SYNOPSIS: AI-generated content, which began to appear in earnest (and at scale) just this year, is poised to grow into an all-out AI Flood, and to utterly dominate, our internet. There are consequences to this event, not all of which are immediately obvious. Key amongst those is the fact that many people, perhaps a majority of the population, will welcome this opiate. With the flood coming, robust defenses, and arks of preservation, must be built in order to safeguard the rich cultural and artistic heritage of humanity.
Have you experienced the latest wave of AI-art? Most assuredly, you have. Whether you recognized it as such is another question. Most people, c. December 2022, have had some experience with it.
You’ve had this experience by seeing either wild paintings of your friends highlighted in their social feeds or profile pictures, or by devouring the ever-growing volume of mind-blowing artwork, “virtual” photography, 3d renders and imaginary sculptures on Instagram. Or perhaps you’ve fallen into the spell and experimented with the #1 app on the entire App Store, Lensa, or the addictively playful Wonder, and tried the AI magic out for yourself. Oh, vanity!
Here are some back-of-the-envelope stats for you, so that you can start to visualise the scene that we see playing out from here atop the mountain with our phat telescopes.
It’s estimated that at present, after only 6 months of generative activity, fully 1/200th of the images across the entire internet (approx. 3 billion of 750 billion total) are thought to be AI-generated.
What the AI Flood Means
My prediction is that within 5 years, the internet will be overwhelmed, completely, by an AI Flood of robo-genned content: articles, fake news, art, fake photos, fake videos on YT, TikTok, snap, FB etc. And all this “content” will be hyper-optimized for two very specific goals: 1) to suck your precious time in the form of attention (screen hours), which is then sold to advertisers for precious billions, and 2) to suck the money right out of your bank accounts and into the corporate coffers for products and services and subscriptions that you had never heard of before but suddenly find impossible to live without.
Most people who have experience with the artform are thinking: “well, even though AI can make an amazing image in 60 seconds on only the merest hint of textual instruction, at most I’m posting 10 photos a day…” but that’s an individual artist or Instagrammer. And while that’s entertaining, that’s not what we’re focused on when we talk about the flood.
Rather, what we’re talking about is the wholesale automation, mass-customization, and robo-posting of hundreds of millions of articles, images, and videos across every platform imaginable. The first shot has already been fired across the bow, when ChatGPT almost annihilated the lofty heights of StackOverflow with incorrect, mass-produced answers to coding challenges (sidenote: yes, AIs can effectively both debug and author code)… and those, by and large, were posted one at a time with the manual assistance of human users.
Imagine, instead of you playing around with the Wonder app on your iPhone, making cute images of teddy bears underwater, that you are a political, criminal, or capitalist organization, and that you write a script that auto-generates a million beautiful images of your product, in every conceivable environment, with every conceivable attractive virtual person interacting with it… every day.
A million images a day. Total cost? Approximately $10,000. In other words, a million custom, unique, aesthetically pleasing images for about 1/2 the cost of a single professional commercial photo shoot. At this price point, automated interests will be able to generate a unique series of advertising images, tailored precisely to YOU, your cultural history, your color preferences, your mood… and hammer you with them at every turn you take on the internets, until you buy. Buy. Buy. And buy again.
And that’s only the tip of the iceberg: commercial imagery by capitalist corporations which, although questionable, are still operating legally. Their only interest is conversion: to take you from an unknown and to convert you into a customer. Next, to convert you from a one-time buyer into a subscriber. Then, to convert you from a subscriber into a viral influencer. We’ve all read and experienced that script before. Nothing new there. New tools, same old script. OK. Follow the money.
But what about actors on earth with less scruples and shadier agendas? What about political candidates (seeking power) and criminal organizations (seeking customers, recruits, and your identity)?
The volume of AI content vs. “organic” human-generated content is predicted to hit 50% within 2 years, and after that? Game over.
The challenge for humanity is that the AI Flood of generative (i.e. “fake”, “deep fake”) photos and videos will be literally indistinguishable from the “reals…” in every respect. We are getting to the point that if you don’t see it in person with your own two eyes, it may not have ever happened. But many many many will be fooled. Indeed, as we have seen in recent history, many want to be fooled. #bigtrouble
Mark my words: the internet is about to become a living hypnotic hallucination… even more so — FAR more so — than it is today.
So, you ask: “OK, I believe that the AI Flood is truly coming. So: What do we do?”Short answer:a) we fortify our defenses. b) we build the ark. |
BONUS MATERIAL 1:
A few historical stats of generated content volume:
Even 2 years ago, loooooong before the epic dawn of ChatGPT and ChatGPT Pro, GPT-3 was already generating the equivalent of an entire US public library (80,000 books) of new content, every day. So we can say that the AI Flood is coming… but in reality, the water level has been precipitously rising… for years.
as cited from the OpenAI blog, March 2021:
“…more than 300 applications are now using GPT-3, and tens of thousands of developers around the globe are building on our platform. We currently generate an average of 4.5 billion words per day, and continue to scale production traffic.”
As per calculations by the prolific Alan Thompson:
Using an average word-count of 55k words per book, 4.5 billion words would equate to ~81,818 books per day. In 2017, there were 9,045 public libraries in the United States, whose collections contained a total of 715 million books and serial volumes (US stats). Following these numbers, we can calculate that the average library holds roughly 79,000 books (715 million books / 9,000 libraries).
Do you like how I round my figures?
Yep. Just like ChatGPT does.
So that math is… comfortably, approximately correct. 🙂
BONUS MATERIAL 2:
Deep Fake, 1400s style:
imagine://prompt: “AI, show me evidence
that bears played basketball with humans
during the Middle Ages.”
In the hands of artists & creatives,
this stuff can be quite entertaining,
even delightfully humorous.
In the hands of hypercapitalist corporations
and power-hungry politicians,
it’s gonna be lethal.