Twitter was worse about this, but yeah. Lots of "uh, actually you're ignoring xyz" when in reality, xyz was either simplified or just...wasn't relevant. And that others have come up with similar layman's explanations to the one I started using a few years ago says A LOT.
Well, primarily one thing very loudly. That these GenAI systems are not actually that complicated. That when you boil them down, you're usually going to get "person inputs a thing, it jumbles the input up a bit (or if an LLM finds the statistically likely word to go next), then spits it out".
That's it. That's all they do, and they can't do any of that without a person's input. Because they're just tools. But unlike a "paint bucket" tool, these tools require intense resources to use and maintain, including stolen data. Making these tools just...Not worth it in the slightest.