NEWSResearcher Working With Royal Air Force Created an 'Undetectable' AI
Nov. 28 2023, Published 6:31 a.m. ET
Amid the clamor surrounding artificial intelligence’s impact on academia, an emergent tool promises to throw a wrench into established AI detection systems.
Bars Juhasz, a 25-year-old PhD candidate at Loughborough University in the UK, has developed a new tool called Undetectable.ai, that makes it significantly more difficult for AI detectors to identify text written by artificial intelligence systems.
Juhasz is currently pursuing a PhD in Computer Science, with a focus on AI and machine learning.
His thesis explores unmanned aircraft system operations in denied command and control environments. In addition to his academic work, Juhasz says he collaborates with the Royal Air Force on projects involving AI research.
When asked for comment, Bars Juhasz told OK! Magazine he “can’t go into the specific details,” but said his research activities “focus on areas of foundational and disruptive machine learning.” Juhasz clarified though that he isn’t file or rank.
And it seems evident that being disruptive is a core trait of Juhasz. In fact, the Gloucester-based co-founder has helped disrupt the entire industry of AI detectors.
This is where Undetectable AI enters the frame – a platform specifically designed to hide AI-written words from the gaze of detection tools like Turnitin. Juhasz co-founded Undetectable AI alongside Christian Perry in May of 2023.
Undetectable AI has gained a high level of notability, from people on both sides of the fence.
Undetectable AI serves dual purposes. The official Undetectable.ai website describes the tool as an “AI text detector and humanizer”.
Not only is it a sophisticated AI detector, but should the content be flagged as AI-generated, its algorithms work to 'humanize' the text.
The goal? To make it undetectable to current AI detection models, and thereby preserving what one could argue is the integrity of automated content.
But the repercussions of such a tool could be considerable. This could be the type of AI tool that makes people panic.
While programs such as Originality.ai boasted around a 99% success rate in pinpointing the creations of AI language models like GPT-4, Undetectable.ai was shown to reduce this figure to approximately 27%, rendering Originality.ai nearly useless (according to a recent AI detection study published in Nature).
What Juhasz has crafted, is a tool that doesn't merely ensure grammatical precision; it simulates the fluid and occasionally unpredictable characteristics of human expression.
Critics, however, aren’t applauding.
Among the naysayers, some (in the AI research community) have dubbed Undetectable AI "A modern threat to academia."
The concerns are not unfounded. Proponents of educational integrity could argue that tools like Juhasz's could further erode the bedrock of academic originality.
Such advancements compel us to interrogate the ethics of AI in learning environments and the complex relationship between technology and human output.
The core technology of Undetectable AI relies on advanced algorithms and linguistic models, which strategically blend AI-generated text with human-like elements, making it exceptionally challenging for detection tools to do their job.
Founders Bars Juhasz and Christian Perry explained in one interview, that “Undetectable AI writes like a human”
This leap in AI-text integration illustrates a profound shift towards seamless human-AI content creation synergy.
However, with the development of any influential tool comes a counterpart designed to circumvent its capabilities.
Undetectable AI reflects this technological cat-and-mouse game, offering a means to bypass the very systems established to uphold originality standards.
Supporters of Juhasz’s tool might argue that it represents an industry evolution, mirroring broader movements towards AI-augmented productivity.
But this raises pivotal questions on the devaluation of human creativity and the preservation of intellectual propriety, tentpoles of academic and journalistic fields.
Academic institutions especially, already grappling with the challenges posed by AI in plagiarism, may find Undetectable.ai a formidable foe.
On the flip side, Undetectable AI poses as a catalyst for innovation in detection technology, demanding more sophisticated methods to discern AI from human input. It’s a testament to the rapidly evolving dynamics between human ingenuity and artificial intelligence.
But it's certainly worth thinking about how Juhasz’s invention will affect the process of writing and content creation.
Will it prioritize subtle and nuanced writing over unmistakable veracity? Will it necessitate a new era of detection tools, or will it compel writers to re-engage with their creative faculties more deeply?
The answers to these questions remain at the immediate crossroads of technological advancement and ethical pragmatism.
Undetectable.ai sits somewhere in between ingenuity and controversy. Devan Leos an ex-Disney celebrity and founding member of Undetectable AI, told RadarOnline that the tool wasn’t for cheating—but some disagree.
As such tools become more sophisticated, the lines between artificial and human creativity may not only blur but also call into question the very essence of original thinking.
It’s a conundrum that begs a critical evaluation of our relationship with technology — where to draw the line, and at what cost to intellectual culture? For now, only time will tell.
