This, here, is a good exercise and analysis for any of us who write online (or write at all, I suppose, but for most of us that’s online).
If you haven’t read it before reading this – and you should read it at some point, at least – it teases apart what could count as “writing with AI”, and argues that rather than either getting all ban-heavy, or requiring labelling of AI-written material, better simply for those of us who write for ourselves to be transparent about how we do what we do.
I like that, so here’s my version, for posterity.
I use AI a lot, partly because it’s fascinating, and partly because it’s really useful. What I don’t use it for, much, is things where I like the process, which includes writing and art.
So increasingly I use AI tools for coding, and analysis, and error-finding, and research. If something has a bug, an AI can scan innumerable lines of text and find what it probably is far faster than I can. There’s also a weird feedback loop developing at this point, where clients send me specs that AI has helped them with, and they tend to be so wordy that I can only makes sense of them by getting AI to tell me what they say. There’s probably a lot to say about that, but I’ll skip it for now.
There are some things AI can do for me in my work, though I’m so far having to create my own means for it to do that, and my main preoccupation is finding ways to ease what I can do without sending client data outside my own control.
But this is really about this blog, and my other personal writing, like LinkedIn posts.
There, AI isn’t much use to me. I like writing, and, as the linked post says, if I automate my hobbies, what am I even doing? Plus I actively dislike the changes AI makes if I ask for suggestions. Yes, I have tried, and try occasionally again because it’s not a static field.
So in the sense that people usually mean, anything you see from me is completely non-AI, all me, no matter how many dashes, rules-of-three and other “tells” it may have.
But transparency should go deeper than that.
The use of AI that readers would probably most want to know about, in my case, comes when I want to say something I know will be provocative or controversial, which I am in the habit of doing because I feel strongly that there are many subjects we avoid that shouldn’t be avoided. In those cases, I have got into the habit of writing what I want to say, then posting it into an AI for feedback, particularly about holes in the arguments, weak spots, or what is likely to provoke unhelpful reactions. That works quite well for me, because it actively makes use of the way AI is intensely average, so I can gauge from it what are typical responses.
In those cases I don’t allow the AI to rewrite for me, because it doesn’t do that in ways I’m ever happy with. But I can no longer say AI had no input into the result, because I do quite often rewrite for myself based on what it says.
I do the same, with less ego involved, with website copy, since that also is a means to an end, and an AI can mimic the audience and tell me what it might be thinking in the way the real audience never will.
My writing isn’t research-heavy, it’s mostly pure opinion and based on knowledge I’ve already gained, but sometimes I’m lazy and will ask AI for some examples of something, although I do then investigate those before using them in my own way. And when I write about books, sometimes I read them long enough ago that I’ll check whether the impression of them that remains is reasonable.
In summary: everything you see me write, I wrote, without so much as a grammar-checking tool.
But what went into the writing may have had AI input, or have been shaped by AI feedback on what I wrote initially.