| Bot detectors are broken. Even human bot detectors are broken. When I'm in the right mood, I can be quite capable of writing with very good formatting, structure, and phrasing. When I actually take the time to do this, there seems to be about a 70% chance that some nimrod will crawl out of the woodwork just to accuse me of being a bot. Even humans who deliberately use lazy formatting and leave obvious errors uncorrected to provide "proof" of being human aren't seeing the big picture, here. --- That bigger picture is that it's easy to make instruct a bot to be lazy, or to avoid the usual quirks. I hate when I'm working on a project and see a constant outflow of negation ("Don't do x, y, or w" is a recent hit) and unfounded exclusive confidence ("The correct answer" as if this is Highlander and there can be only one). Repetitious jargon like overuse of "gate" for things other than fences and skiing is something I can't stand. Plus the usual things — like overuse of unusual punctuation — that are obvious tells. That stuff all drives me nuts. But the bot just follows instructions, and my bot has been instructed to avoid those things. It generally performs very well, though the instructions do need re-hashed every now and then as models ebb and flow. It's super easy to get the bot to write some python or perl that takes a body of text and intentionally some words or lose a comma while mmaking other errors and converting — into --. When it comes to human error in written language, we just aren't that hard to emulate. Now, that all said: You'll just have to take my word for it, but I do not use the bot to help with writing English. But I do have every confidence that if I woke up tomorrow and actually started bulking up my comments using a bot, none of you would be able to tell. |