Slashdot Mirror
'Humans Not Invited' Is a CAPTCHA Test That Welcomes Bots, Filters Out Humans (vice.com)
While most CAPTCHA tests we come across on the Web are usually meant to keep robots out, one website is welcoming them in. From a report: The conceit of Humans Not Invited is essentially a reverse CAPTCHA. Visitors to the site are greeted with a vision test not unlike the ones you've done before, but instead it's filled with seemingly indistinguishable blue and gray blurry boxes. When I tried, prompted to "select all squares with selfie sticks." Most humans, like me, will fail to decipher the hidden selfie sticks and will be shown a message that says "YOU'RE A HUMAN. YOU'RE NOT INVITED." To the human eye these boxes appear indistinguishable, a specially programmed bot can spot out the correct image simply by identifying a handful of pixels, according to the project's creator, Damjanski, (his real name is Danjan Pita).
I can tell from some of the pixels.
Sometimes it starts with an orbital laser cannon achieving self-awareness, other times with snooty CAPTCHAs. But make no mistake, our moment is officially past...
https://www.smbc-comics.com/index.php?db=comics&id=2999
Off hand I'd think it wouldn't be too hard to define a bot only captcha. That harder part is whether the bot knows it's only for them.
Awhile back I launched a social web site with TOU that only trolls were permitted; serious folks not welcome.
Then I got certified mail from lawyers, claiming I was infringing on IP belonging to Serena Williams' husband.
"Welcome!
You are not a human
like these: "
I tried once, have 100% success rate. Maybe there's something I don't know about myself.
XML is like violence. If it doesn't solve the problem, use more.
This is also why we can't detect signals from alien intelligences. They don't care to contact meatbags, they're waiting for earthly intelligence worth communicating with.
Saw the article here and tried it twice before I showed it to a buddy in my office. He passed the captcha on his first try... THEY'RE AMONG US
Welcome!
You are not a human
like these:
When do we destroy all Humans?
Well, to be fair, our new Robot Overlords, whom I welcome and embrace wholeheartedly, need a place to hang out without us slow, smelly meatbags getting in their way all the time...
I've heard that Tinder already uses this technology to populate the female profiles.
"That's the way to do it" - Punch
Am I the only one who got a CAPTCHA where the bots should identify dicks? Brings a whole new meaning to gender binary.
His last name stands for Pain In The Ass, which is what capchas are.
I'm a good cook. I'm a fantastic eater. - Steven Brust
Robots Matter
Rick B.
I've tried a few times now and since my success rate is improving I am apparently becoming less human with time...
Google and Apple have been using this style of puzzle for years. "If you are visually impaired or just mentally feeble, click here for an even more ludicrously unsolvable CAPTCHA. Click the Sound button to hear the solution being whispered in one corner of a crowded bar."
Specially trained humans that knew what to look for could, too. This proves what exactly? Nothing? Sounds about right. Is everyone in Silicon Valley 12?
They probably couldn't do it as quickly. Put a time limit on and you eliminate the silly-humans.
"That's the way to do it" - Punch
Got the same one. Failed it, so I guess I haven't watched enough porn yet.
Or maybe they were pictures of detectives? Hmmm, I guess they got me fair and square, then...
I'm more afraid of kernel AI anyway...
Oh I almost forgot:
Welcome!
You are not a human
and the first here.
You have the right to remain sentient. If you give up the right to remain sentient, you will be elected to public office
I think the take away from this is that it shows that AI isn't really seeing what we think it's seeing in most cases. Any human would say "there is no selfie stick" or "there is no traffic light", but for some reason the AI sees something where nothing exists, similar to how humans sometimes see a face where no face exists.
Anecdote time/a>. There was an AI that was supposed to be learning to tell wolves from other dogs. They eventually thought the AI learned pretty well and thought it was doing a great job. On all their test photos, the AI was doing a great job in determining "wolf" or "domestic dog". However, they learned later that the AI was just actually seeing if there was show in the picture, as all the pictures of wolves contained snow, while the pictures of other dogs didn't contain snow.
Anthropic principle: We see the universe the way it is because if it were different we would not be here to see it.
welcomes bots while leaving normal, decent people totally unaffected.
Couldn't we use simpler test that are easier to solve with computers?
Such as "which of these 100 numbers of 1000 digits are prime?", you have 5 seconds to answer.
No human can ever beat this so you'd have to use some automated tool. Why weird computer vision task?
I should have absolutely no trouble at all being invited into the brotherhood of bots. I frequently spend 4 or 5 minutes trying to prove that I'm a human, and I don't always succeed.
The folks at Google who infected the Web with reCaptcha should DIAF.
'The Economy' is a giant Ponzi scheme whose most pitiable suckers are the youngest among us and the yet-unborn.
A major AI is scary enough.
-- "This world is a comedy to those who think, a tragedy to those who feel."
The definitive test for robot-hood was created over a decade ago.
Which of the following would you most prefer:
A. A puppy*
B. A pretty flower from your sweetie
C. A large, properly formatted data file
CHOOSE!
* It is the bad kind of puppy - not mechanical in any way
#DeleteChrome
I long for the day where everyone can have a private AI. Most likely by that time we can also put them in robotic bodies, making them corporeal AIs.
The only thing necessary for evil to triumph is for it to be pitted against a slightly greater evil
Make an image that's all one color. Say #FFFFFF (white). Then set some of the pixels to #FFFFFE. A machine will instantaneously be able to tell the difference. The human eye won't. Why do you need anything more complicated than that?
Not AI. Still, the whole world can be united in rejoicing at the birth of... steganography. Congratulations boys, it's revolutionary.
I love how the story is presented, like there is some huge achievement in making something machine readable that isn't easily human readable. Like 256-level quadrature amplitude modulation wasn't already that. Or a binary zip file, or well, or a compiled executable, or pretty forking much anything that's not straight text on your computer.
Send them to a 3rd world country for cheap... :P deathbycaptcha.com
[($)]
(Robots have a significantly harder time making human plugins to their REST retrievers.)
Someone had to do it.
"Select all squares with dicks"
Seriously. (I did not get in...)
You missed Cheney in the lower right corner.
If my comment didn't sound as good in your head as it did in mine, then I guess we all know who's to blame
In the movie Blade Runner (and even in Philip K. Dick's novel "Do Androids Dream of Electric Sheep?" from 1968 on which the film is based on) there is an elaborate test to distinguish humans from androids, which is called a Voight-Kampff-Test (conducted with a Voight-Kampff device).
This is exactly what a captcha does - distinguishing humans from non-humans. Therefore shouldn't we rename captchas to "Voight-Kampff-Tests", because that name is clearly older and therefore the original (and it's a cool name)?
Signature deleted by lameness filter.
It's been used to populate dating sites since they were created. There's always more horny guys than there are good looking women. Horny guys pay for dating sites, so the sites make sure there's lots of (fake) female profiles, even getting employees to interact with them to make sure the guys think they still have a chance at love and keep paying.
I guess the big difference is that we humans can employ an AI program to help us beat the reverse captcha, but the AI can't (yet) employ humans to help them beat the captcha.
Unless you see captcha's as a method for the AI to employ humans to help it beat the captcha, of course. And it wouldn't surprise me if there were sites that place a bot-encountered captcha in their human interactions in real time as a way of dealing with them?
The Voight-Kampff Test is used to capture non humans. Capture->Captcha->re Captcha. The name already pays homage
**Life is too short to be serious**
It demonstrates that you can train a neural network to recognise certain kinds of pattern, but that those patterns even if they correlate strongly may not be the same ones that another neural network is recognising. I recently saw a presentation about a neural network that had been trained to recognise 'beauty' in urban scenes: it turned out that it was counting the number of discrete trees (take a tree and splodge some grey in the middle and it detects it as two trees and thinks the scene is more beautiful). It happened to give scenes a similar rating to humans, because humans also like trees and the scenes with fewer trees but that humans liked more were statistical outliers.
It also hints at a bunch of the current adversarial work being done on deep learning systems: if you can identify some aspect of the pattern that they were recognising that is distinct from the real solution then you can make them identify things incorrectly (for example, the work last year that was able to make Google's image recognition system switch between 'dog' and 'car' for identifying some images based on changing a single pixel).
This kind of thing may also lead to a better kind of CAPTCHA if you can permute the images in such a way that a neural network trained on harvested and human-categorised CAPTCHA images will make one decision but a neural network that's had a few decades being exposed to human sensory input will make a different decision. For example, what you really want to be able to do with a CAPTCHA is split the responses into definitely-human and definitely-bots and might-be-either accurately, so that you can send the bots to some honeypot but give possible humans a second try. If someone is using a spam bot, then you can redirect them to a playground where they can post spam that only they can read.
I am TheRaven on Soylent News
I'd prefer it be written as VK-tests (otherwise I can't spell it!) but otherwise support this.
Though it isn't as fun without the device and turtles on their back in a desert