This is not a good use of water
Posts by a wiki editor
hahahaha oh my god
you're using some random no-name app that no one has heard of (until the times kindly gave it a backlink to help its SEO) and you're surprised it sucks?
this is the equivalent of buying a shirt from temu, having it fall apart in the wash, and concluding that fabric is bullshit
"starting to se a pattern?" sure am bro:
- headers wildly inconsistent in punctuation/caps
- dropped words: "Judging [by] human factors" etc
- constructions more typical of human writing: "a number of", "this is/there are a" (vs. "serves as a" etc)
- few to no constructions typical of AI writing
ah yes someone posting examples of their writing that "sounds like AI" but displays zero strong indicators of AI-generated text
it literally never fails
(the explanation for this may be that it's a specific case of the general tendency of LLM to deprioritize simple constructions with "is," "are," or "has," but who knows, it's a black box, the explanation could just as well be 'idk that's how the vector math mathed)
and even more importantly, it's _the specific phrase_, and that if you change the syntax even a little, it may no longer be indicative of AI.
for example: tweaking it just slightly to "the album is a blend," you now have only 31 articles, and the text is more likely to have been added before LLMs
meanwhile, on wikipedia, it appears 158 times. the majority of these are after 2023 (and in many different genres) and probably contain a lot of AI. here's one, for example: (not picking on this person, they just have disclosed using AI)
en.wikipedia.org/wiki/The_Sou...
- billboard uses it more often... as in 8 whole times. many of those are after in 2023, when AI cannot be ruled out
- rolling stone's website uses it the most of all... a whole 17 times. again, many of those are after 2023, and at least one is clearly labeled sponcon on their australian site
- Pitchfork, in its entire 20-year-plus run, has only said "the album blends" 4 times, and one doesn't count since it's "the album [is so boring it] blends together"
- NME's website has only used it once, in a direct quotation
- Spin has only used it once
- Stereogum has only used it once
whenever AI writing discourse crops up on bsky I feel like I'm being gaslit, because I can _see_ certain phrases proliferate that weren't before
example: a common phrase in AI music articles is "the album blends" X with Y, where X and Y are genres.
"but that's not AI, it's just a cliche!" well...
similarly, if your response (not the OP) to accusations of using AI is "I don't sound like AI, AI sounds like me! I did it first!" then you are:
A) probably not correct about your writing style beyond a very superficial level (hint: em-dashes are superficial), and
B) massively telling on yourself
I don't personally use AI but I have read a great deal of AI-generated text from 2023 on. I think this largely comes down to ChatGPT with GPT-4o being extremely "voicey" in a very identifiable way (if I hear the words "crucial role" one more time I will crucify myself) and a lot of that is gone now
kind of a mishmash of gpt-4, gpt-4o and gpt-5 but still chuckleworthy
the translations happened before the ban did
it's not that it "wasn't written properly," it's that no one could agree on what to allow
good to see this was added, it wasn't there yesterday though
text: Not an approximation. Not a summary. A precise sequence of yes-or-no answers about the image's visual frequency content. Here's the complete pipeline.
like come on. what are we even doing here.
this is AI-generated btw
if I could press a button and magically eliminate the em-dash from the world I would do so, for the sole purpose of eliminating forever the cataclysmic fucking amounts of whining by people who are paranoid that their precious uneditable writing will be flagged by some big bad meanie as AI
I mean it sure would be nice to get paid
good and definitive piece on the "is AI writing inevitable" discourse, which has bothered me
"I used AI to find sources and polish the text, but all ideas are my own."
I mean there's also camp three: those who are actively slogging through the enormous amount of AI-generated content being put out into the world, and are watching the proportion of it grow in real time
you don't have to be pro-AI to acknowledge that this is a growing issue!
it's used generative AI since 2023 www.businesswire.com/news/home/20...
good interview
slate.com/technology/2...
pullquote:
"Listen to your user base and start getting a maximum of everyday users involved in the decisionmaking, because a decision that comes from the top down will always be seen with a lot more suspicion than one that’s built from the bottom up. "
actually going to walk this back based on some output I've seen recently, it might just be a bland generic prompt
idk, having now interacted with him, he does not seem to be coming from a malicious place or have ulterior motives (a lot of the AI editors we get do it to pad their edit count until they're allowed to write a promotional article), but could go about the situation better
also a little strange: at least part of the plagiarized passages seemed to be basic plot summary and a description of a major character, which is:
a) extremely low-hanging fruit if you're trying to pad out an article to meet word count
b) ...something you probably should have included already?
which means that defending the process with "AI didn't generate anything of substance" is slightly disingenuous; even if the generated text wasn't the whole review, it still comprised substantial passages
(I realize he is very contrite about the whole mess, but still)
the writer of the New York Times AI book review did an interview clarifying that he used it only to "expand and smooth" it
it's slightly misleading though; he places the emphasis on the "smooth" part, when any plagiarism almost definitely came from "expand"
samleith.substack.com/p/why-llms-r...