As the internet made its way into homes across the globe in the late 1990s and early 2000s, it presented an unprecedented opportunity for people everywhere, presenting unlimited access to boundless information. In this way, the mechanization of the internet and computers that enable users to access it ultimately provided a platform for greater human interactions. By providing a common language, setting, and means of communication, people from different corners of the world could explore one another’s cultures, beliefs, and writings like never before.
However, through more recent developments, Artificial Intelligence (AI) and Machine Learning (ML) tools have become widespread across the internet. AI-powered bots are now found all across social media websites, engaging in full-blown conversations with humans. Bots are even creating AI-generated imagery that utilizes the sum-total of image libraries readily available on the internet to fuel digitized image generators such as Portrait Pal. With the sheer quantity of AI-generated writing proliferating across the internet, it’s easy to see why someone might be interested in using an AI Detector.
The Value of AI Detection
The purpose of AI detection is to better equip, identify, parse, and ultimately understand the role that AI-generated content is playing in the modern internet ecosystem. For example, AI Detector offers universal AI detection that is capable of accurately detecting content from generators and sources such as ChatGPT, GPT-4, Claude, Gemini, and other leading AI models. The system offers real-time analysis, allowing users to get instant results that provide detailed, comprehensive reports that include AI probability scores and further insights.
With the overt reliance upon and surge in popularity of AI technology escalating so rapidly over the course of the last year or so, it has been incredibly difficult for discerning observers to even keep up. Companies such as Google and Meta (owners of Facebook and Instagram, just to name a few) have fully embraced AI within the last several months, leading to some of the most popular sites on the internet being driven by AI. When you search a topic or keyword in Google now, the very first results that appear at the top of the results page are AI-generated answers, which strive to intake information from the top several actual results and distill them into a couple of sentences for the user’s sake. This, however, has not proven universally successful.
If the top results that Google’s AI is using as references offer contradictory data to one another, it can be detrimental to the AI’s answers. However, unless users go through the process of reading the references themselves (something that the vast majority of general audiences are not going to do), they would have no way of knowing that the answers AI provides are incorrect. This is emblematic of a larger issue with AI-generated content: it’s essentially little more than a vaguely repurposed, regurgitated mix of the information that is put into it.
AI Images vs. Reality
Take Portrait Pal, for example, a platform that offers AI-generated headshots. Users feed pictures of themselves and professional headshot templates into the AI, and it then generates an approximation of what a professional headshot of said user would look like. Like all AI uses, this has gained traction in tech spheres for its innovative technology implementation, as it does offer an initially cheaper alternative to booking a professional photoshoot.
Despite the technology, some generated platforms that create headshots resemble little to the user because the generator amalgamates several disparate images of this person into a cohesive whole, which may not be representative of reality.
This is incredibly similar to how AI-generated writing works, often retaining the hallmarks of the writing being input into the generator but sanding off its edges and distilling it into its most broadly palatable form. Because of the ways in which AI generators function, taking elements of other human-made works and digitally blending them together to create ‘new’ work, it has become increasingly common for human-made works to be identified as false positives by AI detection.
How Do False Positives Happen?
If one were to exclusively put the works of Stephen King into an AI generator and task that generator to create new content, it would generate work bearing many of the hallmarks of King’s writing. If that was done for a month, creating hundreds of new AI-generated writing pieces per day, all built from the foundation of King’s actual writing, that might be enough content to get the AI King work noticed by AI detection algorithms. In doing so, the AI detection would internalize the aspects that were distinctly true to King’s writing style and mark those as indicators of AI-written content. As such, if one were to then check the validity of an actual authentic King work following this experiment, there’s a very high chance it would be flagged as AI-generated writing.
Therein lies the problem: AI regurgitates digitally obfuscated actual writing, so if an element of writing is common enough to appear in many of these AI-penned examples, AI detection services will associate that element with AI in future cases. The Stephen King example is an extreme one but valid. Many false positives stem from the human writer using oversimplified or repetitive language. While that may make works more broadly accessible to general audiences, it is also the exact tactic that AI generators employ, which is likely to be flagged in AI detection.
The World of AI: How to Combat False Positives
The best way to combat false positives in your own writing is to avoid these pitfalls. By keeping your writing distinct and idiosyncratic, it is less likely to be flagged as AI-generated. The question of authentic originality is one that has become increasingly hard to define in the internet age, and now with the advent and widespread popularization of AI, its even more difficult. In essence, the best way to avoid being flagged by AI detection is to create work that is unique and distinctively human.