"A lot of the time, AI slop is difficult to discern from authentic content until you’re somewhere in the middle, at which point something starts to feel off abo

"A lot of the time, AI slop is difficult to discern from authentic content until you’re somewhere in the middle, at which point something starts to feel off about it." And it's only going to get worse. I don't know what the limits of LLMs are, even before we start discussing future architectures and capabilities, but it certainly seems to me that if they just scale a bit farther on their current trajectory it is going to be very hard to detect AIs simply by the content. One of the adaptations I've adopted is that if your content looks like it was generated from a simple prompt, no matter how long the resulting content, I just toss it. "Write me a beginner's tutorial for Python" isn't going to produce a document that the internet needs another copy of or that social media sites need to spend time linking to. Unfortunately that means even if the "tutorial" was human-generated, it's probably getting the axe too, unless it has a very interesting or unusual approach, but it isn't clear to me that that is actually a large problem.