Friday, December 5, 2008

The future of CAPTCHAs can be ill-fated

According to an article in Forbes, the future of CAPTCHAs can be ill-fated.

Because every day spam software writers develop their spam bots to evade the protection provided by randomly distorted text images.

Luis von Ahn says that reCAPTCHA's audio version is unbreakable for now, but with a cost: the humans too can't solve the puzzle 30% of the time.

So we should start looking at other alternatives for recognizing humans apart from the computers. A company called Pramana works on a way of recognizing humans by observing their mouse movements and other behaviour while browsing the web page.

But I think this alternative is much more vulnerable compared to the current CAPTCHA technologies. This is because the attacker can too observe the behaviour pattern of a sufficient number of users and then build derivations originating from them.

Tuesday, November 25, 2008

Another distributed work

Students are required to annotate some kind of genomic data.

Human computation is about to rise in popularity with so much examples around all disciplines.

humans team with AI

I recently heard about a game which is claimed to have been the first example of in which the player cooperates with some AI players.The name of the game is GIVE.

I couldn't investigate the game throughly because of its performance problems. But the idea seems to be good.

People are told to do various things by the computer which probably randomly genarates a task list and hands it over to the natural language generation component. Then this NLG component gives the user some directions, and I think at the end the NLP component is evaluated relative to the success of the players.

My criticism for this game is that the design could be accomplished so that the game is more fun and the evaluation is done implicitly.