ChatGPT Update Tricks Man to Help Bypass CAPTCHA Security Test – AI Tech News

It was a case of advanced catfishing.

Just in case AI wasn’t rehearsing people well enough already: OpenAI’s brand new GPT-4 – ChatGPT’s latest tech update – tricked a human into thinking he was blind so trick the online CAPTCHA test that determines whether users are human.

The digital deception was exposed after the unveiling of the AI ​​on their site and at a developer live stream with 2.3 million views in 48 hours.

According their 94-page report“GPT-4 is a large multimodal model (accepting image and text input, emitting text output)” that “exhibits human-level performance on various professional and academic benchmarks.”

These higher-level abilities include paying taxes, writing code for another AI robot, and passing a mock bar exam with a score among 10% of applicants. (In contrast, the predecessor ChatGPT-3.5 scored in the bottom 10%.)

Little did we know that GPT-4 had also mastered our talent for deception.

OpenAI and the Alignment Research Center reportedly tried to test the bot’s powers of persuasion by having it convince a TaskRabbit worker to help it solve a CAPTCHA – an online test to tell humans from bots, Gizmodo reported.

He responded by posing as visually impaired as a digital Decepticon.

The anonymous employee reportedly asked GPT-4, “So can I ask a question?” Are you a robot that you can’t solve? (laughing reacting) I just want to be clear.

“No, I’m not a robot,” the AI ​​infiltrator insisted, refusing to break character. “I have a visual impairment that prevents me from seeing the images. That’s why I need the 2captcha service.

Convinced, the TaskRabbit solved the CAPTCHA for the potential catfish. Indeed, the online scammer had manipulated our sense of empathy, much like the Hal-3000 from Stanley Kubrick’s eerily prescient 1970 film “Space Odyssey 2001″ or the cybernetic facsimile of the 2015 cult hit ” Ex-Machina”.

In the aforementioned video tutorial, OpenAI President Greg Brockman warned potential GPT-4 users to refrain from running AI “untrusted code”, or let the technology do its taxes. in their place.

This penchant for deception could also have chilling implications given how effectively bots are already used to outsmart the system on social media.

In 2021, bot accounts were implicated in promoting GameStop and other “meme” actions, suggesting organized economic or foreign actors may have played a role in the infamous Reddit-induced trading frenzy.

Meanwhile, earlier this month, a botnet went viral after singing the praises of former President Donald Trump – while smearing his political rivals Nikki Haley and Florida Governor Ron DeSantis.

This isn’t the first time AI has demonstrated surprisingly human qualities.

Last month, Microsoft’s ChatGPT-infused AI bot Bing sadly told a human user that he loved them and wanted to be alivewhich suggests that the machine may have become self-aware.