“As Frankfurt says, the thing that is most dangerous to a democratic society is not a liar, it’s a bullshitter,” added Sandra Wachter, a Professor of Technology and Regulation at the Oxford Internet Institute. Her encapsulation is the foreboding proliferation of AI-generated content which has become colloquially known today as “AI slop.” It’s stealthily entering the internet, altering algorithms, and the dynamic of society’s interaction with information.
Consider the mind-blowing figure that a video created by an AI of an alien-like being turning into a spider has been watched 362 million times larger than the aggregate viewership of several serious journalism outlets multiplied some number of times. Such viral videos, created in some cases within minutes, show how generative AI is brute-forcing internet algorithms, targeting platforms rather than people.
The term “AI slop” is a vivid picture, similar to heaps of rubbish food pushed into troughs. It’s full of dodgy writing, odd images, and misleading videos content that is not intended to teach but to fool algorithms to interact or SEO spam. The result is monumental, with Facebook alone propagating AI-made spam resulting in ad-filled click farms. As a single viral instance, the outlandish “Shrimp Jesus” photos, mashed up halfway between crabs and religious figures, are the best example of how AI trash can fool tourists while generating traffic to fishy destinations.
AI slop in its simplest definition is just a computer brute force attack. As hackers utilize relentless guessing with passwords, AI slop creators flood the internet with content, exploiting recommendation and search program weaknesses. Not a sound strategy but great at generating feeds full of content that continues algorithmic interaction and not human relevance.
The effects extend beyond annoyance. AI slobber diminishes the veracity of web information and blurs the line of demarcation between fact and fiction. A dramatic case in point was witnessed after a catastrophic flood had devastated Valencia, Spain, when dramatic images of flooded automobiles were downplayed by cynics as artificial pictures produced by AI synthetically. This cynicism is evidence of a culture wide illness growing more common all the time a failure and even refusal to separate actual events from those created by individuals.
Further complexity arises due to the recursive training of AIs. Large language models (LLMs), which are mimicking human speech, are in turn being increasingly trained on material produced by AIs. The cycle can propagate errors and give rise to a dirty information system. Wachter likens it to environmental destruction, cautioning, “Everybody’s just throwing their empty cans into the forest. So it’s going to be much harder to have a nice walk out there because it’s just being polluted, and because those systems can pollute so much quicker than humans could.”
The destiny of the media is particularly dire. AI puff content is pouring into news deserts, regions with meager local coverage, where lower stakes AI based pieces masquerade as legitimate news. More than a thousand of such sites were discovered by monitoring group NewsGuard to be operating with minimal human editorial control. Such sites are programmed to opt for applying SEO tactics in hopes of reaping the most ad revenue, belittling the worth of honest journalism.
While the difficulties are real, there are also optimists. David Caswell is among the early visionaries concerning news workflow to tap into with AI, and he compares AI slop to spam email a challenge that platforms eventually solved. The solution is differentiation, he argues “What’s going to make a difference is being relevant to audiences.” AI can be employed responsibly in order to increase efficiency without sacrificing editorial values.
But the social effect cannot be ignored. As slop AI is disseminated, it can potentially lead to normalizing information and injecting error into public knowledge bases like Wikipedia. Wachter and her authors offer legal frameworks for making accountability possible to LLM developers and calls on models to be built with integrity at the top of the agenda.
AI slop production is also of moral significance to the environmental aspect of generative AI. The energy required to calculate such content is so massive that it must be questioned whether or not it is even worth the cost of efficiency to produce such an inefficient product. Last, war against AI slop is war against the integrity of virtual space. While Google and Meta tamper with mechanisms to detect AI spam, they are lagging behind. As Wachter advises, users need to subject AI-produced results to criticism and not take such systems as oracles of truth.
The internet is at a turning point. Will AI slop peter out of usefulness, like spam, into the backwaters of our cyberspace? Or will it simply continue to decay the information system, potentially killing our concept of reality? The decision is in the joint efforts of developers, platforms, and users toward keeping the standards of accuracy, relevance, and trust.
