Learning to love what you hate

A few years ago, the folks at Yahoo were concerned that evildoers were using automated programs to register for thousands of free Yahoo email accounts and then using those accounts to send spam. Some Carnegie Mellon computer scientists, including a grad student named Luis von Ahn, came up with a solution, called CAPTCHAs.

By now everyone has seen them--small boxes at the bottom of Web registration forms that ask the user to retype some distorted letters or numbers. It's a task that's usually easy for humans, but difficult for computers. Some people don't give these little puzzles a second thought; others find them irritating and a few detest them.

Nevertheless, they have proliferated because they do a necessary job. But von Ahn, now an assistant professor of computer science, is convinced he can do better. He can't make them less obnoxious, though he can make them do more work.

A new project, which he has dubbed reCAPTCHA , uses these little puzzles to help digitize books and other documents that predate the computer era. Automated optical character recognition (OCR) systems often fail to read text that has been underlined, smudged or poorly printed in the first place. So he and his team use text that OCR fails to decipher as puzzles and lets people who are registering on Web sites decipher them.

A clever idea? Sure. But is it a good idea? Since word of reCAPTCHAs began spreading in the blogosphere and in the mainstream media yesterday, some bloggers have contended von Ahn is just making life worse. Others have doubts about how well it will work.

Time will tell. But many people seem to think that giving CAPTCHAs a dual use--a use that ultimately will benefit humanity by making all text searchable, not just text generated since 1985--makes this test-taking more acceptable.     

Byron Spice