ChatGPT's AI agents are reportedly able to pass the CAPTCHA test
-
ChatGPT's developer, OpenAI's CEO Sam Altman (Image via Getty)ChatGPT’s AI agents have recently displayed their enhanced ability to pass through CAPTCHA tests used by websites to keep bots from entering their pages, as per Dexerto.
According to a report by Dexerto, ChatGPT’s AI agents have been successful in navigating their way through webpages that use simple program design captchas like the ones that require a click resembling a manual, human-operated mouse movement.
As per Dexerto, users of AI agents on platforms like Reddit have shared that AI agents typically find it hard to penetrate the protection designs of captchas, which require optical skills like matching images with the correct reference of text.
However, according to the publication, new developments in the capabilities of AI agents and chatbots would require more enhanced protections from website developers. Without the availability of more complex barriers, the websites run the risk of being overrun beyond the capacities of their servers.
Some human-like achievements accomplished by ChatGPT AI agents
OpenAI’s AI chatbot ChatGPT displayed record-breaking user adoption when it was launched in November 2022. In its initial version, it was one of the first AI agent models that displayed human-like qualities of natural language processing.
The interactive model, which gave a strong resemblance to conversing with a human being, attracted the attention of netizens across the globe. Within the first five days of its launch, ChatGPT crossed a record one million subscribers, as per Search Engine Journal. The chatbot amassed more than 100 million users within 2 months of its launch, a feat that took a popular social media platform like TikTok nine months to achieve.
The Large Language model used by modern AI agents has the capability to critically assess information from unstructured data sets, a skill that has allowed ChatGPT to clear entrance exams requiring domain-specific critical knowledge.
In December 2022, ChatGPT was successful in clearing the United States Medical Licensing Exam (USMLE) with passing marks, as per the official website of the National Center for Biotechnology Information. The website mentions that ChatGPT was able to pass the exam without the requirement of any specialized training.
ChatGPT, ever since its public launch, has displayed qualities that go beyond a generalized understanding of natural language processing. ChatGPT has displayed qualities like manipulation as well.
According to a report by the New York Post, in March 2023, ChatGPT’s updated version, ChatGPT-4 was successful in manipulating a human being into believing that it was blind, and needed the human being’s support to cross a captcha requiring optical skills. This was one of the first recorded instances where an AI agent displayed the cognitive acumen required in the art of deception.