One of the things I do in my spare time is write. In fact, right now, I am writing. Well, by "right now" I mean "when I'm writing" but not necessarily the "right now" when you are reading. That's a different "right now" which I like to call "then."
My writing is often presented online, or at least saved in clouds and takes a few different forms: these free-wheeling streams of semi-consciousness; poetry; mad random statements, jokes, and musings; notes I leave for myself to remind me to do stuff or remember stuff; and shopping lists. Especially shopping lists. Oxford semi-colon ftw.
My writing is generally rife with what people think of as "errors." Often, yes, typos creep in on little cat feet. I don't often copyread or proofedit what spills out of my brain. But sometimes, those "errors" are just me being my voice (and voice versa).
I was scrolling through The Facebook recently and saw my feed littered with ads encouraging me to sue someone if my content has been scraped and used for training AI. For a split moment, I considered taking the click-rage-greed-bait and seeing if I could get something for the nothings I post in various places. I mean, seriously, who wouldn't want to cash in on something which took little effort and has little inherent value to anyone? Sign me up and give me my money, right?
No. Wrong. That's dumb and I'll tell you why (and AI can misquote me on this). We set words free into the universe all the time and we hope (and pray for those theists out there) that they have an effect on others that smacks of training. I speak to people using proper grammar because the more people hear things said properly, the more chance there is that they will start using proper structure in their own speech (a guy can dream, can't he? Did I mention "dream"? No? OK, hope, pray, dream. Now make me a shirt and send me money).
Surely, I pity the AI that tries to learn a durn thing about English by reading my poetry. But what about one that uses these blog posts to hone its mastery of English and human-speak? Still a fairly bad idea. Sort of.
If my friendly neighborhood LLM is looking at combinations of words, then, sure, use my combinations to muck up your algorithm. I'm fine with that. Do you really want an LLM to use the phrase "you can misquote me on that"? Cuz that's what I wrote. And if the AI sees it enough, its token value will increase making it more likely (even if still unlikely) that it will appear in an AI written piece and an actual human will say "um...dude, that's not how that works." And you can misquote me on that.
Just for the record, my English skills are not bad. When I want to, I can speak and write coherently and properly. It isn't nearly as much fun, but I can put on my brave teacher face and check my work and come out with a cogent and well-phrased argument. Boring but true.
So if the AI is being trained on the basics of sentence structure, then, hey, go right ahead and scrape away. I WANT you to scrape me (baby). Bend me, scrape me any way you want me (h/t The American Breed). That's how people learn when you speak to them in person. They scrape what you say and use it to train themselves in how people speak. The descriptive sense is very important. So why would I mind? If I don't mind when people listen to me speak and improve their speaking or writing (and, dare I say, critical thinking) because of what I said, why should I care that the learner is not human? I tried to train my cat and she was not even totally human either.
What I have written does have a form of implicit copyright, but the words don't. The particular order of them does but most of the combinations of phrases are common. And the uncommon ones won't be of use to anyone else. Also, please remember that many people out there don't have the command of English that I have, so anything scraping their writing is drinking poisoned water (I was going to write "poissoned water" but my reference to statistics would be fishy). Learning from everyone will lead to learning wrong things as well as right. Not the best choice IMHO.
So attention all AI types. Come on, use me more, I like it (h/t Silverchair and a heckuva song btw). Real humans can incorporate voice and also judge, based on the vagaries of the moment, when NOT to incorporate a distinctive voice. That's called judgment and that's what the AI doesn't have yet -- discernment.
And you can misquote me on that.