Wrote an essay analyzing Parasyte's many characters as facets of a whole:
Narrative Analysis of Parasyte: A Panoramic Portrait of Humanity
https://t.co/9XosMxLSxH #parasyte#parasytethemaxim#anime#manga#寄生獣
reading non-ai text and literature will give you a wide understanding of what human writing looks like, but known exposure to ai is what inoculates you to noticing ai itself, hence my recommendation to look up guides. for effect, try to see what you notice first and compare.
in my opinion ai is detectable based on "vibes" aka intuition, pattern matching, but since intuition itself isn't proof harassing someone about it isn't the best way to go. pattern matching means you need to know what the patterns themselves look like. 1/6
every company's ai and the updated iterations called models each have their own voice when it comes to a specific kind of writing, so familiarity makes it very detectable over time, given proliferation of its use. however, you can't detect a pattern of nothing.
@9r7g5h@fujocel69@AruBerus The pattern of no cognition repeats continually - no intent, no impact, no significance, no real evidence of an image they're imagining. And AI only gets worse the longer the story is, unlike writers who often develop/refine over the same period - plot, characters, meaning.
@9r7g5h@fujocel69@AruBerus Blocked, so posting here. 1st 2 paragraphs r questionable bc writers would naturally expand on lies/perfume topic or leave it restrained, whether skilled or clumsily. But topic jump says writer somehow forgot topic mid-sentence. A pattern of no cognition.
https://t.co/3EEPzCMPTj
AI writes like it's paid per word yet trying for restraint. AI pattern matches & generates based off humans, so detection needs both familiarity & intuiting patterns before proof. Intuition isn't consistent btwn people or 100% right, but here's what I perceive in this AI pattern
this sort of thing you can't really point to as proof. people interpret text in highly different ways, and the gap increases with lower familiarity, so what sounds like ai to one person (and may not be!) can be entirely invisible to someone else.
honestly ai doesn't process like humans so its logic and the way it writes is always going to be different from human processing. any general model is consistent yet inconsistent bc rlhf drags it around in contradictory ways.
different people process in different ways, but -
this causes a disjointed feeling, like a lack of object permanence, which is even more jarring bc ai is a logical writer overall. if you generously interpret ai work in a scene, it technically works, but the flow of concepts doesn't match the displayed skill level.
@beeenado Restraint for later would avoid barely-related comparisons that overstretch the perfume into multiple subjects, and overuse would continue to how the perfume/lies are sweet/suffocating and then bridge into his rep, instead of the rep being ahead of the perfume's/lies description.
@beeenado The smell went from "perfume and lies" to "perfume and reputation" instead of expanding the 1st metaphor or letting it linger. It's not a good tell on its own but skilled metaphor -> clumsy swerve w detailed themes = no intent behind the metaphor.
@heng9arae It went from "perfume and their relationship's lies" to "perfume and his reputation" instead of continuing the original metaphor or leaving the metaphor to linger in the background. Even then, it's not a good tell by itself bc both good and bad writers love metaphor.
@heng9arae I don't like highlighting such a small portion bc people get the wrong idea. It's not the metaphor, it's the way the metaphor breaks down immediately. If you have parts that are "out of place" it's obvious it's not ai bc ai atmosphere is either samey or abrupt style change.