Covering Disruptive Technology Powering Business in The Digital Age

Written by: Izzat Najmi Abdullah, Journalist, AOPG.

The impact of AI on content creation has been particularly evident in the scientific community, where the pressure to publish has led some researchers to turn to AI tools like ChatGPT to expedite the writing process. However, the use of AI in scientific writing has not been without its significant drawbacks.

For starters, AI has its tendency to generate inaccurate content—a phenomenon known as AI Hallucination—which in most cases can cause quite the embarrassment if caught, and not to forget confusion within the academic world, all while you are on your journey trying to publish reputable journals.

One particularly striking example is an AI-generated infographic of a rat with an unusually large “anatomical feature” (apologies if this is a bit unsettling), which was published in a well-known academic journal before being retracted. Similarly, another study included a horror graphic of human legs with multiple jointed bones that resembled hands—another glaring AI error that was published and quickly retracted. These incidents highlight the dangers of relying on AI-generated content without thorough verification.

The illustrated photo taken from the now retracted study – photo credit: The Telegraph

Andrew Gray, a librarian at University College London, conducted an analysis of papers on AI influence and found that in 2023 alone, over 60,000 papers involved the use of AI—accounting for more than one per cent of the annual total. The sheer volume of AI-generated content in academic research is alarming, and as AI tools continue to advance, this problem is expected to grow.

The Rise of “Slop” Content

The term “slop” has emerged to describe the vast amounts of low-quality, unreliable content generated by AI. This content, often produced with little regard for accuracy, includes everything from bogus articles to dubious social media posts and poorly written e-books. The phenomenon of slop highlights not only the ease with which AI can produce content but also underscores the potential dangers of flooding the Internet with subpar material.

The New York Times recently popularised the term, drawing attention to the growing prevalence of AI-generated content that is both easy and inexpensive to produce. An American journalist, for example, was able to create a fully automated political disinformation site, capable of churning out dozens of fake news stories each day. Such sites could have a profound impact on public opinion, particularly in the lead-up to elections.

The concept of slop is distinct from “botshit,” another term used to describe false or misleading content generated by AI tools, or bots. While botshit is often created with the intent to deceive, slop is more about the sheer volume of low-quality content that is flooding the Internet, making it increasingly difficult for users to discern what is true and what is not.

The Ouroboros Effect

One of the most concerning aspects of the AI content crisis is the potential for AI systems to train on their own generated content, creating a vicious cycle of misinformation. Researchers have warned that as more AI-generated material is uploaded to the Internet, there is a risk that future AI systems will increasingly rely on this content for training, leading to a degradation in the quality of AI outputs over time.

This phenomenon, sometimes referred to as the “dead Internet theory,” suggests that an increasing portion of the Internet is becoming automated, leading to a recursive loop in which AI content begets more AI content. A study by researchers at the University of Toronto found that when AI systems were trained on AI-generated text, the quality of the output deteriorated rapidly, resembling a repetitive list of nonsensical information. This process is akin to making a photocopy of a photocopy, where the quality degrades with each iteration.

The implications of this recursive loop are profound and as AI continues to produce and consume its own content, the Internet could become inundated with increasingly low-quality and misleading information, making it more challenging for users to find reliable sources.

AI Content and the Danger of Blind Trust

As AI-generated content becomes more prevalent, the need for advanced detection tools and increased vigilance is becoming vital, and I hope that it can come as soon as possible.

The potential for AI to be used in manipulative and deceptive ways is growing, and without proper safeguards, the Internet could become a breeding ground for misinformation, which I hope, does not lead to any fatality. I mean this in the best way, especially after some news circling to Google’s very own AI Overview (not the widely used Gemini) that suggested users use glue in order to make the cheese in your pizza stick and even generated a response stating that geologists recommend humans to eat one rock per day which is pretty scary considering many people these days don’t do thorough research might blindly follow such advice if it seems authoritative.

So, for our readers, I have nothing else to advise but this: Do not trust 100% of whatever you’re reading or seeing online. And please, for the love of God, do some research before you actually decide to believe or even share anything you find online.

(0)(0)