Beyond The Kaleidoscope

Kristian Dupont
3 min readJan 4


Generated content has been around for ages. I remember dreaming of video games with unlimited territory to explore, with the obvious solution being for the terrain to be generated by some clever algorithm. As it turned out, every time I actually played a game where something was generated, the experience was consistently underwhelming. Spore was supposed to be the ultimate in terms of emergence, and it was boring!

I struggled to pinpoint what exactly was lacking. Yes, generated content was not “human”, whatever that means, but neither is the real world — nature and the shape of our planet is the result of a bunch of random events (unless you believe in intelligent design I guess, in which case this whole discussion must take on a new level of semantics).

Photo by Malcolm Lightbody on Unsplash

One distinct problem was that even though the generated content was unpredictable in the specifics, you would always detect patterns. The way looking in a kaleidoscope theoretically offers unlimited visuals but still gets boring quickly. But even when such patterns were hard to detect, the mere fact that there was no intent behind anything just made it feel pointless. Which I find interesting philosophically, because why would a made-up but equally unreal world be any less “pointless”?

I think what has me so stunned by the advances in AI that we saw in 2022 is that this limitation seems to have been overcome. The output from various AI generators doesn’t seem pointless to me. ChatGPT writes fiction where I want to follow, I am curious about what it comes up with. In spite of it not being sentient, not knowing anything about Australia, math or why these texts are funny, I find them to be genuinely entertaining:

(Incidentally, John Cleese gave a great talk on creativity which it seems ChatGPT has watched)

Similarly, the image generators create stuff that triggers emotions in me the way looking at (human-made) art does. I have yet to experience this from a music generator but I suspect I will before the end of 2023. One of the strongest indicators of the shift, to me, is that we now seem to be discussing what art is and isn’t. It reminds me of the discussions that appeared when Bitcoin was created about what money is and isn’t.

The thing is, all of this simply illustrates Moravec’s Paradox which was well known years ago. For some reason, I still believed that the progression of automation would go: blue collar → white collar → executive → creative, with the last step being a possible “never”. Turns out that was where it started. While I don’t believe we’re at the singularity yet (the tools that generate code are mind blowing but still quite helpless and certainly not capable of creating more advanced AI themselves), it does feel like the world has been turned upside down.