{"id":585,"date":"2025-12-23T03:56:10","date_gmt":"2025-12-23T03:56:10","guid":{"rendered":"https:\/\/donavan.sg\/blog\/?p=585"},"modified":"2025-12-23T03:56:13","modified_gmt":"2025-12-23T03:56:13","slug":"why-is-ai-slop-everywhere-and-what-can-we-do-about-it","status":"publish","type":"post","link":"https:\/\/donavan.sg\/blog\/index.php\/2025\/12\/23\/why-is-ai-slop-everywhere-and-what-can-we-do-about-it\/","title":{"rendered":"Why is AI Slop Everywhere? And What Can We Do About It?"},"content":{"rendered":"\n

A few months ago, I came across a Channel NewsAsia<\/a> article that provided a treatise (in the eyes of Gen Z\/Alpha, not necessarily Gen X\/Y) of AI slop and why it is so pervasive.<\/p>\n\n\n\n

AI slop is so widespread that most online platforms now suffer from a low signal to noise ratio. For someone that relies on cognitive abilities to remain gainfully employed, this is concerning to both myself as a person and the next generation.<\/p>\n\n\n\n

The Curse of Simple Ideas<\/h2>\n\n\n\n

At the heart of the pervasiveness of AI slop is the near zero marginal cost (to the user) to create graphics or videos that are approximately<\/em> accurate. In the case of impressionistic posts, approximate <\/em>accuracy is sufficient. People simply scroll through posts, pick up some bits of trivia and share them with their friends at conversations. Some of these would include AI slop that is certainly sensational (clickbait drives views), contains some kernels of truth and highly relatable due to the simplicity of the idea. Due to the much higher click-rate, algorithms then tend to prioritize such posts, leading to their virality.<\/p>\n\n\n\n

Now we have an issue. We have approximately <\/em>accurate content (in the best case scenario) that pervades the whole of the Internet quickly due to the mechanisms of said algorithms. And because such methods of content creation require near zero marginal cost, content creators also advertise the ease of creating such videos, attracting an audience and earning “easy money”.<\/p>\n\n\n\n

Accuracy Does Not Really Matter, Does It?<\/h2>\n\n\n\n

But approximate accuracy<\/em> was not even the worst-case assumption. In many of the popular AI slop videos, fake news and fake recipes have mushroomed, leading to real harm.<\/p>\n\n\n

\n
\"\"
Meme highlighting the “fake news” and “fake recipe” trend. Likely generated with AI, original source unknown.<\/figcaption><\/figure>\n<\/div>\n\n\n

Some adults may be able to distinguish between outright falsehood, and content that is likely credible. But children that are learning new knowledge may not be able to tell. And not all adults would be able to distinguish inaccuracies simply because we lack the means to know everything.<\/p>\n\n\n\n

But still, AI slop pervades.<\/p>\n\n\n\n

The Fallacious Chain of “Authority”<\/h2>\n\n\n\n

Perhaps AI slop would not have been such a big issue if it was just seen as spam<\/em>. But one insidious aspect of AI is how people look towards AI as a knowledge authority<\/em> (the next time someone presents an idea, ask them for the provenance and thought process). The lazy AI user would prompt a LLM, may follow some prompt engineering guide, and claim the output is authoritative. But how would we know if said output is correct? The lazy AI user might say, “AI knows everything; it’s most likely correct.” But it is also true that poor prompting and poor problem framing can also lead to serious inaccuracies. Just because an entity is an expert does not mean it has to provide expert opinion reliably. In fact, even experts find the communication of complex ideas difficult. Would an AI necessarily fare better?<\/p>\n\n\n\n

The Curse of Complex Ideas<\/h2>\n\n\n\n

Let us begin this section with a simple experiment with Gemini 3 Flash<\/a>.<\/p>\n\n\n\n

Example 1: Explaining Undergraduate Material: Maxwell’s Equations<\/h3>\n\n\n\n
\"\"<\/figure>\n\n\n\n

Here we have a prompt requesting Gemini to explain Maxwell’s four equations, which govern classical electrodynamics<\/a>. The first law is the simplest, with an approximately accurate<\/em> analogy to explain it. But by the time we come to the last two laws, which are more complex as they involve the interactions between electric and magnetic fields, we get analogies as such:<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

To a Physics-trained person like me, I can pick up the merits and demerits of said analogy. In fact, I reckon such analogies likely leave the primary school student more confused. (My own opinion: Better would be a suggestion to build an experimental set-up to move a magnet in and out of a loop of wire connected to a bulb to explain the point, instead of relying on such a clumsy analogy for Faraday’s Law.)<\/p>\n\n\n\n

But a child left unsupervised might read this fully and may even generate some AI slop trying to over-simplify Maxwell’s equations with confusing analogies.<\/p>\n\n\n\n

Example 2: Explaining Operations in Aviation<\/h3>\n\n\n\n

Here, we try to get Gemini 3 Flash to explain the “hub and spoke” model of SQ and how they structure connection banks for flights through Singapore.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

At this point, the general idea is correct. It is indeed true SQ operates a heavy arrival bank from Europe that lands in the morning, connecting to many Asian destinations with early-morning departures.<\/p>\n\n\n\n

Other examples where Gemini are accurate include SQ’s “dual-brand” strategy (technically there are inaccuracies with Gemini’s output, but we’ll overlook that for now.)<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

But the tactical examples are both incorrect, and also in a format that is not easy to understand how this links to the “flight bank concept”.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Perhaps a simpler way to have explained this would be to obtain an arrival bank information, with the corresponding departure bank information. One way to do this is to extract data from FR24. Let us review Example B: India to US.<\/p>\n\n\n

\n
\"\"<\/figure>\n<\/div>\n\n\n

Here we identify key data points. We have a morning arrival bank from Changi Airport and highlight inbound flights operated by SQ\/Scoot from locations such as Bengaluru, Chennai, Mumbai and Ahmedabad.<\/p>\n\n\n\n

Next, we look at the departures from Singapore.<\/p>\n\n\n

\n
\"\"<\/figure>\n<\/div>\n\n\n

Here, we support two points: SQ operates connections for India-Singapore-US flights, judging by how connection times from India – Singapore to Singapore – US flights (trivia: SQ12 is a SIN-NRT-LAX route, so that’s also included in the pool of US flights) are timed for connection times of between 1.5 hours to 3 hours. Next, we also highlight the “massive departure” banks all over Southeast Asia from this screenshot.<\/p>\n\n\n\n

Thoughts<\/h3>\n\n\n\n

To some extent, these examples highlight how AI can be a great tool to start scaffolding thoughts. The structural frameworks provided help us get from “zero to one” on a topic. But as you can see, I had to use some levels of domain expertise to pull the right sources to exemplify my points. But for me, AI use here is a win-more<\/strong>. I already know the topic, so AI simply speeds up the rate at which I can put together complex ideas. But for those who start from little (or insufficient domain knowledge), AI use leads to very impressionistic understanding. Also, due to the verbosity of the outputs (you can try by recreating these with the exact prompts used), fact-checking becomes an arduous affair (what do I fact-check?). The more instinctive nature is to simply accept the impressionistic output for what it is.<\/p>\n\n\n\n

What Can We Do About It?<\/h2>\n\n\n\n

This is a difficult topic for me personally. Having dealt with multiple disciplines involving complexity of ideas and principles, I feel aggrieved at how AI slop could result in people having even shorter attention spans and hence even less patience to appreciate complexity.<\/p>\n\n\n\n

But appreciation of complexity is a positive feedback loop. The more we spend time to think about different types of topics and piece them together in our heads, the more appreciation and wonder we have about how said ideas can come together to form new insight.<\/p>\n\n\n\n

Writers such as David Lewis have opined on how we, on the other hand, have been inundated with information<\/a> without the corresponding down-time to synthesize thoughts together. In other words, “less is more”.<\/p>\n\n\n\n

I personally find this true of AI slop. Consume less media in general. Think more. If possible, write, and exercise the muscle of thought organization. You don’t even need to publish the writings themselves, but treat the exercise in writing as a means of thinking. With practice, you will quickly find what is “signal” as opposed to “noise” to you, and in so doing, steer towards higher “signal”.<\/p>\n\n\n\n

As for the children, digital literacy is important. And while there are arguments for early exposure to digital devices when younger to be able to adapt to the digital world, said exposure must come with supervision, but more importantly, introspection. Ask questions that not only focus on the “what” of content consumption, but “why” and “how”. Ask questions on what they may find missing or hollow in some types of AI slop. Or perhaps, highlight the exaggerations, and go through the thought process of why your child finds certain content interesting or boring. It is a pity that one paragraph simply cannot do justice to parenting children in the digital world.<\/p>\n\n\n\n

Till next time. This topic is far from over.<\/p>\n","protected":false},"excerpt":{"rendered":"

A few months ago, I came across a Channel NewsAsia article that provided a treatise (in the eyes of Gen Z\/Alpha, not necessarily Gen X\/Y) of AI slop and why it is so pervasive. AI slop is so widespread that most online platforms now suffer from a low signal to noise ratio. For someone that…<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"om_disable_all_campaigns":false,"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"footnotes":""},"categories":[1],"tags":[],"class_list":["post-585","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/donavan.sg\/blog\/index.php\/wp-json\/wp\/v2\/posts\/585","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/donavan.sg\/blog\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/donavan.sg\/blog\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/donavan.sg\/blog\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/donavan.sg\/blog\/index.php\/wp-json\/wp\/v2\/comments?post=585"}],"version-history":[{"count":1,"href":"https:\/\/donavan.sg\/blog\/index.php\/wp-json\/wp\/v2\/posts\/585\/revisions"}],"predecessor-version":[{"id":596,"href":"https:\/\/donavan.sg\/blog\/index.php\/wp-json\/wp\/v2\/posts\/585\/revisions\/596"}],"wp:attachment":[{"href":"https:\/\/donavan.sg\/blog\/index.php\/wp-json\/wp\/v2\/media?parent=585"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/donavan.sg\/blog\/index.php\/wp-json\/wp\/v2\/categories?post=585"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/donavan.sg\/blog\/index.php\/wp-json\/wp\/v2\/tags?post=585"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}