Some of them were for training AI. Nobody thinks about the ways this takes place at this level. You might remember having to go thru 2 layers of these for something at one point. One was the actual robot trap, the other was training AI on something it didn't understand elsewhere. Maps used it for example to get house numbers interpreted.
I work in cybersecurity and privacy, and it’s honestly a pretty mixed bag. On the one hand, you’re absolutely right - this is so much worse. On the other hand, you wouldn’t be able to use the internet without this today - there is too much fraud and cybercrime, and it’s largely because of complex tools like this that actually keeps any of your information actually secure, while also keeping the services that you actively use or even pay to use available for you to be able to use. As an example, one of the ways my organization protects your (colloquial “your” - I have no idea if you use our services or not, and I don’t want to know either) personal accounts, information, and the monetary value our users store in their accounts is by using things like Google’s reCAPTCHA to help prevent hackers using botnets to break into them because users (or roughly 30-35% of them) are notoriously awful at using secure passwords. A few years ago, it used to be super easy to spot hackers trying to get into people’s accounts and steal their information or their money from them. It was very obvious in the logs, and ordinary log monitoring and alerting was sufficient to identify attacks and shut them down easily. Today, it’s not so simple, and the bad guys have improved at a terrifying pace in the last 3-4 years - in part because cybercrime has become a multi-billion dollar a year organized industry today. The good guys just don’t have the resources to keep up without advanced tools like these that rely on intensive and kind of invasive data analysis techniques like reCAPTCHA 3 (“I am not a robot”) and many others that are similarly complex and involved but require enormous deep data analysis capabilities. Here’s an example of what we’ve been able to do in the last few years, and how much harder it is to keep you guys safe online today. Four years ago, we would be able to identify a malicious attack against our users’ accounts within an hour of them starting - and those attacks were big, fast, and loud. They’d typically involve 2-3 times the volume of activity that’s normal for a 24 hour period all occurring within one hour - and not always even a busy hour. Today, with the more advanced analysis we’re doing of the data we have available, we’ve been able to shift that left to the point where we can now identify a “low and slow” attack in which an attacker may only be attempting to get into 4-5 accounts across an entire 24 hour period, with a single malicious access attempt, in which each account’s single access attempt happens from a different IP address. A few years ago, an attack this organized and sophisticated would’ve been invisible to us, and we never would’ve known it was happening, and our users who were compromised would only find out when their money was gone or their information was used to do something they didn’t have anything to do with. Today, we’re able to identify these tiny attacks with over 95% accuracy, and are able to prevent any monetary losses even for our users who have terrible password hygiene and do nothing to secure their own accounts. We still can’t prevent those users’ information from being stolen, but that’s because they basically park a running Ferrari in the middle of the highest auto theft neighborhood in the middle of the night with the engine running and the door open while they go watch a movie at the theater. But we CAN usually protect their monetary value in their accounts by spotting the malicious activity almost the instant it occurs and get their account locked down before any harm can be done to it. None of that would be possible without doing that deep data analysis and keeping track of the information that allows us to tell the difference between a real user and a hacker trying to steal from a real user.
This made me think of Bruce Banner (the Hulk) on the helicarrier in the first Avengers movie, "Really? They want me in a submerged, pressurized metal container? Oh no, this much worse"
And its even worse - it is and always has been possible to design a bot which imitates a human, and successfully can pass the Turing test (it's a cheat if you do it by design, but it works - most typical downside being time, which means costs) : and those kind of bots are better in it than humans as well ...
As a rule, new tech just tends to make things worse cause the incentive is to create profit for shareholders, not improve the world. There's no money in simply improving the world and money is what we structured our entire societies around. They just try to do it in ways that people will just accept as how things are. Give corporations permission to fully surveil you? Sure, that's better than any slight inconvenience, right? It doesn't have to be like this, but few things do.
This explains why Google has gotten very annoying at using check boxes when you browse in private mode. And why you have to solve CAPTCHA-like "check all the buses/crosswalks/cars" tests if you use your phone, as they don't have mouse movement to track.
No, they don't deem it as a bad thing. They just can't tell if you are a real human or a bot without additional information, so they gave you more tests to gather those additional information to confirm that you are a human. Ask yourself what is worse: having Google to give a few more tests to prove yourself, or not having Google to give additional tests and risking bots successfully pretending to be human which may have a huge impact on our future
Yeah, I mean, at this point it isnt even about bots and they know it - They just want to impede people from acessing things without giving their souls away to Google
Very well done, easily understandable and engaging. And you managed to enlighten some people about big techs data hoarding habits, which can't be repeated often enough.
What Lou failed to mention is that CAPTCHAs used to have a very practical side-effect. They would help digitise old texts. The public could look at a piece of scanned text that bots were unable to digitise, and would give the correct letters. Or, eight out of ten people would, and the computer would reject the two wrong answers and consider that piece of text digitised!
I think at least with the images from traffic, it was made so that some of the images shown had preexisting information on which squares are correct, and then there was another pictures shown with which you would train an AI to for example look for zebra crossings.
The worst part is when the box does not work and neither do the pictures which are such low quality that I have no idea what I am looking for. Clearly Google punishing me for restricting my privacy settings.
Google makes their money tracking and selling all your info. You didn't really think google would offer you all these free services out of the goodness of their heart, did you?
The captcha's where you have to identify pictures or part of pictures are also used for services like google maps. You always have to select items that generally exist in traffic and after the captcha proves that you are in fact a human you get one or two more tests. Those are tests where google's automated systems for example are not sure of wether it is a traffic light or not and once they know it is it gets updated to google maps. (obviously a simplified explanation but you get the point)
It’s not impossible to devise an algorithm to randomise mouse cursor movements as it moves towards a target, so I think checking the browser history is pretty much mandatory
I wish captchas didn't use objects that only exist in America like fire hydrants, "crosswalks", and American buses. I don't know what a f*cking American bus or "crosswalk" looks like. What even is a "crosswalk"? In my country it's when you need to go outside for a bit to calm down because you're angry.
Explains why i never get past them. Sites that employ them are inaccessible. NoScript means no mouse tracking. The picture puzzles work, but they never end and just keep adding new "challenges". Nowadays when I see such a captcha I close the tab.
You forgot something! You posted this in 2023, so you know nowadays more than half of Web traffic is mobile phones, and guess what? Mobiles have no mouse movements, only a tap directly on the checkbox! So there's only the timing and the coordinates where the hit lands on the checkbox to make the pre-determination. Is this why we are more and more often asked to select images?
My thought exactly. As I watched this I kept thinking… what about tablets and phones, which is almost exclusively what I use and where I do these tests? Obviously there must be more to it.
The main function is the Google looking at browsing historyand other criteria human operated browsers do while browsing, not so much the click or mouse travel.
Interesting point, but I suspect the touch screen sends and tracks more data than just a single functional tap. After all, it's already calibrated to discriminate between something like your fingertip and the heel of your palm, right?
so are you honestly telling me, noone can design a capcha click button that does movements at random to simulate human idiocy? is that what we doing here?
I absolutely hate the "traffic light test" because I can never understand its logic as well. The one using crosswalk is also an absolute nightmare because I failed the majority of the time. (I click on every individual square that contains even the tiniest amount of white stripe visible to human eyes and it is still wrong!) Whenever I clicked on the checkbox and the test pops up, I always like, "it is going to be an endurance then". As a VPN user, that can be annoying, as you might be forced to do that before your search result pops out.
So now the bots just have to do some random searching (and some not so random) plus add some fractal and/or random mouse moves and timing and away they go again.
Robots can check that box, but if they have to check that box often then it results in an image recognition test I've automated different bots that checked this box, but usually after running it for 10 minutes it isnt enough anymore to just check the box.
This really represents the advancement in computing over the last couple of decades. To summarise: In the early days it confirmed you were human by seeing if you were smarter than a bot. Now it confirms you're human by seeing if you're dumber than a bot. 😂
I hate those stupid image tests; I always get some object that has literally like one pixel on one square, and I am never sure if I am supposed to click that square or not.
The new reCAPTCHA tests are a really good way to determine how well your browser protects your privacy. A properly private browser will fail reCAPTCHA, because it doesn't give the test enough information to determine whether you are human.
I don't have any vision impairments, but those text and number puzzles are such a hassle. The same goes for clicking the images. I have to do 10 of them before I'm accepted. As for the box checking, you can easily macro human mouse movements and click the box. The browser history I'm sure can be spoofed. Am curious how it does against private browsers with tracking blockers.
It is so easy now to reprogram a robot to employ the randomize timer option to insert deviations from that straight line and to reduce the effectiveness of recognition. Thanks for giving the bot makers a clear path to now thwart captcha. Brilliant.
The check box mouse-movement test is easy to get around as well: just add noise to the mouse movement, while still landing on the check box - which is super easy to do. The site test.. why would one need to automate sites anyhow - but the solution is probably to queue random solution searches. Those curvy letters will be the death of me as a human person though.
After you spent an hour or more to locate where they hide those privacy pages to click on those options that said not to track or share, and then few weeks later, they sent an update to upgrade to a newer version. So you have to go through the same process again and again whenever there’s an update, which until one day, you would say, take whatever you want.
The mouse movement thing is a red herring thrown in by google, they don't even mention that by finding 'sidewalks' and 'traffic lights' we're actually training self-driving cars
While fairly technically accurate, it is missing a lot of the details that help answer some questions in the comments. It may use google account history as a flag of sorts, but from what I can tell as someone who has implemented it on their site, it primarily tracks you through the site you're on. The tracking script must be on the site for an extended period of time for them to be able to collect enough data to statistically define what is bot versus humans traffic on the site.
Why does the voice audio sounds like some 64kbps audio that was compressed to infinity and beyond was sent to Andromeda and back and then got reuploaded 5 times, then someone played it on a speaker and recorded it with his phone and then it got reuploaded 10 more times just for extra measure?
that "select all squares with..." stuff is so annoying. sometimes there is a tiny bit of edge of the object in a square and i just dont know whether to select it or not. need to try multiple times
Well explained. Regarding the box, I thought it was something like this, which is why I often find myself delaying / taking a circuitous route so it doesn't seem suspiciously robotic. I still get the picture stuff though. Interesting to hear about the history tracking, and the reason you don't see wobbly text anymore. I believe that was a dual-purpose feature to help digitise books that image processing had failed to do well, but as it explains, that's now so much better than humans so it makes sense that it's been scrapped!
It's not that complicated to add some randomness to the cursor's movement to simulate a shaky or indecisive human, and AI image recognition is getting good enough to render the "click the boxes with" pointless too. As for browsing history, just load up a nice fake history for each new session. None of this is something the average user will do, but it's trivial for the professional botters to do. So, once again, the average user has to suffer while the actual people they want to stop laugh and watch their bots pass the tests anyways.
![Lp. Сердце Вселенной #60 РОЖДЕНИЕ ЛОЛОЛОШКИ [Финал] • Майнкрафт](http://i.ytimg.com/vi/YoR0pAV9FVQ/mqdefault.jpg)
Having to pass 1000 captchas for only 30 cents sounds like an actual level of hell.
Ah, so this explains why I have to complete so many image recognition tests in succession. I’m actually a robot- thanks Google!
Same here. Do them slooooowly. Robots are fast with images. Humes are slow & clumsy
Some of them were for training AI. Nobody thinks about the ways this takes place at this level.
You might remember having to go thru 2 layers of these for something at one point. One was the actual robot trap, the other was training AI on something it didn't understand elsewhere. Maps used it for example to get house numbers interpreted.
Happens to me from time to time. Maybe I should move my mouse differently. I think I movey mouse in a straight line
It also helps train googles image recognition AI. Free labor yay!
“You passed by exhibiting incompetence at every turn” I love it!
First half of the video: "Yay they solved an accessibility problem and made the world better". Second half: "Oh... Oh no, this is so much worse."
I work in cybersecurity and privacy, and it’s honestly a pretty mixed bag. On the one hand, you’re absolutely right - this is so much worse. On the other hand, you wouldn’t be able to use the internet without this today - there is too much fraud and cybercrime, and it’s largely because of complex tools like this that actually keeps any of your information actually secure, while also keeping the services that you actively use or even pay to use available for you to be able to use.
As an example, one of the ways my organization protects your (colloquial “your” - I have no idea if you use our services or not, and I don’t want to know either) personal accounts, information, and the monetary value our users store in their accounts is by using things like Google’s reCAPTCHA to help prevent hackers using botnets to break into them because users (or roughly 30-35% of them) are notoriously awful at using secure passwords. A few years ago, it used to be super easy to spot hackers trying to get into people’s accounts and steal their information or their money from them. It was very obvious in the logs, and ordinary log monitoring and alerting was sufficient to identify attacks and shut them down easily.
Today, it’s not so simple, and the bad guys have improved at a terrifying pace in the last 3-4 years - in part because cybercrime has become a multi-billion dollar a year organized industry today. The good guys just don’t have the resources to keep up without advanced tools like these that rely on intensive and kind of invasive data analysis techniques like reCAPTCHA 3 (“I am not a robot”) and many others that are similarly complex and involved but require enormous deep data analysis capabilities. Here’s an example of what we’ve been able to do in the last few years, and how much harder it is to keep you guys safe online today. Four years ago, we would be able to identify a malicious attack against our users’ accounts within an hour of them starting - and those attacks were big, fast, and loud. They’d typically involve 2-3 times the volume of activity that’s normal for a 24 hour period all occurring within one hour - and not always even a busy hour.
Today, with the more advanced analysis we’re doing of the data we have available, we’ve been able to shift that left to the point where we can now identify a “low and slow” attack in which an attacker may only be attempting to get into 4-5 accounts across an entire 24 hour period, with a single malicious access attempt, in which each account’s single access attempt happens from a different IP address. A few years ago, an attack this organized and sophisticated would’ve been invisible to us, and we never would’ve known it was happening, and our users who were compromised would only find out when their money was gone or their information was used to do something they didn’t have anything to do with. Today, we’re able to identify these tiny attacks with over 95% accuracy, and are able to prevent any monetary losses even for our users who have terrible password hygiene and do nothing to secure their own accounts. We still can’t prevent those users’ information from being stolen, but that’s because they basically park a running Ferrari in the middle of the highest auto theft neighborhood in the middle of the night with the engine running and the door open while they go watch a movie at the theater. But we CAN usually protect their monetary value in their accounts by spotting the malicious activity almost the instant it occurs and get their account locked down before any harm can be done to it.
None of that would be possible without doing that deep data analysis and keeping track of the information that allows us to tell the difference between a real user and a hacker trying to steal from a real user.
Exactly. I feel a little sick, actually...
This made me think of Bruce Banner (the Hulk) on the helicarrier in the first Avengers movie, "Really? They want me in a submerged, pressurized metal container? Oh no, this much worse"
And its even worse - it is and always has been possible to design a bot which imitates a human, and successfully can pass the Turing test (it's a cheat if you do it by design, but it works - most typical downside being time, which means costs) : and those kind of bots are better in it than humans as well ...
As a rule, new tech just tends to make things worse cause the incentive is to create profit for shareholders, not improve the world. There's no money in simply improving the world and money is what we structured our entire societies around. They just try to do it in ways that people will just accept as how things are. Give corporations permission to fully surveil you? Sure, that's better than any slight inconvenience, right? It doesn't have to be like this, but few things do.
A friend of mine did the opposite for his website, he made a box he humans don't see, so only the bots click the box.
This explains why Google has gotten very annoying at using check boxes when you browse in private mode. And why you have to solve CAPTCHA-like "check all the buses/crosswalks/cars" tests if you use your phone, as they don't have mouse movement to track.
My favourite thing to do with recaptcha is not click the box. Anywhere near the box counts as an attempt, so you still get through.
That's never worked for me
So that's why it fails so often for me. I use tracking protection. Google deems tracking protection a bad thing it seems.
You say you use tracking protection but can we believe you? Are you sure you are not a bot? LOL
Google makes money tracking and selling your info. If you block that, then google blocks you.
Well of course, ads and web trackers is how Google makes money
No, they don't deem it as a bad thing. They just can't tell if you are a real human or a bot without additional information, so they gave you more tests to gather those additional information to confirm that you are a human. Ask yourself what is worse: having Google to give a few more tests to prove yourself, or not having Google to give additional tests and risking bots successfully pretending to be human which may have a huge impact on our future
Yeah, I mean, at this point it isnt even about bots and they know it - They just want to impede people from acessing things without giving their souls away to Google
Well...that went from kind of neat to horribly creepy _really_ fast.
Very well done, easily understandable and engaging. And you managed to enlighten some people about big techs data hoarding habits, which can't be repeated often enough.
What Lou failed to mention is that CAPTCHAs used to have a very practical side-effect. They would help digitise old texts. The public could look at a piece of scanned text that bots were unable to digitise, and would give the correct letters. Or, eight out of ten people would, and the computer would reject the two wrong answers and consider that piece of text digitised!
now they're used to help train the AI that steals artists' hard work instead
I think at least with the images from traffic, it was made so that some of the images shown had preexisting information on which squares are correct, and then there was another pictures shown with which you would train an AI to for example look for zebra crossings.
I miss the days when all you had to do was click the button. Now it's 10 rounds of "select all images taken on Tuesday"
"assumes permission" the most important phrase in this whole video.
The worst part is when the box does not work and neither do the pictures which are such low quality that I have no idea what I am looking for. Clearly Google punishing me for restricting my privacy settings.
Are you sure you are not a bot? LOL
Now that makes more sense since i have stoped google from everywere
Google makes their money tracking and selling all your info. You didn't really think google would offer you all these free services out of the goodness of their heart, did you?
This happens to me too, since I switch off cookies almost everywhere. It's especially bad in incognito mode. Guess this explains it!
I have always said that someday the captcha will be inverted such that only robots will be able to pass, thereby locking humans out of their world...
The captcha's where you have to identify pictures or part of pictures are also used for services like google maps. You always have to select items that generally exist in traffic and after the captcha proves that you are in fact a human you get one or two more tests. Those are tests where google's automated systems for example are not sure of wether it is a traffic light or not and once they know it is it gets updated to google maps. (obviously a simplified explanation but you get the point)
reCaptha is a perfect example of Google Being Google
You passed the captcha by showing incompetence at every step. This line summarises, humanity, like forever.
You forgot to mention that all versions of captcha are using the data collected to train robots to complete the same tests.
0:32 "Is That Even A ?!#@ing letter?!"
Very educational and fun to watch, good job
PMSL !
Very funny "Is that even a F$#KING letter?"
I fail that test all the time.
🤣🤣🤣
Great video, I didn't expect an explanation of captchas to be so entertaining
It’s not impossible to devise an algorithm to randomise mouse cursor movements as it moves towards a target, so I think checking the browser history is pretty much mandatory
I wish captchas didn't use objects that only exist in America like fire hydrants, "crosswalks", and American buses. I don't know what a f*cking American bus or "crosswalk" looks like. What even is a "crosswalk"? In my country it's when you need to go outside for a bit to calm down because you're angry.
Because they’re just so honest they can’t lie.
Explains why i never get past them. Sites that employ them are inaccessible. NoScript means no mouse tracking. The picture puzzles work, but they never end and just keep adding new "challenges". Nowadays when I see such a captcha I close the tab.
You did not allow Google to peer into your soul...and this is its revenge.
Yes as many webpages the captcha is broken and you can sit for hours in endless loop of captchas...
You forgot something! You posted this in 2023, so you know nowadays more than half of Web traffic is mobile phones, and guess what? Mobiles have no mouse movements, only a tap directly on the checkbox! So there's only the timing and the coordinates where the hit lands on the checkbox to make the pre-determination. Is this why we are more and more often asked to select images?
I don't allow my robots to have mobile phones...
My thought exactly. As I watched this I kept thinking… what about tablets and phones, which is almost exclusively what I use and where I do these tests? Obviously there must be more to it.
The main function is the Google looking at browsing historyand other criteria human operated browsers do while browsing, not so much the click or mouse travel.
Interesting point, but I suspect the touch screen sends and tracks more data than just a single functional tap. After all, it's already calibrated to discriminate between something like your fingertip and the heel of your palm, right?
I dont see why a bot couldnt be programmed to take a route that mimicks a humans.
so are you honestly telling me, noone can design a capcha click button that does movements at random to simulate human idiocy? is that what we doing here?
I absolutely hate the "traffic light test" because I can never understand its logic as well. The one using crosswalk is also an absolute nightmare because I failed the majority of the time. (I click on every individual square that contains even the tiniest amount of white stripe visible to human eyes and it is still wrong!) Whenever I clicked on the checkbox and the test pops up, I always like, "it is going to be an endurance then". As a VPN user, that can be annoying, as you might be forced to do that before your search result pops out.
Her voice is perfect... too perfect...
So now the bots just have to do some random searching (and some not so random) plus add some fractal and/or random mouse moves and timing and away they go again.
Robots can check that box, but if they have to check that box often then it results in an image recognition test
I've automated different bots that checked this box, but usually after running it for 10 minutes it isnt enough anymore to just check the box.
I thing the bigger crime is people getting only $1 for doing it 1000 times. WTF!
This was actually interesting. Good work!
Wow, your video was the most entertaining way to learn new things. Very informative. Thank you.
Loved: the educational and comical nature
Disloved: being reminded I’m being watched 24/7
Excellent video.
Interesting, entertaining and only as long as it needed to be. 👍
good explanation and great humor that aren't over the top!
This was both informative and funny, which is the best combo.
This really represents the advancement in computing over the last couple of decades. To summarise:
In the early days it confirmed you were human by seeing if you were smarter than a bot.
Now it confirms you're human by seeing if you're dumber than a bot. 😂
2:52 It was at this moment that everything started going haywire.
2:09 wth so all these years google knew that I've been searching p0rn on a daily basis?!!
Excellent video! Informative without being long winded, short without losing meaning. Saying something concisely yet entertains us too? Perfect.
I hate those stupid image tests;
I always get some object that has literally like one pixel on one square, and I am never sure if I am supposed to click that square or not.
"you passed by exhibiting incompetence at every turn" - is my new favorite phrase
Well done video. Informative and entertaining at the same time, without being long and drawn out. A rarity.
It’s amazing that reCAPTCHA considered the imperfections in humans to track robots. Seems cool for some reason
Excellently written and presented! Thanks for sharing!
The new reCAPTCHA tests are a really good way to determine how well your browser protects your privacy. A properly private browser will fail reCAPTCHA, because it doesn't give the test enough information to determine whether you are human.
I don't have any vision impairments, but those text and number puzzles are such a hassle. The same goes for clicking the images. I have to do 10 of them before I'm accepted. As for the box checking, you can easily macro human mouse movements and click the box. The browser history I'm sure can be spoofed. Am curious how it does against private browsers with tracking blockers.
I wonder how a robot’s search history looks like 😂😅
God damn, I must have a really steady hand because I ALWAYS have to find the damn sidewalk!
"Is that even a f***ing letter?!" Made my day
It is so easy now to reprogram a robot to employ the randomize timer option to insert deviations from that straight line and to reduce the effectiveness of recognition. Thanks for giving the bot makers a clear path to now thwart captcha. Brilliant.
Her voice is soothing.
It's equally scary and brilliant.
I feel like a bot, I once had to do 7 recaptchas in a row
Amazing explanation, thank you
rip people who have an ocd of always moving the mouse in a straight line
The check box mouse-movement test is easy to get around as well: just add noise to the mouse movement, while still landing on the check box - which is super easy to do.
The site test.. why would one need to automate sites anyhow - but the solution is probably to queue random solution searches.
Those curvy letters will be the death of me as a human person though.
Now that is one of the best explanatory vids I've seen. Detailed and funny. Brilliant
1:25 but how about smartphone? Touchscreen? There is no cursor to track
"I Think I see a dolphin!" HAHAHAHA, just perfect!
Ahh, I fail these captchas all the time, so less accuracy might help…. 😂
“Don’t worry, human error looks really good on you, girl.”
After you spent an hour or more to locate where they hide those privacy pages to click on those options that said not to track or share, and then few weeks later, they sent an update to upgrade to a newer version. So you have to go through the same process again and again whenever there’s an update, which until one day, you would say, take whatever you want.
A lot of people commenting without even watching the video and it shows 😅
They saw the title and got mad lmao
using robots to reconize humans from robots by checking the consistency of his memories ? that's literally bladerunner
The mouse movement thing is a red herring thrown in by google, they don't even mention that by finding 'sidewalks' and 'traffic lights' we're actually training self-driving cars
While fairly technically accurate, it is missing a lot of the details that help answer some questions in the comments. It may use google account history as a flag of sorts, but from what I can tell as someone who has implemented it on their site, it primarily tracks you through the site you're on. The tracking script must be on the site for an extended period of time for them to be able to collect enough data to statistically define what is bot versus humans traffic on the site.
This makes so much sense, I just need to make sure to stop moving in a straight line whenever I solve a captcha.
This video is actually pretty good.
Why does the voice audio sounds like some 64kbps audio that was compressed to infinity and beyond was sent to Andromeda and back and then got reuploaded 5 times, then someone played it on a speaker and recorded it with his phone and then it got reuploaded 10 more times just for extra measure?
that "select all squares with..." stuff is so annoying. sometimes there is a tiny bit of edge of the object in a square and i just dont know whether to select it or not. need to try multiple times
Excellent production values, script and even passable acting 😂. Well done.
Great video! I am not a robot.
Got the image test today and instantly thought about this video. I'm impressed by my skill of moving my mouse in a straight line.
1:30 i move my mouse in perfect straight line, lol
That "later humans" line freaked the hell outta me!
Will nobody else mention the tattoo! It’s awesome
Wow didn't know that captcha was an acronym.
Well explained. Regarding the box, I thought it was something like this, which is why I often find myself delaying / taking a circuitous route so it doesn't seem suspiciously robotic. I still get the picture stuff though.
Interesting to hear about the history tracking, and the reason you don't see wobbly text anymore. I believe that was a dual-purpose feature to help digitise books that image processing had failed to do well, but as it explains, that's now so much better than humans so it makes sense that it's been scrapped!
What do you do when you find a recaptcha with never ending images, because I gave up at the sixth attempt.
Very informative, quite entertaining and tightly packed. Good work 🦶
Your assistant is great! 👌🏼
Good information! 👍🏼
(From India) ☺️
Do you work in this channel or this a one off thing? I want to see more stuff like this, educational or preferably purely comedic!
Oops I many to say, they're one of the regular presenters on the TV show. Also look out for their show next Comedy Festival
Why is no one talking about the fact that they check your browser history...
It's not that complicated to add some randomness to the cursor's movement to simulate a shaky or indecisive human, and AI image recognition is getting good enough to render the "click the boxes with" pointless too. As for browsing history, just load up a nice fake history for each new session. None of this is something the average user will do, but it's trivial for the professional botters to do. So, once again, the average user has to suffer while the actual people they want to stop laugh and watch their bots pass the tests anyways.
Hmmm... I almost always get the traffic light or motor bike pictures. I guess I'm an efficient mouse mover.
0:46 😂magic 🧝♂️ eye 👁🗨... 😅🐬 Dolohins
thx! that was new for me :-)
Greetings from germany
2:59
Look!
It’s Santa!
That ending tho, i'm not expected that kind of thing
How this works on touchphones btw
It's simple. Robots can't lie 🤣
❤ thanks for the knowledge
Lol. Love it. That plot twist at the end 😅
I don't imagine it's so hard to program a bot to make human-like mouse movements.
Thank goodness these exist