Driven by a natural curiosity of what the competitors in my local market are building, I recently found myself stepping through the website of one of the top Google results for, "web agencies in my area". They had a blog, with a plethora of entries to the tune of, "How to do X in Y" (Y, being a local town). Each of these posts repeated the same, "X in Y", so many times that by paragraph 2 of post 1, I was questioning whether the post had been created with the sole intent of optimising for these keywords. By paragraph 4, I was certain of it. Needless to say, having my time wasted reading, most likely, AI generated drivel in a vapid attempt to increase site ranking, left me with a bitter impression of the company. They succeeded at getting their agency near the top of Google rankings, but what use is that if you squander your one and only chance at making a good first impression on a potential customer?

The Rise of (Generative) AI

Bots attempting to imitate humans predates the internet (see ELIZA, 1967). The sophistication of these bots has grown exponentially over time, exploding with the invent of the generative pre-trained transformer (GPT e.g. ChatGPT). With cheap, easy access to Natural Language Processing models, it has never been easier for players to produce coherent wish-wash with a quick tapping of their eager fingers. It is even cheaper and easier for those fingers to extend to an unsuspecting person's ear.

User (me): Give me a two line SEO optimised blog post for doing business in Narnia

Bot (GPT-4o): “Unlock Business Success in Narnia: A Magical Guide
Discover how to navigate Narnia’s enchanted economy, from trade opportunities in Cair Paravel to partnerships with mythical creatures—your gateway to thriving in a land of wonder!”

Perhaps this is a poor example, the model has more accurate inbaked knowledge about Narnia's economy than the current economic state of my local town, but you get my drift.

The Death of the Internet

The Dead Internet theory is a conspiracy theory about the rampant increase in bot driven content since approximately 2016/2017. By some accounts, the theory goes so far as to allege that governments propel the bots in order to manipulate public perception (e.g. Trump's 2016 election) but this is outside the scope of this post.

The crux of the theory is that at some point in recent history, online content generated by bots exceeded that generated by humans. If search engines reward key word optimisation above quality of content, players will always be incentivised to sacrifice quality in lieu of an arbitrary demand of an algorithm. There are many other incentives, including, but not limited to: mediating the cold start problem; simulated testimonials for an emerging product; influencer promotion. All inevitably lead to the manipulation of the unsuspecting consumer.

Less is More

There was a point of time when the majority of the accessible information available on the internet was valuable. Information was scarce and the price of entry was high, people weren't so willing to pollute their newly founded data utopia. Then came the commoditisation of data, social media gave everyone a voice, a voice for sale. The more a person spoke into the net, the more data collected on them by the internet behemoths. Voracious corporations waited (and continue to wait) on the sidelines to be fed a user deemed (algorithmically) most likely to buy their product. When there is minimal visible cost to expressing one's self, there is no reason not to.

People are rightly rewarded for their ability to provide noiseless, valuable, cohesively constructive information to the internet. But the converse isn't true, there's no penalty for people who spin up bots armed with social accounts, sent out to farm impressions. There is only a (limited) potential upside to these practices: a click is valuable, regardless of whether a user was misled to make that click. I don't remember when I started appending "reddit" to my Google searches in search of a genuine human opinion, but I know that this is a symptom of the larger problem. The problem of yesterday was a lack of data, today's is how to divide an abundance of data into the meaningful and the worthless.

Guidance

The advent of GPT led to almost immediate services available purporting to report whether text is AI generated or human written. They barely worked on release and they've only become more useless over time as models have become more advanced. I'm not arguing that anything created with the use of generative AI is worthless. On the contrary, I believe that a symbiotic relationship between a human's creativity and the perspective widening lens of generative AI assistance can greatly help a person refine their ideas and execution.

At each step in our technological evolution there has been an increasing temptation to slacken the reins and give way entirely to automated processes. But without human oversight, how can an AI agent embody our unique creative direction? Our creative spark has been formed as a result of all the experiences we have uniquely lived through. The AI's "creative" identity is defined by a model created from the aggregation of the entire (in reduced format) human experience. Without guidance, there is no soul. People subconsciously, or consciously, detect this and wisely avoid the content.

Call to Action

Be conservative with the use of generative AI tools for creative purposes, remain in the driving seat. Spurn the AI generated drivel as much as you applaud the soulful.