I always click at least one wrong square on purpose, just to fuck with their training data. Most of the time it accepts it anyways.
Why?
To fuck with the training data. But don’t worry, most of the time it accepts it anyways.
That’s what he’s doing. Not why he’s doing it?
I’d assume it’s because they don’t want to hand out free labor.
You are on Lemmy asking why would anybody fuck with Google?
Lemmy is cool and mainstream now bucko. Have you seen the amount of shit posts on your front page lately? If not, congrats, you’ve curated your feed about as much as you probably already did on reddit too lol.
The lolz.
Or to slow the inevitable AI takeover of society.
Is it going to put a lot of… traffic light recognisers out of work?
Yes. If you count taxi drivers, lorry drivers, bus drivers, delivery drivers.
I think mostly it’ll be the first option though, and a happy feeling from “sticking it to the man”.
Are you for real?
Because fuck them, that’s why. Nobody, not a single person, is out there trying to make AI that will make the lives of average humans more livable. 100% of it is for corporate profit. Oh yeah, they’re gonna replace your job. And then where do you think your income will come from? It won’t. Come from anywhere. You’ll just be out of a job, no longer able to afford rent or food, but don’t worry because for a brief moment, they really maximized shareholder value. In the short term at least. And that’s all that matters anymore right?
I don’t want to: not only work directly toward that goal, but also do it for free. There isn’t much you or I could do to stop this shit short of revolting, but if enough people were to fuck with Captchas, it would absolutely have an impact on the rate at which this tech is ruining our lives.
I don’t want to do a job that a machine would do better than me.
And then where do you think your income will come from? It won’t. Come from anywhere. You’ll just be out of a job, no longer able to afford rent or food, but don’t worry because for a brief moment, they really maximized shareholder value.
Yeah, that’s my problem, and, on a broader scale, society’s problem. When enough people get hungry, things will have to change. The alternative is…doing pointless labor that a machine could do better, just to prop up a failed system. Forced inefficiency. How depressing is that?
Well if the two options I have are: continue being able to afford to live while doing a job that can now be done by a machine, or to just let it happen with zero safety net and let the good vibes from the naive optimism of “if enough people go hungry, then things will have to change.” I know what I’m choosing and I’m sure that will be a great comfort in that time
Those are not your two options. Society will have to reform from technological pressure. Political activism, such as lobbying for a universal basic income, is part of this.
Trying to fight progress just so you can keep your shitty job helps neither you nor society. That energy would be better spent on a future that’s better for everybody.
You’re not doing work for free. That’s a really weird way to look at it. You’re doing it to gain access to a service, an entirely optional one. It’s somewhat believable when people say oh, yes I have a car but I have to have a car, I’d like not to. But nobody is required to use GMail to live.
And yes, feel free to be an AI pessimist. You could quite well be right. It seems bizarre to me however to make an effort to sabotage this one specific AI training (which might stop cars from running us down) so that you can log into something which will also be used to train AI (to sell you shit). And then posting your opinion about it here, where it will be used to train an AI.
I feel like you think you just came up with something truly groundbreaking, “You criticize society yet you live in society. Curious.” You have to believe me when I say you haven’t.
… I don’t see how you can genuinely have understood that to be a point I was making. That’s the opposite of what I was saying. I even gave an example of how you might be forced to drive a car. Because you are in fact forced to live in society. But you are here on Lemmy so you must understand that you are not forced to use Gmail? So either you’re being disingenuous or you’re not actually reading all of what I’m posting. Given how you completely ignored the majority of my other post and went straight to “oH sO iF wE dO iT fOr 20 yEaRs”, I’m going with the latter.
The data collected from these tools are used to train models that detect cars and stuff with precise accuracy. Decades of data from millions of users each day. Once these are perfected, they will be sold to smart car users as auto-driving mode and what not. These services are likely going to be subscription based to maximize the profits.
Not to mention that they’ll probably be used by cops to profile people long before then…
Profile people as… traffic lights?
You’re being overly literal, but…maybe that too? 🤷
Yes?
Sorry I’m not sure “why?” is a difficult question. People keep explaining that is happening as if this hasn’t been happening in various forms for 20 years.
If its “I don’t want to contribute to machine learning”, I’ve got exciting news for you about every post you are making on this site where every comment is publicly accessible and consumable.
I just don’t think many of us believe that, “it’s been happening in various forms for 20 years” is an acceptable reason to continue doing or accepting something.
You act as if Captchas themselves weren’t created to prevent corporations from gaming the system. Like you do realize that the only reason that “every post you are making on this site blahblah” is because corporations just had to exploit the system for profit at the expense of literally everyone. Captchas don’t exist because a handful of clever coders made bots that were able to skim some profits from Google. It’s because corporations turned that into a billion dollar business. Capitalism ruins everything it wraps it’s putrid tendrils around.
So yeah, no. “You’ve been doing it already for 20 years” doesn’t cut it for me.
The 20 years thing was my bafflement as to why people felt the need to explain what Captchas are. Not a justification for their existence. When I ask “why” I think I’m pretty clearly not asking to have how Captchas work explained to me.
Putting that aside, can you explain to me how asking you to identify a traffic light is “at the expense of literally everyone”?
Someone was saying “they’ll train smart cars! And sell it to us as a subscription!” which I think is the same route you’re going down. But… how is that at the expense of everyone? And would not training smart cars benefit anyone? Is someone else going to do it? Is there a FOSS smart car I can contribue to? And is that FOSS going to prevent bad actors from training it to misbehave?
What? Why?
I mean if you think about it, you’re doing unpaid labor to prove you’re a human.
Is that a reason not to do it?
Obviously.
Might not be for you, but for them it seems to be
Because assuming you are human is not so much about the correctest of selections as much as it’s about moving the cursor around the screen in imprecise, non robotic, humanly ways.
These are actually self taught for what is right, based on what most people click, do just for shits n giggs he’s talking about peppering the data used for verification with incorrect responses.
Honestly, we should be paid for every CAPTCHA we complete. The amount of free labour big tech steals from us is astounding.
There’s some anti-CAPTCHA services where they do just that. You pay so you don’t have to fill them out. I’m guessing they have a bunch of Indian guys just filling out CAPTCHAs.
Theres ons that uses ai to do it
Pfft, AI is just thousands of Indians behind a curtain
Just ask Amazon
The Mechanical Turk thing?
Those would be Turks, not Indians.
But still not AI
Isn’t that Amazon’s Mechanical Turk?
Is that still a thing?
The payout for Mechanical Turk was abysmal. You’d make more money walking down the street and picking up the occasional quarter.
Yeah, it always seemed like flat-out exploitation to me.
Ever since image captchas were created, they’ve fail me 50% of the time, so I’ve spammed a lot of random shit that’s been accepted. If you get run over by an AI truck, I’m sorry.
You might be a bot.
I FUCKING HATE GOOGLE CAPTCHAS
Sometimes when I’m on VPN or TOR these little fucks won’t let me in no matter what, but instead of showing error message or something, they just keep throwing new traffic lights or bicycles at me
If you are using Firefox, switch your default search engine to DuckDuckGo, haven’t gotten a captcha from them once. And right in the search bar you can switch to Google if DDG isn’t doing too well on a particular search, I do end up doing that about 5% of the time.
I’ve been using VPN, Firefox and DDG as default search engine for years and I’m filling captchas every single day.
Huh, I’ve been using a VPN for under a year, and DDG for just a couple of months, but I haven’t had a captcha from them yet.
A lot of times those are the backup captchas, the normal one looks at your browser history instead. Using a VPN will change your IP and throw off the history logs, or a setting on your browser
Nah, it’s cause you’re on an ip blacklist. While lots of people use vpns just for privacy reasons, there is also a lot of abuse from them and the ips get blacklisted frequently.
There is something fundamentally wrong with having to prove I’m human. Especially to a machine.
On its surface, it’s a good idea. Website hosts don’t want to be inundated with fake traffic and fake data inserted by bots.
But it’s entirely unnecessary when your local library or doctor’s office is using captcha just for you to fill out a form, it’s a bit excessive because it’s highly unlikely any type of botnet would be targeting sign up fields on their sites. The attackers wouldn’t get anything out of it.
Eh, not that unlikely.
The bots just scan everything they can.For example, I’ve seen a guy do some testing with SSH server on default port (22). On average, there was constant 10Mbps of traffic just from the login attempts.
I can imagine it to be similar with websites.
I never know if I’m supposed to click the squares that have a tiny bit of the object or not. I hate the ones like this where you have to do 3-4 screens and then it fails you.
I like some of the newer captchas where it’s like “choose the heaviest animal” and it’s trippy pics of birds, dogs, and elephants.
It doesn’t matter because they show the images to multiple people and even shift the images around. If a square is only halfway there some people will click it, some won’t and this way you can generate some sort of heat map which is all you need to label your training data.
And then you’ll fail 50% of those people because fuck them, fill out more captchas you monkey
The secret is, it doesn’t matter.
For those who are confused by what we mean by “free training”, read this article:
Great article.
Data always goes somewhere…
Ever since I’ve understood that it accepts objectively wrong answers as long as it somehow seems as if you gave it some thought, I’ve made sure to hinder the accuracy of models that try to use my data.
Meanwhile I use a VPN, so half the time it rejects me no matter what. It’s really fucking annoying.
There is an extension that solves those:
https://addons.mozilla.org/en-US/firefox/addon/buster-captcha-solver/
I fucking hate that scenario. Click on buses: ok, that the corner by a few pixels. Does it expect that or will it make me do another captcha?
It expects you to behave like the other humans in its training sample. Just act like a neurotypical and you’ll pass the captcha. And if you can’t act like a neurotypical, then you’re fucked.
Just tell us if you people click the sliver box or not!
Us people? You mean autistic people? Most of us click the box and that’s why we fail the captcha.
Most of the time they already know whether you’re human or robot from your user agent string and the speed of your request. I encountered multiple layer of Captchas if I turned in my VPN and blocked all trackers.
I’ve found that using a VPN tends to get e caught in more Captcha loops than not. I think google, etc. has gotten better at knowing which swathes of IP addresses belong to VPNs. I thought that maybe it was specifically an issue with NordVPN because it had gotten so popular, so I switched to Mullvad and nope. Still get way more captchas. Still keep it on.
Especially notice this whenever I use Tor, sometimes even get stuck in captcha loops
Then why does it fail me every other time if it knows I’m human and I don’t use a VPN? Very annoying!
If it doesn’t like your answer it doesn’t say you’re wrong. It just gives you another puzzle to solve. Even if you’re right, sometimes it gives you multiple puzzles anyway. Don’t sweat the small stuff.
Something about the expression on his face in that scene is so fucking funny to me. That look is just perfect for the scene.
It’s mostly just making sure you move your mouse like a human.
The test and the answer as proposed is irrelevant.
Why does it keep asking me over and over again, then…and then other times, it doesn’t. Did I accidentally move my mouse like a robot? Several times in a row?
It’s more likely your IP was flagged for a reason unrelated to the captcha so it’s giving you more challenges to verify.
These always come up when you’re in a huge rush to do something
After 362879 wrong answers you will pass. Or after 2,0922789888×10¹³ tries if it’s a fancy 4x4 grid.
