In “The Further Adventures of an Artificial Intelligence Refusenik”, I have realised that, if I ever need to prove that I am indeed the author of anything to which I have placed my name, I may need to go back to writing out that work longhand or, at the very least, plan them out using pen and paper.
This may sound like the latest stop on a road to paranoia, but for as much flak as people gave Ed Sheeran for revealing how he employs a videographer to record his songwriting process, to avoid further lawsuits over perceived breaches in copyright, his need for incontrovertible proof of his own creative ability speaks of how much the assertion of authorship has, well, been taken as written up to now.
I have found it hard to write much in a creative capacity recently because the presence of A.I. makes the act of writing mechanical in a way that threatens my dream of making it a livelihood, a threat I could not have envisaged when I started writing articles in 2016. The continued use of A.I. may require people to prove they did not use it to write anything, from a letter to a news article, from a short story to a complete novel. Rather than just needing to have something in my writing to help make me stand out, or to protect my ability to write, what we all need is something that proves that consideration was made into what words were used.
What highlighted this issue to me the most — although, to be honest, sniping about A.I. reliably rouses me anyway — was the use of em-dashes. The apparent story regarding these is that, because the ChatGPT program uses em-dashes as its default dash, not distinguishing its use with that of an en-dash or a hyphen, marks it as a red flag. An em-dash is used when you want to make a separate point within a sentence, like I did two sentences ago, but I know I am guilty of not selecting the correct dash while typing, which ironically could save me here.
It makes me wonder if this is acting as a kind of A.I. “watermark”, like imperceptible watermarks that can be added to A.I.-generated images, an irony when A.I. is often used to remove more visible ones. If A.I. doesn’t know when to use the right dash, but consistently uses the same wrong one, is this evidence of a wrong-footed style that acts as a deterrent for people to choose their own words instead?
Not really, as while Google has created an open-source “SynthID” that records the weighting given to the choice of words by its text-generating programs, something similar is required for every other program that does the same. Until then, self-declaration is the name of the game, suggesting its own questions of motive depending on the answer.
People shouldn’t be left with the words they need to get by, or to have them mean enough for what they need. Playfulness needs to replace the paranoia. All we have our words, which I will need to write in ink.