Smart computers - what about Captchas ?

[matt-browne-sfw]

By 2015 all new personal computers will have the computing power and intelligent software to recognise any Captcha character patterns. Do you agree?

In science fiction a lot of discussions revolve around the Turing test. Captchas are a somewhat simpler problem, but still complex enough that today in 2008 only humans can recognize, say more than 99.9% of all Captchas devised. Complex pattern recognition is a human brain stronghold.

A bit of a background. Here's what Wikipedia has to say:

CAPTCHA - Wikipedia, the free encyclopedia


A CAPTCHA is a type of challenge-response test used in computing to determine whether the user is human. "CAPTCHA" is a contrived acronym for "Completely Automated Public Turing test to tell Computers and Humans Apart", trademarked by Carnegie Mellon University. The process involves one computer (a server) asking a user to complete a simple test which the computer is able to generate and grade, but not able to solve on its own. Because computers are unable to solve the CAPTCHA, any user entering a correct solution is presumed to be human.

Any thoughts?

[jp306972]

Did you type this message? I'm betting you did.

How much did the general public use Optical Character Recognition and Voice Recognition Software. After seeing Scotty speak into the mouse in one of the early Star Trek movies I was lead to think these technologies should be in use by now. What happened, how come this stuff isn't being used today?

Captchas might be cracked by 2015 but will the public use this technology for things besides cracking Captchas or will it be something that only spammers are interested in for cracking Captchas. Maybe by 2015 people will be finally begin speaking to their computers.

What would be a future use of technology that could crack a Captcha?

JP

[Steve Jordan]

Originally Posted by jp306972

What would be a future use of technology that could crack a Captcha?

One thing that comes to mind is verification cracking. Presently a version of Captchas are used to verify that a human, and not a computer, are accessing/sending e-mails or secure web sites (those small images of distorted letters and numbers that are randomly generated, and must be physically retyped to ID you as a human). Once computers can handle Captchas, this system goes right out the window, forcing us to develop better identity verification systems to secure sites and data.

Outside of spying on terrorists and other hostile countries/organizations, I can't see that as a positive development. But if it helps us to get to better verification, such as biometric data, it will have served a good purpose there.

Another thing would be defeating camoflage. The human eye is excellent at seeing through camoflage, given enough time... but in case of war or similar attack, the attacker often does not give you that time. If a Captcha-enabled computer system were available to scan ahead, it could recognize camoflaged people and devices in milliseconds, and warn soldiers of impending danger or ambush.

[Lucien21]

I actually think speech is an inefficient and undesirable interface for computers. I can't imagine sitting in an office with everyone chatting to the PC.

A neural or thought controlled interface would be more efficent, but an old fashioned touch technology works fine (touch screen interfaces are the new fashion)

[Dave]

Originally Posted by Lucien21

A neural or thought controlled interface would be more efficient, but an old fashioned touch technology works fine (touch screen interfaces are the new fashion)

Sight-controlled or eye-tracking controlled interfaces are already possible and in use.
Eye-controlled Computer Operation
They can be used by paraplegics, but I expect they are currently too expensive for widespread use, otherwise I thought they would be more common.

[pyan]

Originally Posted by Lucien21

I actually think speech is an inefficient and undesirable interface for computers. I can't imagine sitting in an office with everyone chatting to the PC.

Agreed...would you really want someone to hear everything you were putting into the computer? Might be a bit awkward if the boss is passing....

[n25an]

Now here's the thing... you are making the presumption that the office will be a brick and mortar building that you will have to report to every day... with the way things are going... it might just be cheaper if the officer were portable... any place you would want it to be... not just a place where your boss can see you... with the only constant being you need to turn the work in ontime... and with the evolution of employees... the boss just might not even want to see you except in a teleconference...

I propose that offices will go from being a building to back to the idea of a title... and in such a situation... talking to your computer is possible... it might be the only sane producing factor... and with the constant move towards automation... it may be the only way for some companies to compete...

[matt-browne-sfw]

Chatting with computers could be interesting in a car. Unless you let the computer do the driving then you got your hands free to use a keyboard...

[Lenny]

I've just come across on article on ZDNet about Microsoft's CAPTCHA being broken, and thought of this thread.

The article
The paper about it

The article gives a few examples of CAPTCHA's being successfully broken from as early as July 2006 - a number of which were broken using publically available, albeit possibly illegal, software.

By 2015 all new personal computers will have the computing power and intelligent software to recognise any Captcha character patterns. Do you agree?

It definitely looks to be going that way, and I agree that by 2015 it will be possible for every computer to do it, if not earlier.

[Dave]

So, does that mean that their computer has beaten a Turing Test? or does it simply mean that CAPTCHAs are not good enough to be classed as a Turing Test?

[Lenny]

I'd imagine that the programs used to crack the CAPTCHAs have been programmed to recognise specific patterns, which I think would mean that CAPTCHAs aren't good enough to be classed as a Turing Test. I haven't a clue how it's been done, but I'd guess that the people who have created programs to crack CAPTCHAs took examples of the different CAPTCHAs (some are shown in the ZDNet article) and used some sort of highly advanced recognition software (maybe something like your standard handwriting recognition software that you can train to recognise your handwriting).

You can program a computer to recognise inputs and respond accordingly (like cracking CAPTCHAs, or things like AIM bots which respond to your text based on keywords within that text), but although it gives the impression of intelligence, it doesn't actually demonstrate its own intelligence, rather it uses a database of responses.

[Interference]

Originally Posted by jp306972

Maybe by 2015 people will be finally begin speaking to their computers.

I hope I'm not around the office when that starts

[PTeppic]

People already say things to the computers in our office... not, generally, the sorts of things that can published, but "things"

[Dave]

You mean like this:

[PTeppic]

Yes, just like that...