Does warmer mean temperature? Color? Something else?
You must log in or register to comment.
“Please select all images which evoke melancholy as opposed to existential dread.”
“choose the roads that fill you with ennui moreso than the avenues that give you a sense of listless”
That’s like, your opinion man
Please try again
Okay I’ll click the green ones
Sorry, but we cannot verify that you are human, please try again later.
Hey man,open up. It’s me, Dave. I got the stuff.
I love how so many of these images meant to fool bots are literally generated by AI
Because their real purpose is to train the bots. The captcha features are just there to get us to use it.
It’s more complicated than that. Only 2 or 3 images are meant to train AI, the rest are meant to establish whether you’re human enough that AI can learn from you.
That’s also only partially true. Most of the human detection is done by collecting metadata about your browser and how you interact with the widgets.
I think the actual website detection is usually done that way, while the AI training is generally separate and runs it’s own standalone game.
It’s all a ranking system. Who knows the score - my goal is to get through in as dishonest a way as possible. If they want my honest answers they should pay me for them.
You have to win the game however. Training the AI is an essential part of it.
Nah I disagree, it’s possible to win the game, gain access to the website, but train AI the wrong way. That’s the sweet spot I aim for.
It’s also possible to win the game and gain access, but the website decides you’re an idiot and disregards your training.
AI: “Um, …these aren’t the droids you want. Squirrel!”
I’d be with you if it weren’t so obvious.
If you’re wondering why does it seem so strange, it’s because the learning model is actually hyper sophisticated now. It knows what a bus, a bicycle, and a sailboat looks like, now it’s asking for comparative assessments of complex images. It clearly understands that snow is covering houses and that snow is cold.
OK, but that doesn’t explain why anybody thinks that this is good UX.
It isn’t supposed to be a good user experience, it’s supposed to train their AI models, and they figured out how to get you to do it for free.
And a perfect opportunity to poison ML models.
You wish lol. In truth they can serve this to as many users as they want, and catch garbage outlier responses.
Only if those responses are outliers. If enough people do it, the data gets corrupted.
The most surprising thing is how you and some commenters dont see how obvious and dead simple the answer is
Like, should they show you a block of ice and a fire next time?
[This comment has been deleted by an automated system]
Am neurodivergent, didn’t even occur to me they’d be talking about snow vs indoors. I thought because it is a visual test they meant color temp which for me registered as the middle left, center bottom, and maybe an argument for bottom right.
I thought the same thing. Was looking for the overall color temperature of the scene, but none really fit.
[This comment has been deleted by an automated system]
Thanks for reminding me of my neurodivergence, I couldn’t figure it out either.
I was starting to think “ok this one looks lit and cozy, and how am I supposed to tell how well insulated these houses are?”
It even clearly gives an example of a picture of a living room.
Something that’s obvious to you isn’t necessarily obvious to everyone.
More surprising is how apparently some of you haven’t encountered captchas that employ nuance, and what seems like the obvious answer sometimes isn’t.
The thing is that a captcha is made to be solvable by almost anyone.
So whatever you think the answer is, is probably one of the many correct responses.
my man, its a blizzard and indoors, what part of that has any more nuance than being beaten over the head with the answer
The different snow images have different color tones, some matching that of the example image. The center image has a cool color tone, which doesn’t match. Captchas are made to defeat AI logic, so sometimes it’s not the obvious thing. It could very well possibly be selecting all images that match the color tone, something a bot may not work out. It could be just selecting indoor images. I wouldn’t know for certain until I got one of these and succeeded or failed. Personally I think it would be too easy for a bot to just ignore all images that have snow, or are mostly white, because that doesn’t resemble the example image at all.
edit: and in case it needs to be said, getting beaten over the head by anything doesn’t involve nuance. That’s the opposite of nuance.
Captchas aren’t made to “defeat AI logic”, the human detection happens in part outside the picture selection part. The picture selection is for training AI. In this case you are training an AI to distinguish the (potentially abstract) concept of warmth.
Semantics, whatever. In truth it’s both, if you stop long enough to actually think about it instead of parroting other replies that solely focus on AI training.
I couldn’t get past “pick the smallest animal”
There was a large picture of a hummingbird, and a tiny panda. Both choices were wrong, apparently. They probably meant that I should pick the pettiest animal.
Captchas are made to defeat AI logic, so sometimes it’s not the obvious thing. It could very well possibly be selecting all images that match the color tone, something a bot may not work out.
IMO the idea here is that most users are not thinking very hard, so they are going to see the word “warmer”, think “snow = cold” and leave their analysis at that. AI on the other hand is going to put more effort into interpreting the specific meaning of the request in context of the images. The primary challenge for captchas now is to defeat AI, so the captcha ideas that get through probably did so because they gave the AI trouble in testing, but did not give most users trouble.
I think that going forward, people who put thought into following specific directions accurately are going to have a lot of trouble with captchas.
I can agree with most of this. Still don’t know for certain which interpretation would be the correct one for this captcha. Thank you for acknowledging that defeating AI is one of the goals, which seems obvious since they’re meant to determine if you’re human. idk why that’s difficult for some people to accept.
There’s a sample picture of a living room, then pictures of living rooms and snowy houses.
Yes, thank you for your entirely original response, and demonstrating again that many people lack depth in their thinking. Honestly it’s sad that you replied to that explanation with this.
I’m also starting to believe that some of you don’t understand what “sometimes the obvious looking answer isn’t the correct one” means.
The most surprising thing is how you and some commenters dont see how obvious and dead simple the answer is
Like, should they show you a block of ice and a fire next time?
This is an incredibly narrow view of people, and what ‘obvious’ is. This sentence is absolutely awful if you’re ESL in any way:
Please select all images of one type that appear warmer in comparison to other images
Even I stumbled for a second on that sentence. What the hell does ‘appear warmer’ mean? Colour, hue, saturation, is there a temperature reading on them? It can snow at zero degrees, but that middle image could be -20 for all we know; it’s in shadow and the only non-cool-colour in it is that orange rectangle.
I mean, to me, it’s obvious that you add an apostrophe to ‘don’t’ but you didn’t. Your sentence also doesn’t end with a period. Does that mean I get to call you out for missing such an ‘obvious’ thing, and insult you for not doing it? You know, how obvious and dead simple writing your sentence correctly would be.
Great, now they expect you to be thinking about lighting temperature terms. People who don’t do photography or haven’t read light bulb boxes won’t know wtf this means.
They could easily mean ambient temperature.
The example shows an interior room which would indeed be warmer. There are two which could be what they want you to select.
I think it means warmer than the thumbnail photo it shows at the top. It could mean temperature, it could even be trying to understand our sense of coziness, who knows what the ai is fishing for at any given moment
Can’t say I’ve had a captcha with only one result as the answer.
I had a captcha that had you click to place a single dot at the end of a line. They get weird sometimes.
Well yeah the word ones are single answer, but pick the pictures? Always more than one.
Don’t overthink it. Just select the interior ones.
Are you sure about that? Because I’m not. I would have to actually get this captcha and succeed or fail.
All the other ones are literally in the snow.
Yet different snow images have warm or cool color tones, some matching that of the example image, and the center shot appears to be a cooler color tone. You could be correct, but you could also be wrong. Again, it seems like some of you haven’t encountered captchas that employ nuance, where sometimes the obvious looking answer isn’t the correct one.
I assume this just means “pick inside” without saying it directly. The sample photo is of an inside space. No? The two in the middle row, I assume, are the “correct” answer.
Often the correct answer is only half the puzzle - how you answer (mouse movement) also can be to determine things
No, there’s far more depth to that. The goal isn’t for you to prove yourself human, the goal is to teach an AI how to “think more human”.
1, 2 and 7 are obviously cold. They’re oustide, with no “warm” colour lighting.
3 and 6 are both green houses, the green house could be considered “warm”, but 3 has light on the inside. This is perhaps a test against AI readers. To a human, they both seem warm inside, but an AI might differentiate based on the lighting.
9 is a dark brown house, but 8 is a light brown house that is illuminated by external lighting. This contrasts with 3 and 6, because 6 has external lighting but it does not illuminate much.
4 and 5 are both internal shots. 4 is light and airy, meanwhile 5 is a bit more grey - but then, grey is the fashion these days.
All in all this is a bullshit test made up by bullshit people looking to get a bullshit result, with which they hope to make money off of.
You’re working to help them make more money, meanwhile they don’t pay you for your labor. They also collect data from your connection to their servers - as well as the website you’re trying to access, you will almost certainly be connecting to at least 2 other servers to deliver this hcaptcha, and thanks to cooperation with the website host hcaptcha will triangulate the internet routing and fingerprinting information to attain a significantly accurate identification of you as an indvidual (which they will then consolidate with whatever other information they have).
Much like a disgruntled worker might “phone it in”, or work within the requirements of their paid employment, or “quiet quit”; you should limit and perhaps even poison the output you give in proportion to what you’re being paid for your labor.
The goal isn’t to satisfy captcha, the goal is to get passed it while giving as little commercial value as they compensate you for.
Your data has value. If it didn’t, then Facebook and Google wouldn’t be amongst the wealthiest businesses in the world. You own the value they establish themselves with, they just claim a license.
Way to throw a shit fit over paying 5 seconds of labor to use a service
The service is free to use. My access to it is not conditional to my authentic participation in unpaid labor, and nor is it equivalent.
Personal data has value. Thought has value. Commercial enterprises like this attempt to suppress that value, while simultaneously using it to position themselves amongst the wealthiest businesses in the world. They should pay us for our data.
What they do is akin to a car manufacturer saying they shouldn’t pay the person who makes nuts and bolts, because nuts and bolts have far less value than a car, and the people who make nuts and bolts do not know how to build a car. This is would be a ridiculous scenario; it is also ridiculous that users aren’t paid fairly for their data.
If people were paid fairly for their data, then these businesses would have no scope to raise the price of their product in line with this new (fair) material cost. This is because the cost of their product is already an exaggeration of the value they provide. They sell their product for more than it’s worth, meanwhile they pay their data suppliers (every single human being) nothing. Of course they don’t want you to realise the value they’re taking, doing so could only reduce their profits.
The service is free to use
No it isn’t.
My access to it is not conditional to my authentic participation in unpaid labor
Yes it is, and you’re being paid by access.
Is paying in labor instead of money such an alien concept to you?
That isn’t the deal as described in the contract. That is how they try to frame the deal after the fact, to convince people to let them get away with it.
The site is free to access. While you access it, they claim rights to your data, or in this case the output of your work. It is not an exchange of access for data/labor, it is a free provision with terms snuck in via the fine print.
It’s really not that complicated. If you don’t want to pay to use a service, that’s your perogative, but it’s not a deceptive trade at all.
The website gets to avoid bots spamming forms, you get access to the forms, and Captcha gets some training data. Everyone benefits
It absolutely is deceptive. The captcha does not openly tell the user “if you complete this you’re going to train an AI system for us which we will eventually sell for profit”. The user is merely told to “prove that you’re human”. The terms and conditions or privacy policy also don’t spell things out in plain English, it’s all generalised statements meant to disguise what they’re doing.
It’s also not true that everyone benefits. The user is supposed to gain access to the website for free - the website wants users to visit. However, the website wants to prevent non-user bots from accessing the website. Instead of the website paying for a service to prevent bots and taking that as part of their overhead costs, the website is getting the user to provide unpaid labor to pay a third party for the service that the website wants. The service gets a benefit, the website doesn’t have to pay, the user has to do all the work with no fair reward.
If it was literally just proving the user was human, that would be different. These systems extract further value from the user, for which the user is not compensated.
It might only be a small thing, a few pennies here and there, but they’re stealing pennies from everyone.
There’s an example at the top along with the instructions.
This ain’t exactly the Mississippi.
That’s exactly what the model is trying to learn.
I worry about commenters in this post that seem to take this as some sort of highly complex problem verging on philosophical rather than a silly little riddle to go through as fast as possible to get to the primary part of website.
It’s the “of one type” that gets me - to me that says I should be examining either the outdoor or the indoor pictures, not comparing between those two types of picture. So I should somehow pick the warmest outdoor or warmest indoor pictures.
It clearly says to select images of one type, not examine images of one type.
I would write it “Select all the images of the type which is warmer than the other type of image”
I think it’s just asking you to pick the indoor pictures because they don’t have snow in them. The confusing wording is to trick AI trying to get through captchas.
Whole lot of AIs in this thread.
Go home, robot, you’re drunk
Numbering left to right, top to bottom, I think the answer should be 3, 5, 6, 9.
Fuck you for trying to get me to train your AI. If you want my work, fucking pay me.
Edit: To be clear, I think those answers would be most likely to almost seem correct to an alrgorithm, but actually break their objective for training.
You think 3,6 and 9 are warmer than 4?
I explicitly left out 4.
3, 4, 6, 8 surely.
I’m serious, do not call my Shirley.
3, 4, 6, 8 have the kind of lighting that a person might be drawn towards. The kind of thing they’re trying to train AI towards. My answers are meant to seem like the kind that AI would accept as a human answer, while also being wrong to the human eye.
It’s 4D chess. You have to predict what the AI thinks you would think, and agree with that, while providing an objectionable answer to the things AI is uncertain about.
If they want the right answer, they should be paying us for it. They’re a business, labor shouldn’t be free for them.
I always click randomly on the first 2 attempts to mislead the AI. Hopefully you did the same, and when the robots come to kill us the grease will freeze up and they won’t be able to move.
