With image generation, you can generate seemingly infinite images with nothing but a prompt, but you’ll never get what you really want. You’ll get an approximation at best. You have all this agency, but also no real agency that matters.
But of course if you’re a visual professional of some kind then I imagine it’s even worse, a net decrease in agency. Not only can you never quite get what you want, but having to go in and edit things is tedious and dull. Even if you’re saving time this way, the lack of enthusiasm and flow makes it feel like the job takes much longer than before.
Likewise, with programming, at the end of the day it feels as if I’d be more productive and just wrote down my thoughts as code rather than delegate to a copilot or ChatGPT and waste time making sure this actually works or solving for the occasional bug that came from a chunk of mystery code.
At the end of the day, you aren’t creating the image or the video, the AI is and you make adjustments where you can but otherwise accept the results. Extrapolating this to software, I don’t imagine a future of where people are empowered to write their own software, but simply a future where people ask for software from an AI, it gives it to them, and they do an awkward back and forth to try and get the details right until they inevitably accept what’s been given to them.
It’s depressing, but also so goofy that it’s hard to imagine this being the final outcome.