For Instagram, my art is "Likely made with AI". Hard to be an artist in 2026...
26 Boosts
2 Likes
@davidrevoy Here's the dirty secret nobody tells you: these "AI detection algorithms" are themselves AI, and subject to the same "hallucination" issues.
@me @davidrevoy yep, reminds me of this gem
@maxlath @me @davidrevoy put som HG. Wells through and see if that dude had time travel after all. If he was using AI we know he could travel through time, meaning it was all made up, so he didn’t have…shit this might be one of those paradox things we should avoid.
@maxlath @me @davidrevoy Seems to confuse 19th century flowery prose with AI verbosity.
@maxlath HAHAHA oh god, I made the ugliest laugh.
@davidrevoy Ugh. I dread the day that people start using AI to make fake "process" videos.
Sorry you need to be on Insta for the reach.
@raven @davidrevoy guaranteed that it has already been done, unfortunately.
@davidrevoy AI detectors are just pure bullshit. I've fed my own hand-written code into a few of those and they all said the code was between 70 and 100% AI (when some of it was written before ChatGPT could spit out a fizzbuzz that actually worked)
@davidrevoy AI detectors are just pure bullshit. I've fed my own hand-written code into a few of those and they all said the code was between 70 and 100% AI (when some of it was written before ChatGPT could spit out a fizzbuzz that actually worked)
@davidrevoy sounds like a confession:
AI was trained to copy whose art?
I put some of my works, even actual scanned pencil drawings that were never touched by a painting software, through some AI detectors just to see what would happen and my traditional pieces were often rated as being likely AI.
@Takiro that raises the question: what are the criteria for tagging a picture as ”not likely AI“?
(Respectively, on what basis is the likelihood for ”not AI“ computed?)
🤔
@davidrevoy
@davidrevoy, this is so sad. I committed digital suicide on these platforms. I was too annoyed, and not only once. Following you on mastodon now.
@davidrevoy That's because AI was trained on your many great CC-licensed webcomics! Instagram doesn't know you're the source. 😉
I'm surprised (not) to see that even the few models trained solely with copyleft-compatible data still don't bother tracking which tokens correspond to which files for attribution purposes - at most they're tacked in after the model is trained already.
Is AI detection the biggest scam in the AI bubble?
@davidrevoy
i wrote a whole rant about it with all my lungs
essentially:
the problem of #AI is not fakes (although that is a problem, it's not the main problem)
the problem of AI is that it enables this lowest common denominator asshole going "fake, AI" under EVERY GODDAMN NEWS ITEM AND ARTWORK
that lazy behavior becomes a #disinfo vector all by itself
the point of the #propaganda routine the Firehose of Falsehood is to make us distrust everything, even the real
https://mastodon.social/@benroyce/115435710841470406