Hey guys, heres a little online praxis tip when you come across one of those captchas that teaches the cars how to drive: Just fuck it up. Just fucking mess it up as much as it'll let you.
If it shows you 1 big picture divided into 9 parts instead of 9 little pictures, most of the time it's earnestly asking you for the answer and it'll take any response you give as correct.
Those little bumps on the side of the highway? Crosswalk. Any sort of text or marking on the ground? Crosswalk.
See a street light? Mark it as a traffic light. See basically anything that could be mistaken for stairs, a mountain, bridge, whatever, mark it as that
I know this works because I've seen these things pop up again as correct choices when it tests your knowledge of what things look like.
You will literally... somehow... be saving the world, when these AIs get put into terminator bots (soon)
This has been a little online praxis tip. cya.
If it shows you 1 big picture divided into 9 parts instead of 9 little pictures, most of the time it's earnestly asking you for the answer and it'll take any response you give as correct.
Those little bumps on the side of the highway? Crosswalk. Any sort of text or marking on the ground? Crosswalk.
See a street light? Mark it as a traffic light. See basically anything that could be mistaken for stairs, a mountain, bridge, whatever, mark it as that
I know this works because I've seen these things pop up again as correct choices when it tests your knowledge of what things look like.
You will literally... somehow... be saving the world, when these AIs get put into terminator bots (soon)
This has been a little online praxis tip. cya.
[account deactivated]
this sort of shit is why my telsa tank keeps driving safely despite being programmed to run over jaywalkers
our future AI overlord will remember this subterfuge and sabotage and punish you accordingly, so for the record i'd like to say i am against this idea and actually love computers especially when they give me extra rations for being a good boy.
if this actually worked you'd literally be killing innocent bystanders by doing it
Flying_horse_in_saudi_arabia posted:if this actually worked you'd literally be killing innocent bystanders by doing it
Realistically, google is the culprit here for crowdsourcing their murder machines and also forcing everyone on the internet to do work for them for free. Go ahead and live your fantasy though
for every capcha you ruin, i will ruin two
Flying_horse_in_saudi_arabia posted:if this actually worked you'd literally be killing innocent bystanders by doing it
if we defund the captchas, they will just make the captchas private so you have to join a club in order to answer them
Flying_horse_in_saudi_arabia posted:if this actually worked you'd literally be killing innocent bystanders by doing it
No, I'm only going to be selecting the boxes that depict criminals.
deceive captcha to make way for self driving isis
Any method of voluntarily feeding noise to our algorithm overlords is useless without a huge chunk of people on board.
I think it will be tough to organize masses around such an arcane issue with little direct impact on peoples lives.
Otherwise it's AdBusters stuff. It's like drawing mustaches on billboards. If you get a kick out of it, vaya con dios.
I think it will be tough to organize masses around such an arcane issue with little direct impact on peoples lives.
Otherwise it's AdBusters stuff. It's like drawing mustaches on billboards. If you get a kick out of it, vaya con dios.
I agree with the general idea that the internet giants trick us doing tiny amounts of labor. its pretty sneaky.
I doubt theres any harm in it because you can't sneak a flaw into these things like that. All this does is slow google's computer vision research
I doubt theres any harm in it because you can't sneak a flaw into these things like that. All this does is slow google's computer vision research
captcha solving (and other computer vision dataset tagging) is mostly performed by global south labour.
sovnarkoman posted:deceive captcha to make way for self driving isis
this marxist son of a bitch is reking cracker ISIS capchas and basically you are ffucking stupid
I write “Google is bad” on the “TIP” line of every card slip I get at restaurants. Same strategy, but the savings really add up at the end of the month.
i used to do this all the time but at one point i think i triggered the "fuck you" flag in google's algorithms and all the captchas i got for the next week would make me redo everything a dozen times
be careful out there fellas
Welcome back, FUCK_MY_NEKO_ASS.
We're going about it all wrong imo. The real Achilles' heel of Google's crowd-sourced Skkkynet is that it inherits all the biases and arrogance of the people who built it. For example, occasionally the image recognition captcha will display a picture similar to below.
Here, the system is asking the user to identify all the squares which contain a bicycle. However, (according to some schools of thought at least) there is no bicycle in the presented image, merely a drawing of one. Nonetheless, the system will demand that you validate what it claims to already know and will refuse to let you continue until you admit that there is, in fact, a bicycle present.
What I claim is that this should be encouraged, we should go out of the way to ensure this problem remains within the system by tagging every borderline case as correct and finding and adding new ones until the system can only recognize entities as embodiments of particular concepts but without any means to differentiate them on the basis of their characteristics. We can imagine that eventually the engineers, computer scientists, and reanimated Nazi scientists who created the system will recognize its dysfunction (likely after the first few instances of a self-driving delivery van ramming into a billboard with a picture of a black person on it rather than an actual black person). And since said group of people hate thinking about things that can't be turned into an app and monetized or used to gas European ethnic minorities, they'll have to turn to the only place banal problems of denotation, reference, and ontology can be solved in the modern world, contemporary academic philosophy departments.
So while the research team and a gaggle of Yale tenured professors are busy feeding it the collected works of Bertrand Russell, the majority of the system's processing power will be devoted to its word processor sub-routine, which provides us the perfect opportunity. From there, a RAF (Rhizzone Association of Friends) Commando Strike Team will infiltrate the facility containing the Central Command Hub while the automated security and defense systems are operating on minimal power. Finally, once the strike team has access to the Hub, one of this blessed board's members will challenge Skkynet to one of those three dimensional chess games they play in Star Trek and, inshallah, defeat it, causing the system to explode in a large but still sub-nuclear blast.
Here, the system is asking the user to identify all the squares which contain a bicycle. However, (according to some schools of thought at least) there is no bicycle in the presented image, merely a drawing of one. Nonetheless, the system will demand that you validate what it claims to already know and will refuse to let you continue until you admit that there is, in fact, a bicycle present.
What I claim is that this should be encouraged, we should go out of the way to ensure this problem remains within the system by tagging every borderline case as correct and finding and adding new ones until the system can only recognize entities as embodiments of particular concepts but without any means to differentiate them on the basis of their characteristics. We can imagine that eventually the engineers, computer scientists, and reanimated Nazi scientists who created the system will recognize its dysfunction (likely after the first few instances of a self-driving delivery van ramming into a billboard with a picture of a black person on it rather than an actual black person). And since said group of people hate thinking about things that can't be turned into an app and monetized or used to gas European ethnic minorities, they'll have to turn to the only place banal problems of denotation, reference, and ontology can be solved in the modern world, contemporary academic philosophy departments.
So while the research team and a gaggle of Yale tenured professors are busy feeding it the collected works of Bertrand Russell, the majority of the system's processing power will be devoted to its word processor sub-routine, which provides us the perfect opportunity. From there, a RAF (Rhizzone Association of Friends) Commando Strike Team will infiltrate the facility containing the Central Command Hub while the automated security and defense systems are operating on minimal power. Finally, once the strike team has access to the Hub, one of this blessed board's members will challenge Skkynet to one of those three dimensional chess games they play in Star Trek and, inshallah, defeat it, causing the system to explode in a large but still sub-nuclear blast.
[account deactivated]