Will AI End Up Eating Itself or Just Become A Giant Content McFlurry?

When AI blends everything into algorithmic smooth cream - what is left?

Thanks for reading the Creative Stack. To support my writing please consider taking out a subscription.

Every month a group of writers meet over Zoom to read and discuss our work. Group members are from Ireland, the UK, California and Brazil and we work in tech, politics, medicine, taxation, financial services, screenwriting and theatre. All of us have been exposed to AI in our regular work and, until now, none have us has used AI to create anything.

This month we discussed what the limits could be. I have tried Chat GPT and Gemini for proof-reading which is excellent and editorial feedback which is interesting and logical. It often offers to create ‘line tightened’ versions of my chapters but they come out flat and smoothed down like an auto-tuned vocal. They’re bad but they are also not me; sometimes I write a sentence that jars, or has out of kilter tempo because that is how life is. AI cannot replicate this.

Our group discussion got us all thinking about how far we should take the use of AI. In Hollywood there is a move to replace writers’ rooms with AI and a script supervisor which will have the result of making US made movies and TV even more homogenous, filtered, and dehumanised - and even less compelling.

The technology that gives us AI is undoubtedly incredible. The problem for AI in the creative sphere is that it cannot add anything; it can only re-use and recycle existing ideas. Each time creativity is pushed through the blender the result comes back smoother and blander. All the pips in the strawberry milkshake are smoothed out until you have so little texture that you might as well not even use strawberries anymore; so you just use a syrup. The syrup is artificial, it is loaded with sugar to make up for its lack of real fruit flavour. Eventually (very quickly), your simple milkshake becomes a McFlurry; a hybrid Frankenstein dish that needs stabilisers, sugars and syrups just to exist.

Many people use AI for their social media posts and Substack articles. AI is supposedly tuned in to what algorithms are looking for. But if we use AI to create content that will be picked up by algorithms and those algorithms are run by AI then the technology will eventually just eat its own tail; AI Bots will write articles, post on social media and publish books, produce movies and manufacture music that is designed to be picked up by other AI algorithms.

This AI generated product will appear to be popular because all the other AI bots are triggered to respond to it. AI powered Spotify playlists pick up songs with the right AI calculated musical cues ensuring that they get streamed. But they get streamed not through any qualitative classification, rather as a result of a quantitive analysis.

There is no human listening and saying “this sounds like an interesting track, people should hear it.” Instead AI checks it for song length, bass sound, vocals and whatever other machine generated data points it needs.

This apparent success feeds into what AI then uses as the basis for its next work because it only recognises success. AI cannot measure quality; it can only measure success. The problem is that AI’s KPI is achieved through AI; we are at the point, or very soon will be, that AI is just talking to itself. Its sphere of influence and its own influences reduced to its own kind; to those who share its algorithmic DNA.

Just like humans get radicalised online by burrowing into internet rabbit holes. AI will only see what it already knows and will magnify that. The technology becomes the Gogglebox of the modern world; infinitely self referential until the Gogglebox participants are sat on their sofas eating chocolate biscuits and commenting on themselves on Gogglebox, sat on their sofas eating chocolate biscuits commenting on themselves sat on their sofas eating chocolate biscuits.

AI is a tool, in the same way that steam engines were tools that drove the Industrial Revolution. It can do things for us - not things that we are incapable of - and it might be able to do them quicker than we can. It might be able to impress us with its easy solutions, its credible sounding academic essays or its ability to analyse thousands of phone calls in seconds. But it is not us.

It cannot explain its decisions, make moral judgements or empathise. Its inability to cope with these very human tasks means that AI cannot be spontaneously creative either.

Humans can guide it to create a song or a painting but as IBM admits;

“while computers can be coached on some parameters of creativity, experts question the extent to which AI can develop its own sense of creativity”

Without its own sense of creativity, AI will only continue to chase its own tail in an infinite zero sum game. Once it has refined and smoothed out everything that it has already produced it will have become irrelevant, boring and in need of human guidance.

We can only hope.

I’ll probably keep using AI to proofread but not to rewrite or smooth out. Because it is the textures, the jarring rhythms, the voice that makes what we write worth reading in the first place.

Thanks for reading. To support my work please consider taking out a subscription.