Get Mentally Prepared for (Lots of) AI Content

In the race between OpenAI, Google, Meta & Amazon to create the most advanced language models, they’ve opened a digital Pandora's box. And what's pouring out? Content. Endless, infinite, A.I.-generated content. Are we prepared for this?

Imagine a scenario where, in just weeks, A.I. models produce more content than all of human history combined. That’s the future we’re sprinting towards, if we’re not there already.

Gone are the days when we were overwhelmed by too many browser tabs filled with articles, videos, and memes. Today, every digital platform is an A.I.-powered content machine, producing and tailoring content faster than ever before. We thought we’d reached the peak of the information age, but A.I. is telling us that we ain’t seen nothing yet.

The Atlantic termed our situation a "Textpocalypse", pointing to our struggle to handle this content overload.

In essence, A.I.-generated content is becoming the new dark matter of the web.

Timothy Shoup of the Copenhagen Institute for Futures Studies suggests that by 2025 to 2030, models like GPT-3 and GPT-4 could be responsible for generating nearly 99.9% of online content.

The internet's evolution has seen various phases over the decades, from its foundational protocols like HTTP, TCP/IP, and HTML to the dynamic social platforms and commerce hubs we have today. Yet, as we move into 2023 and beyond, it's not AR, VR, or IoT that's taking center stage. It's the A.I. content layer – growing at a rate that's almost inconceivable.

This is all fed by a large ecosystem of startups and tools emerging, all aimed at empowering creators. These tools are refining and optimizing existing models, making it more efficient for creators to produce vast amounts of content using generative A.I.

‍Models aren't training on A.I.-generated content.

I recently had a conversation with Jonathan White, Chief of Staff at GPTZero, a startup at the intersection of AI and human authenticity. He believes that as our interactions with peers and businesses evolve, specialized language models could play a pivotal role in enhancing customer service and streamlining workflows.

But not without challenges. There's a growing concern about model collapse and hallucinations in LLMs. To address this, companies are currently focusing on training LLMs using human-written content from before 2022. As the AI landscape keeps evolving, Jonathan anticipates the emergence of specialized, domain-specific LLMs trained only on human-created text.

White highlighted responsible AI use, but cautioned against relying solely on top-down regulation. "I would not expect regulation to be anybody's savior in this space," he shared. 

As we move forward, he believes the solution starts with individual users understanding how to best interact with A.I.’s outputs.

Because soon, we'll have models tailored for every market, every niche, and every kind of training data. Companies are pouring vast resources into models like Google's Bard, Anthropic's Claude, and Meta's LlaMA. And every day the list grows, with countless other open-source models being developed globally.

This surge empowers novice writers and creators to produce and publish boundless content. We've seen a hint of this with social media, but now these tools are potentially the authors.

Jeremy Shere, founder and CEO of Connversa, a digital content agency, comments, "Using AI to generate an outline may be convenient, but it allows you to skip a crucial step in learning to write well, namely thinking through how best to structure what you're writing to most effectively communicate with your audience. When you outsource that task to AI, you essentially cede control of the entire thrust of the piece."

On the one hand, emerging technology always brings change. We've seen the internet evolve with bots, photoshop, clickbait headlines, autoplay videos, and eerily accurate ads. Now, A.I.-generated content seems to be the next phase.

But the scale of this change is unprecedented. We're looking at an influx of content – countless articles, stories, and ideas. It's repetition, but on a scale we've never seen before.

As we've seen with the demand for organic labels and GMO-free certifications in the food industry, I anticipate a similar trend in content. 

Just as consumers want to know what's in their food, they'll want assurance that the articles they read have been vetted by humans for accuracy and context.

This counter-movement is already underway with tools like GPTZero, designed to detect and label A.I.-generated content. In the same way we seek out "clean eating," might we soon demand "clean" content that's labeled "A.I.-free"?

Consumers will want assurance that the content they're reading has undergone human quality checks to eliminate false information and provide accurate context. In response, tools like GPTZero have been developed to detect and label this content, ensuring users are aware of its origins.

Lauren Lang, formerly an English 101 university instructor and currently a content marketer for AI startups, expressed her reservations about the increasing dependence on A.I. for writing.

"Let's not forget the 'artificial' in 'artificial intelligence'," Lauren remarked, emphasizing that A.I. might fall short in grasping the subtle intricacies of human language. She added, “Promising students they can skip the 'shitty first drafts' is worrisome. Those drafts are essential for their growth.”

At an Art Festival in Charlotte, North Carolina, where I live, I commissioned a charcoal drawing of our family cat, Ozzie, as a Christmas gift for my dad.

Curious, I also used Midjourney to digitally transform Ozzie's photo into a Pixar-style character.

While the digital version was cute, it was created in mere seconds and lacked the personal touch. In the end, the hand-drawn portrait, filled with genuine effort and emotion, was the standout choice for the gift. Safe to say, Ozzie approved.

We process information to shape our actions. But what happens when the information we consume is a never-ending loop of A.I.-generated content? It's a concerning thought. This could just be noise, or worse, it could mislead and stunt our growth.

Even in something as simple as a family cat portrait, we show a preference for the hand-drawn, human-created piece. It's a testament to our innate appreciation for genuine human effort and creativity, even with all the utility A.I. offers.

Yet, A.I. content might give us a new priority: cherish human insights over volume.