American startup has developed a neural network that recognizes the popular CAPTCHA with an accuracy of more than 90%

    Technology startup Vicarious has announced the development of a solution that allows it to successfully pass modern CAPTCHA tests , including the most popular reCAPTCHA on the modern Internet , acquired in 2009 by Google.

    Using machine learning and using the principles of the structure of the human brain, researchers were able to achieve 90% CAPTCHA recognition accuracy from Google, Yahoo, PayPal, and other projects. This progress shows that modern CAPTCHAs are no longer effective as a Turing test .

    According to Scott Phoenix, co-founder of Vicarious, the system uses a minimal amount of computing power and differs from traditional neural networks operating on clusters and devouring megawatts of energy, such as IBM Watson. We managed to achieve such results by creating an architecture as close as possible to the human brain.

    The developers of his company, who created the algorithm, do not pursue any selfish goals and do not intend to sell the technology. At the same time, the main idea of ​​the project is to create a "machine that works like a human brain."

    Recognition accuracy for different types of CAPTCHA:

    Unofficial data from

    Also popular now: