The fear of AI: how I got dehumanized
I'm human, a fellow human
Taking a break for the muse-related content to have a little story time.
The other day I signed up for Vocal - a Medium, blog-esque platform for writers (this is my profile, by the way). I was interested in it because I’m always on the lookout for creative venues. Ideally, I would love to post fiction where people actually read it and where I’m not kept out by obscure editorial reasons.
So, I went ahead and tried uploading an old short story about the capitalism applications of a time-compressing device. Your run of the mill thing.
This is what happened.
Uh-uh. That’s strange, I thought.
Disclaimer (feel free to unfollow me): I’m not against using AI. I’ll probably write on why and how to use it someday, but this is not the day. What matters to the context of this story is that the short story in question (titled “Essentially, a tool”) was written in late 2022, before generic availability of writing AIs.
The warning didn’t mark which sections were suspicious, but now I was on a quest to publish something. Anything.
Of seven stories I submitted, all written by my own hand, only two passed the check. That means there’s a 71% of false positive rate. That was high enough to make me doubt myself.
Am I human, after all, or am I a statistical software?
I got in touch with support.
Due to the influx of spam and AI-generated content we are experiencing, which makes the platform unreadable, we have automated some screening procedures. We are continuing to refine this process daily, and will ensure this does not happen to you again going forward.
This is fine and dandy, except for the 71% false positive rate seen above. I was still high-spirited tho (and I find it incredibly funny). Basically, this is a conundrum.
Most of my stories were submitted on a Sunday. I am quite sure that the screening procedures that blocked my posts were automated, sure - but that’s just another word for algorithm.
Now, we are in one of two cases:
The “automated screening procedures” are AI-based. There are AI services trained against Bard, ChatGPT and so to recognize AI-input.
The “automated screening procedures” are human-coded, meaning a program running some kind of static statistical analysis. Maybe there’s a correlation on the usage of commas, or paragraph breaks, or the word “perchance” that correlates strongly with AI.
Basically it’s either a poorly trained AI or a poorly written software. Now, using AI to ban AI brings its own subset of problems, but that’s not even the end of it.
Want to know the cherry on top of the cake? One of the few (at the time of writing, two) stories that managed to get through the filter was titled “This piece was AI-generated”. You can read it on vocal. It’s a short reflection originally intended for the Archive of the Odd1 , were I pretend to be an AI.
In short: apparently, when I pose as a human, I’m flagged as an AI. When I pose as an AI, I get recognize as human.
Who am I, really?
Addendum.
More or less when text-based AI became widespread, Clarkesworld closed submissions to combat the influx of cheap AI-pieces2. What Vocal is experiencing, albeit not new, is a issue.
I recognize that.
As of someone who uses AI in the creative process, I know that it tends to repeat the same styles over and over. State of the art AI seems unable to escape a certain overall amateurish quality. The idea of using AI start to finish, while interesting on a technical level, doesn’t yield good results without heavy human intervention.
Yet. Vocal and Medium pay their writers based on views. Magazines pay by word. Publishing used to, and now rarely does, pay in royalties.
What we are seeing is just people using a tool to play well inside the rules of the market. Sure enough, many “AI-bros” are hacks. But isn’t that how capitalism works? Isn’t everyone on the lookout to make a quick buck from the gig-economy? Maybe the issue lays somewhere in the system.
There is, something rotten, in Denmark after all.
If people start hitting themselves with hammers3, we wouldn’t put a ban on hammers… or would we?
On another line of inquiry, I am aware of the fact that English is my second language. Maybe that’s what triggers the filter. Especially if I’m tired my sentence structure tends to be less than idiomatic. Not to mention my vocabulary and my hopping between British and American terms (I cannot tell the difference and I couldn’t care less).
So, am I being discriminated?
If you by any chance liked this, I assure you I have even more useful posts in store.
https://archiveoftheodd.com
Who you are?
A robot Pal or a bio Pal?