Computer scientists have developed artificial intelligence that can outsmart the Captcha website security check system.
Captcha challenges people to prove they are human by recognising combinations of letters and numbers that machines would struggle to complete correctly.
Researchers developed an algorithm that imitates how the human brain responds to these visual clues.
The neural network could identify letters and numbers from their shapes.
The research, conducted by Vicarious - a Californian artificial intelligence firm funded by Amazon founder Jeff Bezos and Facebook's Mark Zuckerberg - is published in the journal Science.
Good. Now maybe I can get past Captchas.
(Score: 3, Insightful) by KilroySmith on Sunday October 29 2017, @01:11AM (9 children)
In some areas, being better than a human is a remarkable achievement - Chess and Go were such.
In some areas, humans set a pretty low bar that an AI would have to surpass. Driving and solving Captcha's are like this.
(Score: 0) by Anonymous Coward on Sunday October 29 2017, @01:56AM
That's where slashcode was ahead of the curve. An AI can't determine if something is complete crap but a human can. Unfortunately, (youtube adpocolypse) the powers that be do not like the results and think AI is the answer. [youtube.com] LOL!
(Score: 3, Interesting) by Runaway1956 on Sunday October 29 2017, @02:16AM (6 children)
You're probably right, that humans in general set a low bar for captcha crap. But, I can't even get up to that bar. A lot of them just kick my ass, and I just give up on them, because I just can't see whatever it is that I'm supposed to see.
“Take me to the Brig. I want to see the “real Marines”. – Major General Chesty Puller, USMC
(Score: 0) by Anonymous Coward on Sunday October 29 2017, @04:01AM (2 children)
Google's captchas are absolutely terrible. It'll give you a picture of a guy on a bike and ask you to find the street signs. And you have to pick the guy on the bike if you want to get it right.
(Score: 3, Insightful) by Runaway1956 on Sunday October 29 2017, @09:56AM (1 child)
Actually, those captchas are much better than some of the others. The ones with a bunch of poorly drawn numerals, mixed in with a bunch of squiggly lines were among the worst. And, then there were the ones with poorly written letters, sometimes cursive, which were only a tiny bit better. I much prefer the math ones - a simple equation. Or even a not-so-simple one. Solve the problem put the solution in the box, and you're good to go. A logic question works, or even a history question. Physical sciences? That would work too. I hate that I have bad eyes, and that half the world thinks that hiding text in plain sight is the way to go.
“Take me to the Brig. I want to see the “real Marines”. – Major General Chesty Puller, USMC
(Score: 0) by Anonymous Coward on Sunday October 29 2017, @07:23PM
I have poor eyesight as well and have many friends that are blind or near-blind. I challenge almost anyone who thinks that visual CAPTCHAs are bad to try the audio ones. Some of those are just insane. They are random assemblages of noise and whatever. The Google ones can be frustrating too, as the answers have to match what most people put in and the audio is ~4 seconds clipped from random Google Voice phone messages. So, do you put the whole word in when it is cut off or just part? When two people talk over each other, which is the right answer? What about messages in foreign languages? Literal gibberish (like young children)? Completely made up of noise?
(Score: 0) by Anonymous Coward on Sunday October 29 2017, @05:28AM
Maybe that's because you mistakenly think your hero is called "Methanol-Fueled", not "Ethanol-fueled."
(Score: 0) by Anonymous Coward on Sunday October 29 2017, @12:45PM
Sounds like the captchas are working as designed.
(Score: 0) by Anonymous Coward on Sunday October 29 2017, @06:35PM
I have great eyesight (official test a couple of months ago), but within the last year I think I needed to repeat two or three captcha tests.
They're just dumb.
(Score: 0) by Anonymous Coward on Sunday October 29 2017, @02:37AM
Unless you are using surpass the bar in a strange way, I think you have it the wrong way round. Chess and Go are fixed rules in a limited universe.
Solving self-driving is going to be much more harder and more complicated. Captchas are somewhere in the middle depending on their difficulty and acceptable failure rate.
(Score: 2) by takyon on Sunday October 29 2017, @01:48AM (3 children)
They will phase out captchas within 3 years and replace it with "AI" to detect and delete spam and objectionable content.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 1, Touché) by Anonymous Coward on Sunday October 29 2017, @02:01AM
Dear Takyon,
Soylent news comment #588870 [soylentnews.org] has been deleted due to spam and objectionable content. Please click here [google.co.uk] to help resolve the issue. Hate speech is not tolerated, thank you for understanding.
(Score: 0) by Anonymous Coward on Sunday October 29 2017, @02:44AM
https://xkcd.com/810/ [xkcd.com]
(Score: 0) by Anonymous Coward on Sunday October 29 2017, @12:47PM
I may be "objectionable content" (just ask my ex) but I am not going down without a fight.
(Score: 0) by Anonymous Coward on Sunday October 29 2017, @03:07AM (4 children)
It is not, there is no consciousness and never will be , can ray please upload himself ASAP because we need to stop with the useless AI, oh yay please DIAF
(Score: 0) by Anonymous Coward on Sunday October 29 2017, @03:12AM (2 children)
Speak for yourself, meatbag.
(Score: 0) by Anonymous Coward on Sunday October 29 2017, @03:42AM (1 child)
Well that is about as bright as AI but not as bright as bender
(Score: 1, Funny) by Anonymous Coward on Sunday October 29 2017, @05:11AM
Bite my shiny virtual ass!
(Score: 0) by Anonymous Coward on Sunday October 29 2017, @12:50PM
The ghost of Steve Jobs disagrees.
(Score: 2) by krishnoid on Sunday October 29 2017, @03:18AM
The AI's don't need to be smart enough to fool the captchas. They just need to be smart enough to fool us into believing they're not in cahoots.
(Score: 1, Informative) by Anonymous Coward on Sunday October 29 2017, @03:46AM
People that invented them have spoken about them on this site, they are good at bounded problems, but they are not intelligent they are not even close to the SCA
(Score: 0) by Anonymous Coward on Monday October 30 2017, @02:57PM
Will this mean that A.I.'s can do Free Labor for Google instead of us meatbags?
So an AI will now train an AI?