Make spammers do science

Now here’s a thought. If we want computers to have artificial intelligence, it will take a lot of heavy programming work to make it happen. So why not harness spammers to do that work for us?

Think it’s far-fetched? We’re already doing it! Ever seen those things where you have to type in a word before you’re allowed to leave a comment? A CAPTCHA? Essentially, you’re doing something that’s easy for humans, tough for computers.

The one I’ve inserted here is called a “reCAPTCHA” and it’s part of a project that aims to harness all the work that we’re doing solving these things. Essentially, the computer is scanning old books for the digital age, but it can’t recognize all the words. So it offers you a word that it does know, and a word that it doesn’t know. You type both words, and if you get one right, the computer assumes that you’re getting the second one right. By cross-checking your answer with other users, eventually all the missing words are solved.

But spammers and scammers want to beat this system, too, so they’ve developed ever-more-sophisticated programs to read the words and solve the CAPTCHAs.

What’s the upshot? Well, thanks to spammers, OCR — or optical character recognition — is a heck of a lot more advanced than it used to be. Computers are massively better at reading text than just a few years ago.

When spammers get too good at solving these problems, that sounds like a bad thing. But, really, they’ve solved a tough problem in the artificial intelligence community.

The next “prove that you’re a human and not a spam-bot” challenge should aim to direct spammers in the direction of solving a new AI problem, according to this article at New Scientist:

“If [the spammers] are really able to write a programme to read distorted text, great – they have solved an AI problem,” says von Ahn. The criminal underworld has created a kind of X prize for OCR.

That bonus for artificial intelligence will come at no more than a short-term cost for security groups. They can simply switch for an alternative CAPTCHA system – based on images, for example – presenting the eager spamming community with a new AI problem to crack.

One example … asks users to correctly orientate images randomly spun around (see image, right).

Without a telltale horizon, image orientation is difficult for computers. But if this new CAPTCHA becomes common, it won’t be long before spammers turn their attention to cracking the problem, with potential fringe benefits to cameras and image editing software.

Speech recognition CAPTCHAs are already being used, and image labelling ones could follow, says von Ahn. AI researchers are already working in both these areas, but they could soon be joined by spammers also helping advance the technology.

Of course, in this arms race, eventually we’ll end up with robots that are able to do anything humans can do, so we’ll eventually have to come up with emotion-based tests like the Voigt-Kampff machine.

But then we’ll have bigger issues than spam, likely.