaik this is precisely what the captcha data was intended for - training AI models. Originally leveraged machine learning. LLMs are a slightly different paradigm but same purpose and results here.
i hope you’re joking. please, tell me you’re joking?
Removed by mod
Its never been confirmed by Google, so I may be wrong. It still tracks that the data harvesting company with a AI self driving car project would use free human labor to identify road hazards.
I was referring to the “This is actually a good sign for self driving” part of their comment.
The captcha circumvention arms race has been going on for over two decades, and every new type of captcha has and will continue to be broken as soon as it’s widely deployed enough that someone is motivated to spend the time to.
So, the notion that an academic paper about breaking the current generation of traffic-related captchas (something which the captcha solving industry has been doing for years with a pretty high success rate already) is “good news” for the autonomous vehicle industry (who has also been able to identify such objects well enough to continue existing and getting more regulatory approval for years now) is…
Not really. I’m not even sure what you’re disagreeing with based on the above comment.
My point is that if bog standard AI can accurately identify all of the road information from pictures, that is good news for self driving.
What was once a nearly impossible task for computers is now mundane, and can be used to improve safety/utility for self driving, especially for FOSS projects like comma.ai
So can we stop using those damn things? They’re super annoying!
I’m kind of hoping the AI permanently beats them. I hate them too.
Well yeah, I’d hope so, that’s the entire point.
Catcha’s data collection always was with the intent for training ai on these skills. That’s “the point” of them.
It’s reasonable to expect that the older version of captchas can now be beaten by modern ai, because they’re often literally trained on that exact data to beat it.
Captcha effectively is free to use on websites as a tool because the data collection is the “payment”, they then license that data out to people like OpenAI to train with for stuff like image recognition.
It’s why ai is progressing so fast, captchas are one of humanity’s long term collected data silos that are very full now.
We are going to have to keep progressing the complexity of catches as it will be the only way to catch modern AIs, and in turn it will collect more data to improve it.
Yeah, my understanding is that these capchas were made to harvest data to use for AI/Autopilot driven cars. That’s why they are always having you identify motorcycles, bycicles, crosswalks, stoplights, busses, etc. It’s all stuff that automatic driving cars have had a hard time identifying.
Wait, so if a visitor fails the v3 Captcha, v2 is used as a fallback?
That makes absolutely no sense.V3 isn’t necessarily more effective than V2, it’s just less obtrusive.
Not quite: it’ll drop a v2 captcha for you to solve when a v3 one can’t clearly classify you one way or another.
So if v3 isn’t entirely sure you’re human, it’ll make you do a v2.
And if you fail the V2, it’ll just take your word on it and let you pass anyway.
deleted by creator
god I fucking wish
When it’s asking for motorcycles but it’s clearly a scooter
Or, like, “there’s the bottom 10% of a traffic light in this one. Do I click that box? Ia that supposed to count?”
Does the backside of a traffic light even count? What about these strange traffic lights that have more boarder than light?
How about “do they want just the bulbs or the pole holding it up?”
What they are doing is comparing your answer and seeing if it is consistent with how it has been answered previously. They realize that not everyone is going to give the exact same answer, so as long as you answer it in a way that enough other people have answered it, it should let you in.
I’ll usually go with the minimum number of clicks that I think will get me through, since I’m lazy and it’ll also at times slow down how fast you can click which is annoying.
I’ll also answer them wrong if I think it’s a mistake that enough other people will make. “Yes… that RV over there is a bus…”
They are also overly US centric.
One of the questions asks you to click on only the school buses. I had to Google how you tell the difference between a school bus and not a school bus.
Also is it a crosswalk if it’s at an intersection or is it only a crosswalk if it’s in the middle of a road somewhere?
The questions either need to be not cultural or they need to be adapted for where they detect the user is coming from, the first option seems easier.
Interesting. Do you not have school buses, or are school buses not distinctly marked? How do kids get to school when it’s beyond walking distance?
They are just buses.
I guess the British government just assume that school children are smart enough to get on the right bus without them being individually distinct.
I knew school buses are yellow but I did not realize that they are always yellow. I did not realize that the yellow color meant school. I just assumed that the yellow color was a color busses could be.
The size of the UK verses the exponentially larger size of the US probably has a lot to do with it.
And if you knew school busses where yellow… Where’s the problem?
I had one with one of those Motorcycles with the long handles, apparently they aren’t part of the bike, but the dudes foot holding it up is.
That tip of a handle bar that makes you wonder if that square counts or not.
And yet I can’t beat the CAPTCHAs because reCAPTCHA doesn’t like VPNs lol
I was going to say I’ve straight up just left whatever website I was trying to access because I was stuck in some endless loop of clicking on street crossings, buses, bikes, and street lights.
But, I cannot pass those 50% of the time… what does that mean?
Ever watched Bladerunner?
It means Rick Deckard won’t be coming for you any time soon.
You’re walking in the desert and you see a tortoise
Is it an Agassiz or a Morafka?
You should schedule an eye appointment.
Everywhere I look, all I see is fire hydrant, buses, and bicycles
My score is lower.
I mean, we literally train them by completing the CAPTCHAs. Why do you think you were picking things like bikes, traffic lights, cars, and busses? The only question now is what’s next…
they embed dark souls into the browser
In order to pay your utility bill, you have to beat the Undertale Sans fight in Genocide mode
Thank God this means i can stop wondering if i should click on the… the 13 pixels from the fucking bike in that one corner square or wondering if i should count the scooter as a motorcycle fuck i am so tired of that shit
Complete the obligatory “is this a staircase or street crossing” round only to be roundhouse kicked back to the beginning.
I never get the first one and rarely the second one. If it says to click all the squares with motorcycles and it’s just the one big picture, am I supposed to click stuff like the tire and mirrors? I always do and never get it right. Then most of the time they ask me to identify motorcycles, they show me motor scooters and what am I supposed to do then? I think I just need to get one of these bots to do it for me.
Fwiw they aren’t really asking about the motorcycle. I mean they are but they are washing your mouse movements and how fast you click through the images. It’s okay to get a few images wrong.
Not quite.
It’s mostly wisdom of the crowd, as it always has been.
As long as you mostly click the same squares most other people click, you pass.
You often at random get 2-3 images because 2 of them are actual checks, but the third is a new image that you auto pass and they’re using it to gather data on what the average clicks are on it.
A motor scooter is a motorcycle in the eyes of the law.
I bet you use the word “actually” in conversions regularly.
Well we trained them
Great, so now can I get an add-on to my browser that skips these?
As someone who can not decide if 3 pixels of a motorcycle counts as a correct square, I need this add on.
In use an add-on that does 90% of these for me already on Firefox. I would tell you what it’s called but I’m not at my PC.
Which (on a side note) I’d totally go downstairs and check for you, but I just sprained my ankle real bad, and am dreading stairs. Sorry :(
Sorry to hear about your ankle. When you’re able to, I’d also like to know what the add-on is
CAPTCHA doesn’t stop bots, and let us be honest, it never really did. It frustrated the hell out of people though, and caused people to waste time doing these challenges. Meanwhile even before AI bad actors and bots could get past it simply by using captcha solver services run by exploited humans solving captchas for the service.
It’s a display of security theater meant to make normies feel safe but in reality doesn’t stop most bad actors.
I fail more of those checks then these AI bots do. Surreal.
