Can we deal with the downslope?

In 1945, Belgian surrealist E. L. T. Mesens wrote a perfect musical composition. Above the first bars of traditional scoring, he wrote "In the beginning, everything is fine". Then the composition descends into chaos. Drawings of massive hands, bugles and made-up musical notations appear. In the end, there is nothing sensible to play – and that is the point.

I love that phrase: "Au debut, tout va bien". It is both reassuring and chilling. It applies perfectly to this new AI-era we are entering.

AI is a magic trickster. By shuffling and reassembling trillions of existing words, images and styles, it can mimic new human output perfectly. This is nothing to sneeze at: great value is added. We can now do twice the work in half the time. Everything seems fine indeed. Smart copywriters who see the writing on the wall mutate into AI-empowered super-fast editors. Mediocre armchair directors create exploding dinosaurs on Mars without having to justify their creative visions to a stingy film producer. Anyone can now produce content – and so much of it. No wonder everybody is happy.

But then, just like in E.L.T.  Mesen's  composition, the first ominous drawings appear, disrupting the music.

First, AI and human content become indistinguishable. Eager to be seen as real, we sell the blend as a true human product, like unscrupulous traders spiking cheap rapeseed oil with green coloring to create fake virgin olive oil.

Second: we start to make too much stuff. So much that no one has the time to consume it all. Netflix, YouTube, LinkedIn, Facebook, Instagram all pitched in and now our average attention span on TikTok has dropped to about 1.7 seconds per video. Like a fly trapped in honey, we blissfully drown in an abundance of our own content.

Then comes the downslope.

To capture the fleeting interest of an exhausted audience, we pivot to harvesting attention instead of interest. We grab the audience, hook them with big explosions and "wow" statements up front, all in a bold and assertive tone that demands: watch me!  AI becomes like the Clippy from Microsoft: "It looks like you are trying to explode a dinosaur. Can I help?" Always eager to assist and take things to the next level, it helps us amuse ourselves to death.

Now for the musical finale.

AI trains itself on our output. But by now, this output is increasingly a blend of human and AI content. With no way to distinguish the two, AI is starting to train more and on its own work. Our massive output of attention-grabbing, syrupy ersatz blend is now becoming the main source from which AI distills its new content.

And so the spiral tightens. In the next five years, the first cracks will appear. Nonsensical or wrong information will be asserted with bold confidence. Long-form text that allows room for nuance or background will be largely replaced on screen by soundbites. Truth in image - seeing is believing – is already under scrutiny and will be forever lost in the form that we know it. Soon we start to live on a diet of rehashed soundbites, infographics and quick facts, forgetting what nutrition is like. We lock ourselves inside a candy factory, forgetting what real food is like.

And next? The apocalypse? I don't think so. This too shall pass.

The invention of a practical photo camera did not kill painting as an art form. The miracle
"one-pill" dinners and prefab meals from the sixties did not replace the art of cooking. AI will find its niche, and we humans will continue to occupy the rest of the space.

In the beginning, everything will be fine. All's well that ends well.  It is the transition period in the middle that has me worried.