Overturned CAPTCHA security measures surrendered to innovative AI technique
In a groundbreaking development, scientists have created an innovative artificial intelligence (AI) system capable of cracking CAPTCHA systems designed to distinguish between human and automated website access. The new AI system, named 'TechXplore', was developed by a team at Vicarious AI based in California.
The key to this new AI system is a modified neural network known as a recursive cortical network. Unlike traditional neural networks, where nodes hold new information and a network is built from these nodes to judge new data, the recursive cortical network uses a different approach.
Recursion, a software technique where data is used to learn something new, is at the heart of this innovative system. In the recursive cortical network, the results of the learning process are fed back into the software, allowing the system to refine its responses over and over until a solution is reached.
By applying recursion to a neural network, the researchers found they were able to reduce the learning curve of their software dramatically. Previously, researchers have shown that defeating CAPTCHA was possible using neural networks, but it required the system to study thousands or even millions of examples. However, the new AI system, developed by Vicarious, needed just five training steps to crack Google's reCAPTCHA 67 per cent of the time.
The system's ability to crack CAPTCHA after studying just a few examples is a significant improvement over previous methods. CAPTCHA challenges users to prove they are human by recognizing distorted combinations of letters and numbers. The new AI system uses a different approach, but the details of this approach remain confidential, as the researchers have chosen not to reveal the specific names or identities of those who developed the system.
The development of this new AI system could potentially allow bots to access websites that currently use CAPTCHA for protection. While the implications of this are far-reaching, it is important to note that the system's creators have not disclosed any plans for malicious use.
Recursion has long been used to solve mazes, and its application to AI systems opens up exciting possibilities for the future of artificial intelligence. As more research is conducted and the capabilities of these systems continue to grow, it will be interesting to see how they are used and what impact they will have on our digital world.
