Ai in writing — Don’t lose yourself
by Eric E. Cane
I think a problem people have with using AI in writing is that they are too trusting of the output without double-checking or having the AI check itself. This and that most of the output is similar in its pale imitation of the original human expression, but that’s for another article.,
In my own writing (novels, short stories, screenplays, essays like this one), I use the AI for checking punctuation and grammar. I explicitly make sure that it doesn’t try to add its own words or modify any of my content. I have it simply put in brackets suggestions for grammar changes or notification of spelling errors inline. Other times that I use it is when I have tons of notes written over the course of days, weeks, or years and I have it organize them chronologically or by heading and that sort of thing.
I still value the content and expression of my own writing. I think the reader also values that. It’s sometimes easy to see when somebody is using AI for creative content, especially after you’ve studied the different models and their output. Another easy tell is when the output is factually incorrect — like “off the walls” incorrect. Now, humans can do this incorrect thing with great skill and frequency, but that’s why content editors exist. Different magazines, online and otherwise, are getting caught using AI for content creation in their rush to be first with some article. The problem with this is that good writing, and especially good journalistic writing, takes time to fact and source check so that public trust isn’t eroded.
AI can do things very quickly, but it’s up to us to actually make sure the expression comes from us if we want to have any sort of personal expression put out into the world. Where AI excels is in its ability to process a lot of text quickly for checking errors, but you have to make sure that the AI itself is double-checked. You can do this simply by telling it to double-check its own reply. You can also tell it to process your directions step-by-step and then add the direction of “double-check your own reply” at the end. Then have it double check its own reply again. Not kidding. Even with the best models, I’ve had to do this and find that it frequently back walks on its recommendations, catching itself in error. This goes for very simple things like grammar and spelling — and it most certainly must go for anything you’re putting into the world where you want to be seen as a source for accurate information. In cases where the information is really important, this needs to be verified through several different sources. It doesn’t require that much work in the grand scheme of things, but it does take a little bit of time. Search engines like consistent output, but if that output is frequently error-soaked rubbish, then viewership will suffer. Trust will take a hit that is difficult to recapture.
I’ve been testing AI in various forms since the release of the very first open source models. Nothing I’ve seen to date compares with a true and heartfelt human expression. In some cases, it’s getting close. You might laugh and say that some people don’t write well enough or connect to their own internal source for expression well-enough that their writing is somewhat bland, passive, robotic, already. So in those cases, there isn’t that great a difference between the AI output and a writer who really hasn’t found themselves yet or mastered the craft of the medium. And you would be right.
In my own life, I try to continually refine my expression. Not just the expression itself, but strengthening the connection to put into words my own internal thoughts and feelings in a way I think others will understand. That’s my goal. Once it hits the screen, I have to then rearrange my dyslexic version of this dialog that my own internal checks missed. It’s part of why my writing takes much longer. But I believe that process shapes my expression in a rather original way. It’s not always clean, not always perfect, but the effort brings its own rewards. I think that’s humanity in a nutshell.
(Please feel free to clap and share this so my work gets out to more people. Thank you.)
