Some time ago, in this post, I expressed my concern that AI, especially LLM-driven AI, would lead us into an age of rampant misinformation. My fear was rooted in the way LLMs like ChatGPT operate. These models function as advanced predictive text systems. They generate responses by processing vast amounts of text scraped from the internet, recognizing patterns in human speech, and then predicting what should come next based on the given context. Essentially, they don’t “think”—they just predict.

So, my concern was that since these systems aren’t perfect, inaccuracies would inevitably slip in. And now, with LLMs being used in almost every industry, from medical journals to tech magazines, mainstream news media, and even books, I feared we were on the verge of an unprecedented explosion of misinformation.

A friend of mine, who ghostwrites medical articles for doctors, uses LLMs in her work, and I realized that because she isn’t an expert, she can’t identify inaccuracies. The danger, as I saw it, was that AI-generated content would keep being used to train future AI models, creating a feedback loop of errors. I worried that this could lead to a “singularity of misinformation”, where the entire internet becomes riddled with inaccuracies, rendering it useless.

But here’s the surprise: I no longer hold that concern.

You might be wondering how such a valid concern suddenly dissolved. The logic seemed so sound, right? But two key realizations made me rethink the whole issue.

The Internet Was Never Accurate to Begin With

The first thing that hit me—and I feel a bit foolish for not realizing it sooner—is that my concern assumed the internet, as written by humans, was accurate. That assumption is laughable. I knew this, but I didn’t think of it at the time.

The reality is that most human-generated content, especially when it matters, is tainted by biases and agendas. Whether it’s information about climate change, health guidelines, or the pandemic, vested interests shape what we’re told. Corporations fund studies to promote their products. Governments push narratives that help them exert control or raise taxes. In short, wherever money or power is at stake, the information is corrupted.

This has been the case since the dawn of civilization. The “Gell-Mann amnesia effect” explains how this works. If you’ve ever read something in the media about a topic you’re deeply familiar with, you’ll know how wrong they often get it. But then, you turn the page and trust everything else. The same happens when articles are written about people. Anyone who’s experienced some degree of fame will tell you that what’s written about them is often distorted, yet we trust media reports on everything else.

The truth is, human-generated content is already full of inaccuracies. Writers, even when covering scientific topics, rely on published literature that’s often tainted by misinformation. So, the process isn’t all that different from what’s happening with AI. If AI is trained on misinformation and reproduces it, that’s just how humans have been operating for centuries.

The Human Brain Isn’t So Different From AI

The second realization is this: I’ve argued before that I don’t see much difference between LLMs and the human brain. I don’t believe human intelligence is as special as we like to think. In creating AI, we’re essentially learning more about how our own consciousness operates.

The brain, like LLMs, relies on data (our experiences), complexity (our neural networks), and connections (our synapses). The more complex the system, the more intricate the thoughts. But fundamentally, the brain isn’t producing anything truly original—it’s all drawn from the data it has accumulated over time, much like an LLM.

So, if LLMs can’t generate original thoughts because they work from their training data, the same can be said for humans. We don’t generate original content either. Every idea or thought we have is influenced by our experiences, environment, and the information we’ve absorbed over our lifetimes. The notion that humans create “original” content is an illusion. In that sense, we’re no different from AI.

By tying these two observations together, my concern evaporates. The rise of AI won’t drastically change the accuracy of content—human content has always been filled with inaccuracies. The only thing that will change with AI is the speed at which content is produced. So, whether it’s humans or AI generating it, the degree of accuracy (or inaccuracy) will remain largely the same.