Completely Automated Public Turing Tests to tell Computers and Humans Apart (CAPTCHAs) appear to be heading toward obsolescence, with Vicarious demonstrating an algorithm that can solve the test with a recursive cortical network by using 5,000 times fewer training images than other methods.
This and other advances are making CAPTCHAs less secure and less relevant, pushing researchers toward a rethinking of the test to strengthen it in the era of machine learning.
Nan Jiang at Bournemouth University in the U.K. has developed a mobile CAPTCHA named Tapcha, which builds on the traditional distorted text approach while still relying on human knowledge. "We use this approach to create the instruction," Nan says.
He believes a machine can only defeat this approach by understanding what is occurring in the image and coming to a conclusion on its own, a breakthrough that currently is very difficult to realize with artificial intelligence.
From Wired.co.uk
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA
No entries found