AI on a trip: Artificially intelligent garbage takes over the internet

Today, one cannot do without artificial intelligence. Many self-proclaimed experts assure us of this, popping up from every corner over the past year to push artificial intelligence where it is and isn’t needed. Although just a few years ago, AI was discussed only timidly, and experts in the field could be counted on one hand, with the emergence of ChatGPT, everyone suddenly became an expert, and artificial intelligence is being integrated into every aspect of our lives. Everyone praised OpenAI’s ChatGPT, claiming it would replace many people and jobs, promising more free time and who knows what else. But it soon became clear that ChatGPT, like all other models, such as Gemini, Claude, and Le Chat, suffers from what experts have termed – hallucinations.

Namely, no matter how many swear by AI, it is still far from true artificial intelligence, and when it comes to hallucinations, sometimes you can really think that some of these AI models are actually on acid or, as they say on the street, tripping. What’s worse, in 99 percent of cases, these are woke-tripped language models, making the situation even more comical. But language models are not the only problem. All other AI outputs, from AI music, photography to video, suffer from hallucinations and often produce total nonsense. For instance, most AI models that convert words into images still do not know how many fingers a person has on one hand, so we often see perfectly generated AI women (who earn on social media) with extra fingers, strange legs, and so on.

Spam, spam, spam

AI garbage is already abundant in the vastness of the internet, and as things stand, garbage generated by artificial intelligence will only increase. This AI garbage has recently received its name – slop. Slop is actually a distant relative of the well-known spam. Spam, as we know, is unwanted mail that is mostly sent in large quantities for the purpose of advertising or defrauding the recipient. The emergence of spam can be traced back to the beginning of electronic mail, and it became increasingly present in the 90s. Interestingly, the first documented case of spam occurred on May 3, 1978, when Gary Thuerk, a marketing manager for Digital Equipment Corporation (DEC), sent a mass email to Arpanet users (the precursor to the internet). He sent a message to 393 addresses promoting a new series of DEC computers. This event became known as the first case of spam via electronic mail. Another interesting fact about spam is that the linguistic meaning of the word spam is attributed to the British comedy troupe Monty Python in the now legendary sketch from their TV series ‘Flying Circus’ in which a group of Vikings asks for spam, spam, spam.

Shoveling it in

And after spam, here comes slop. The term slop became popular a few months ago when Google incorporated its Gemini AI into its search engine, which instead of directing users to links offers its AI solutions and answers at the top of the search results as ‘AI overview’. This Google change was a reaction to Microsoft also embedding AI into its search results on Bing. The problem, however, is that most of this AI content is either inaccurate or misleading, and it exists in vast quantities. Hence the name slop, which denotes ‘a pile of tasteless food that is shoveled into troughs for livestock’. An interesting analogy, but quite accurate. Thus, our holy AI quickly shovels its misleading and not necessarily accurate information in large quantities before us.

At least it’s not interactive

Slop, however, is not interactive unlike various chatbots and language models, and its purpose is rarely to answer user questions. Instead, it mainly functions to create the impression of content created by humans, exploit advertising revenue, and direct search engine attention to other pages. Just like unwanted mail, almost no one wants to look at this garbage, but the internet economy still creates it. And just like unwanted mail, its overall effect is negative, wasting users’ time and effort as they now have to sift through the dirt to find the content they are actually looking for.

Kristian Hammond, director of the Center for the Advancement of Machine Intelligence at Northwestern University, noted that the problem with the current AI overview model is that it presents itself as a final answer to a query, rather than as a starting point for users to explore a topic. Hammond says that ‘searching the internet should prompt you to think, and AI results do not encourage you to think but rather to accept those answers as facts, which is quite dangerous’.

Zombie internet

And there are already plenty of examples of slop. For instance, an article generated by Microsoft Travel states that ‘The Food Bank in Ottawa’ is an attraction that every tourist must see when visiting the capital of Canada. Additionally, AI-generated books are becoming an increasing problem. One example is a book about mushrooms that appeared on Amazon. Shortly after its publication, amateur foragers warned that the book was likely written by AI, as it is full of dangerous advice for those hoping to distinguish poisonous from edible mushrooms. However, the ‘author’ of that book apparently did not care whether someone foraging for mushrooms would eat their last one because of the book.

On Facebook, AI-generated images have also flourished, with images of Jesus Christ with shrimp instead of hands, children in plastic toy cars, fake dream homes, and incredible old ladies claiming to have baked a cake for their 122nd birthday garnering thousands of shares and likes.

Jason Koebler from the tech news site 404 Media believes that this trend represents what he calls the ‘zombie internet’. The rise of slop, he says, has turned that social network into a space where ‘bots, people, and accounts that were once people mix to create a catastrophic social network with very few social connections’.

So even though Facebook has recently been inundated with bots, AI images, and videos, Nick Clegg, head of global affairs at Facebook’s parent company Meta, wrote in February that the social network is training its systems to recognize content made by artificial intelligence.

– As the line between human and synthetic content blurs, people want to know where the boundary is – he wrote.

Serving nonsense

However, as we see, Facebook still has significant problems recognizing what is human and what is not, and the issue has begun to concern the main source of revenue for the social media industry: advertising agencies, which pay to place ads alongside content. Farhad Divecha, director of the digital marketing agency AccuraCast based in the UK, says he is now encountering cases where users mistakenly label ads as slop created by artificial intelligence when they are not.

– We have seen cases where people commented that the ad is nonsense created by artificial intelligence when it was not – he emphasizes, adding that this could become a problem for the social media industry if consumers start to feel that they are being served nonsense all the time.

And that will inevitably happen because the increasing power of artificial intelligence means that its ability to generate new text, images, videos, and other types of content will become better, faster, and larger, which means that such material will flood the internet, so we will not be able to escape slop, but we will have to learn to recognize it. So we can conclude this text with a common acronym in the world of cryptocurrencies, which is DYOR or do your own research, which is becoming mandatory for everything you see and read on the internet, because as we see, slop or garbage is taking over our world wide web.