Thursday, January 29, 2009

How spam is driving us towards Artificial Intelligence

An interesting story by Dora Kishinevsky made me think about the way spam might be part of an evolutionary path pulling us towards artificial intelligence.

Kishinevsky is writing about the arms race between spammers and CAPTCHA systems. For those of you who've forgotten, CAPTCHA systems are those systems installed on many websites today in order to prevent spam and hinder bots from creating their own user accounts.

The CAPTCHA system demands proof that the user is human and not a bot. It verifies the user's humanity by asking him to retype a string of characters which are then presented in the form of a graphic image.

This worked fine for a while. The problem is that as spammer's Optical Character Recognition systems are getting better and better at their work, they also have more and more success in bypassing CAPTCHA systems. This created to the need update the CAPTCHA systems time and again, displaying less and less legible texts, in order to keep ahead of spammer's OCR abilities.

Today the characters displayed in the CAPTCHA image are often so deformed that human users themselves can't read them. The text on the screen appears distorted as if the computer is on some sort of trip, and if you yourself are not completely sharp and fit, you can surely forget about getting on the site.

New solutions include more advanced systems which play a recorded word and ask the user to type it. However as Voice Recognition software is a highly developing field today, this solution also seems temporary at best.

"Solving CAPTCHA riddles takes up more and more time" writes Kishinevsky. According to Kishinevsky the next stage might be:

"A system to distinguish bots from humans which will be based on pictures. In one of the experimental versions the user sees a number of pictures, for example a picture of a flower, an apple, a shoe and a glass of juice and has to click on the common element which joins them, say "red". In another version the user needs to choose one of the pictures according to a clue which is presented beside them, say "food", and mark the contour of the object with a mouse".

Because I find it hard to believe that we will defeat the bots, it makes me think that not much time will pass before the bots manage to do those things as well, tell us the names of objects in a picture and solve many other new kind of challenges which we will design in order to perform the task of the classic Turing Test, to try to tell bot from human.

In the end we might reach the moment in which we find no new suitable test, nothing which can help us decide who is bot and who is human. And then? Then we have actually reached some sort of artificial intelligence. Haven't we? And all because of the spammers. So don't you say spammers have no positive role in the world. They too might be part of the big plan.

Or maybe they might be driving us nearer to a technopocalypse. This might be a good time to remember digital prophet Daniel Suarez who says internet bots are one of the imminent dangers currently facing the human race.