The Hunt Blog

In May of 2023, seemingly ages ago in AI time, I wrote an article for my friend Alex Horovitz’s blog entitled ‘Low-Background Content’. Within, I speculated about a distant future where much of the content to which we were exposed would be created by generative AI systems. I also imagined that AI purveyors would start ‘eating their own dogfood’ as it became harder to find human-generated source content, leading to a self-referencing AI death spiral that at first stifles innovation and skews human voices, but eventually causes these AI systems to fail. 

I also suggested that ‘Low-Background’ content, content created by real human prior to AI’s arrival, would become increasingly valuable. I compared it to the salvaging of sunken battleships after WWII because steel on those vessels was created before radioactive isotopes triggered by the nuclear age made all steel made afterwards slightly radioactive.  

I now have an update on my musings.

Update on the AI Tsunami – It has come ashore

In a short 15 months since my article, my darkest concerns about the speed of an AI-generated content tsunami seem naive. I recently read this article from Business Insider India entitled “The internet has more AI-generated content today than human-created” which details just how quickly AI content has taken over. Here’s the payoff paragraph:

“A study by Amazon Web Services, as reported by a recent Forbes report, says that 57 percent content on the internet is generated by AI. The Forbes report also quoted an expert who said that 90 percent of all internet content is likely to be AI-generated by 2025.”

We are now officially inundated, engulfed and overwhelmed by the AI content flood. And the prediction that 90 percent of the internet will be AI-generated by next year? Take a moment to think about what this means. 

As you navigate the internet in all its guises (websites, cloud services, apps, etc.) almost none of what you find will be written by a human being who has experienced the non-digital ‘real world’ outside of computers. The Emerald has a brilliant podcast entitled “So you want to be a sorcerer in the age of mythic powers” that explores the implications of this. And worse, as Alex points out in his own recent article exploring AI (he references Wolfram’s excellent explanation of LLMs), there’s no indication that AI is anything more than a statistical parlor trick. This means that AI can’t make those leaps of logic that lead to intellectual, scientific or artistic breakthroughs that have moved the human race forward in the past. This is very sobering. But wait! There’s more!

The AI Death Spiral now has an Official Name:  Model Collapse

As to my predictions about ‘Low-Background Content’, I recommend this Forbes article, “Is AI quietly killing itself – and the Internet?”. They talk about the AI Self-Referential Death Spiral that I mentioned and give it a technical name: ‘Model Collapse’. Here is the payoff paragraph in this article:

“For some reason—and the researchers don’t fully know why—when AI feeds only on a steady diet of its own synthetic data, it loses touch with original threads of reality and tends to create its own best answer based on its own best recycled data points.”

If almost all of the content in the world is generated by AI, and AI models must constantly be trained on content, what happens? Where do the AI companies get the content to train their models? And how do they keep AI-generated content out? I suspect that the answer is this:  given the depths and pervasiveness of the AI content, it will be increasingly difficult to police the authenticity of content. Some AI generated content will slip through, polluting future models. I suspect we will see weird things happening with LLMs as this plays out, in the same way that inbreeding led to some very weird Habsburg aristocrats.

The Scramble for ‘Low-Background Content”

What about my prediction that there would be a scramble for certifiably human-created content? There appears to be an explosion of lawsuits by content owners against AI companies who have been using their content without paying. AI companies are scraping content until the cows come home. They are also making a legal claim that they have a right to use any content they find without compensation. To quote the article, ‘AI and the copyright liability overhang’ by McIntosh, Suzuki and Schall:

“[AI Companies] assert that after the models are trained, they do not contain any copies of the ‘scraped’ content, or any copy of the material that was used to train the model. However, the defendants have generally not yet provided a theory as to why copies made in the process of training the AI models or creating the training databases are not infringement.”  

Get your popcorn ready. It’s going to be a dramatic battle. But I suspect that AI companies will get their way. Those rights holders that can shield their content from being ‘slurped’ will no doubt be able to cut a better deal in the future as AI companies look ravenously for source material. But will anybody be creating new, human-generated content? Even if they do, it will be hard for them to break through the tidal wave of generated stuff.

Surfing the Tsunami

So, dear reader and job hunter, what might all of this mean for us? Well, soon, almost everything we’re exposed to will be AI-authored or AI influenced. This will have one obvious, painful implication for job seekers. Hiring entities will be extremely skeptical of what you wrote on your resume. If they read it at all (75% of resumes are now evaluated and rejected without human intervention). Here is how I’m holding all of this.

First, I don’t feel that there is any way to stop the tsunami. So I’m learning how to surf. I’ve jumped into AI with my company, Callings.ai. Our mission: level the playing field for job seekers with a comprehensive, AI-powered platform that augments their very human efforts. I feel much less afraid of AI the more I work with it. As my partner, Philippe Cailloux reminds me, “AI is just a tool”.

Second, I am doubling down on real human interactions. Nobody wants to have lunch with an LLM. It can’t smile, smirk, or in any way share a real human moment. It may drone on about the finer points of racing in the Tour De France but it can’t actually ride a bike. I am optimistic that one outcome of the AI tsunami is that people will start to gravitate again to other people.

Finally, I’m investing—both at Callings.ai and for myself—in the concept of a personal brand. While ChatGPT can bury you with content, it can’t replicate that unique combination of skills, experiences, personality—the humanity— that we embody. It is our authenticity that will stand out in a deluge of artificiality, as this article by Michael Todasco suggests.

So embrace the inevitability of the AI revolution with authenticity and humanity. Perhaps the tsunami will also usher in a parallel era of increased humanity as well.

CODA: My eldest, starting college this fall, called out of the blue this morning. Top of mind for him is the impact of AI on his future as a Liberal Arts student. He’s been working on the following poem to express his concerns, using ChatGPT to prompt him. Perhaps a lesson in using the tool while remaining authentic? He let me share it with you all.

Write Me a Poem About AI Taking Over The Arts
By Thomas Rice

Is there something I can say now that hasn’t already been said?
In a million databases in dark rooms, the clicking of
Digital keyboards in perfected movements, unpracticed.

Can soul be replicated?
I feel mine can be. And It* says

“Yet in the art of crafted line,
Where human touch and heart entwine,
Can circuits grasp the soul’s deep yearn,
Or feel the flame for which we burn?”

It (did not) think about what I would say
It (did not) feel my fear and rage
It (did not) play with lines of rhyme
It (did not) forget to claim it’s mine

And it won’t ever, but does it matter?
Can you tell which one is better?
The shaking hand of frustration, bitter.
I needed its help, its thin, dead letters.

I can’t rhyme like it can, I know,
the words don’t come, there’s nothing to show.
And maybe it’s because I am not a poet
But I feel it should not be, and I know that you know it.

Unsettling, seeing my emotions in line, that flame,
That claims of flesh, blood, and spine
Of arts to be stolen (it’s us that owns it)
replication of thought, feeling, soul,
Naught to show for it
and I asked if the essence could be replicated, refit
And it Said, we cannot
And I don’t believe it.

*This portion was written by ChatGPT