AI, Douglas Hofstadter and the Letter A: Vision, Recursive Cortical Network (RCN) & CAPTCHAs

AI, Douglas Hofstadter and the Letter A: Vision, Recursive Cortical Network (RCN) & CAPTCHAs
1 Star2 Stars3 Stars4 Stars5 Stars (No Ratings Yet)
Loading...

A blog post from the AI company Vicarious, says that they have developed a computer model that can solve Completely Automated Public Turing Tests to Tell Computers and Humans Apart (CAPTCHAs).

Vicarious co-founder Dileep George says the model cracks a CAPTCHA’s defenses by parsing the text the test presents more effectively than earlier deep-learning models and with less training using a Recursive Cortical Network (RCN) vision model.

Traditional deep learning models such as Variational Autoencoders (VAE) and Convolutional Neural Networks (CNN) start the learning process from scratch while the Recursive Cortical Network (RCN) model used by Vicarious, is an object-based model that assumes factorization of contours and surfaces, and objects and background.

CAPTCHA-style letter A: The combinatorial number of ways in which this letter can be rendered and recognized by humans, without being explicitly trained on those kinds of variations (Vicarious)

Quoting Vicarious blog: “In 2013, we announced an early success of RCN: its ability to break text-based CAPTCHAs like those illustrated below (left column). With one model, we achieve an accuracy rate of 66.6% on reCAPTCHAs, 64.4% on BotDetect, 57.4% on Yahoo, and 57.1% on PayPal, all significantly above the 1% rate at which CAPTCHAs are considered ineffective. When we optimize a single model for a specific style, we can achieve up to 90% accuracy.

Source: Common Sense, Cortex and Captcha. Vicarious. October 26th, 2017.

Photos Credits: Vicarious