I think there’s a “time is a flat circle” aspect to this. Like in the early days LLM’s would spit out useless but entertaining content. Like that early AI art, it was this chaotic jumbled mess with a very surrealistic tether to whatever the prompt was. Neither useful as a product nor decent art, but entertaining in its absurdity. Then it got to the point where, as long as people put at least a little effort into massaging the prompts, it’ll spit something believable at first glance. You’ve got break out the photo tools to tell, and most people aren’t going to zoom in to see if there’s a sudden jump between jpeg and png edges.

But the technical finesse has brought a plasticness. My wife’s boss, absolute walking definition of failing upwards McKinsey brain, uses ChatGPT for everything. She showed me one of his emails, it looks professionally written and all. But good Christ it’s painful to read. It doesn’t look like something a human being would earnestly write, more like a satire of corporate speak that would be considered too obvious to work as satire.

And that’s where the circle comes around, all this AI text, art, movies, etc have this feel to them that makes you aware that it’s AI-generated. That breaks the spell, it makes you aware that it’s a plastic facsimile of creative output, and that devalues it. Which will conversely create a value for creative output that feels distinctly not AI.

I’m not sure if there’s an exact 1-to-1 when it comes to “programmers” who don’t really know how to code, they just know how to ask ChatGPT for a function. But I’d have to imagine that they’re in a position where they’re fine if things are running smoothly, but the moment things get FUBAR’d and ChatGPT can’t spit out a solution, or the solution is so convoluted and messy that it makes more problems than it fixes, that will separate the chaff from those who actually know how to wrangle with the code.