What's new
By:
Choice of Mods

Join our amazing community to find Choicescript games and modifications to improve your gameplay!

Stories that left you disappointed for a reason

Grace5963 said:
I did look at the proof post she linked to for her fanfic and by the looks of it, it's human writing as I don't think ai does edits or things like that
Click to expand...
I hate to be a downer, but the thing she provided doesn't prove anything, sadly. One of my friends works in education, and she says that at this point even asking students to hand in a document with an edit history is pointless now because a lot of them just type the stuff they get from AI manually.

Also, the way she showed the edits made me go ???. As far as I can see, this is not edit history from Google doc (it highlights any added words), so it is done in editing\proofreading mode in a text editor like Microsoft Word or smth similar. Now I am not saying no one ever edits their work in editing mode, but this is very... errr... unusual, to put it mildly. I'd also say that she either doesn't know basic grammar (one of the examples in her fic is: you watch her for a moment — the line of her jaw, the way the light catches... This should be a colon, not a dash. There are two reasons why there would be a dash: you don't know rules or you used AI.)

When AI first came out I played around a bit with it; not character ai, but more sophisticated 'writer' models. Her stuff reads like what these models put out almost to a T. The biggest tell for me is not even the improper dashes or antithesis, but what I call 'triplet parallelism': for some reason AI loves to create sequences of three near-identical rhythmically or grammatically structured sentences.
 
Licklo98 said:
Summer of love rewrite reads like ai so bad
Click to expand...
It does. Then again, Tumblr writers tend to write a certain way so it's possible that AI learned from them and not the other way around. I've read something on Tumblr before 2022 that would look like AI writing now. The rewrite wasn't needed though. There's no flow and no feel of action somehow, the prose feels very purple, and it's very flat. Also, thesauruses are a thing. I got tired of seeing the word "clock", which makes me feel like it might be AI.
I'll give her the benefit of the doubt though, but every passage seems longer for no good reason. I skim descriptions often so when you take them out it's almost like nothing has actually happened.
 
xeroxo said:
It does. Then again, Tumblr writers tend to write a certain way so it's possible that AI learned from them and not the other way around. I've read something on Tumblr before 2022 that would look like AI writing now. The rewrite wasn't needed though. There's no flow and no feel of action somehow, the prose feels very purple, and it's very flat. Also, thesauruses are a thing. I got tired of seeing the word "clock", which makes me feel like it might be AI.
I'll give her the benefit of the doubt though, but every passage seems longer for no good reason. I skim descriptions often so when you take them out it's almost like nothing has actually happened.
Click to expand...
Also, AI tends to cut off pronouns and seems to dislike using conjunctions, which happens a lot here e.g

> Good," she says, and she doesn't elaborate on what's good about it. Just picks up her clipboard, smooths her jacket, and opens the door.

Nothing wrong with cutting off the sentences to make it "just" instead of "she just", but it happens a lot. AI tends to do that.
 
xeroxo said:
The rewrite wasn't needed though.
Click to expand...
Imo, while the original had some grammar and spelling mistakes, it did read as a honest text, and there was nothing wrong with it. The new version seems both sanitised and too much at the same time to me, idk.

I feel bad for her if it's genuinely her writing and she has to deflect the suspicions, and being accused of using AI when you haven't sucks, but at the same time I have hard time believing it's all authentic.
 
Ireth313 said:
Imo, while the original had some grammar and spelling mistakes, it did read as a honest text, and there was nothing wrong with it. The new version seems both sanitised and too much at the same time to me, idk.

I feel bad for her if it's genuinely her writing and she has to deflect the suspicions, and being accused of using AI when you haven't sucks, but at the same time I have hard time believing it's all authentic.
Click to expand...
I don't know enough about AI to really comment, I'll be honest, the closest I came was the grammarly rewrite tool before I realised that's AI, but I've seen so many people unfairly accused, usually ESL people or neurodivergent, that I have a hard time believing anyone can accurately detect these things. And she's on tumblr, so that doesn't help.

I will say 400k does seem too long for an AI to hold though, I heard something about them hallucinating if you get it to write more than 10k or something but I think that was in the early days of ai. Do they hallucinate still? Is the length realistic for ai? Genuinely curious (i know this is technically off topic but I feel it's relevant)
 
Ultimately, there isn't really any good high confidence way to completely eliminate the possibility that a text is LLM produced that can't be spoofed. All the "quirks" people point out exist because they common quirks of human generated text. If your LLM did produce text that could be easily and conclusively detected than your LLM is a abject failure and fundamentally unfit for the base purpose it was designed for.
 
That is why I'm hesitant to say that this author is undoubtedly using AI as AI is based off of real writing. A lot of modern authors were or still are fanfiction authors, which was also scraped and stolen from. This author is a fanfiction author. I have also suspected AI use in IFs and online but instead of voicing it publically or using an AI detection tool (those suck, and are also AI), I don't interact with their content anymore. AI art is much easier to detect than written word and a lot of the "signs" towards AI writing are almost always natural forms of human expression that people scraped for their models. I've seen too many authors be accused of AI just because they didn't hire an editor and or because people think mistakes or things unique to authors have to be from AI use.
 
Last edited:
Grace5963 said:
I will say 400k does seem too long for an AI to hold though, I heard something about them hallucinating if you get it to write more than 10k or something but I think that was in the early days of ai. Do they hallucinate still? Is the length realistic for ai? Genuinely curious (i know this is technically off topic but I feel it's relevant)
Click to expand...
Nah, 10k is free chatgpt memory. I took a quick peep at the processing capacity of more advanced ones, and it seems that the specialised text-heavy llms can handle around 50-200k words as of right now without hallucinating but with some continuity errors. So if you, let's say, split the whole text into several big blocks and feed the AI a summary of the events + a snippet of text? Should be doable, though you'll have to check for continuity and repetitions. And the rewrite does have strange continuity\parallelism mishaps, e.g. two characters being described holding the glass in an almost same manner but a passage apart. Which could be a result of her writing a placeholder scene, or could be a result of such mitigation.

So just to show you what everyone means, I went to one model I remember having a somewhat similar lingo and basically told it: write me an intro for a reality-show bachelor style thing and listed like 5 qualities I remember about one of the dudes from the game. I think you can see how the style does resemble the IF. Note the Not an offering. A classification. bit in the 3rd passage and the em-dash love. I love my em-dashes because I am a foreigner and they are common in my mother tongue (as are colons and overly long sentences, tbh), but AI surely does love them more. But, yeah, ultimately we can only speculate.

1778793280435.png
 
f8butterfly said:
That is why I'm hesitant to say that this author is undoubtedly using AI as AI is based off of real writing. A lot of modern authors were or still are fanfiction authors, which was also scraped and stolen from. This author is a fanfiction author. I have also suspected AI use in IFs and online but instead of voicing it publically or using an AI dectection tool (those suck, and are also AI), I don't interact with their content anymore. AI art is much easier to detect than written word and a lot of the "signs" towards AI writing are almost always natural forms of human expression that people scraped for their models. I've seen too many authors be accused of AI just because they didn't hire an editor and or because people think mistakes or things unique to authors have to be from AI use.
Click to expand...
I think the difference is that AI is always hyperfixating on the most common tropes and expression forms it detects in a body of texts it was trained on, so if an author also happens to have such hyperfixation... Speaking of editors: at least now when I see dumb typos I know the person wrote the thing themselves lol
 
I fed one of my college papers into a AI detector about a year ago and it told me it was 88% likely to be AI, it was written in 2011 sufficed to say it was all my writing. So yeah I don't put any stock in those AI detectors.
 
isahale said:
I fed one of my college papers into a AI detector about a year ago and it told me it was 88% likely to be AI, it was written in 2011 sufficed to say it was all my writing. So yeah I don't put any stock in those AI detectors.
Click to expand...
Mood, most of my writing from between 2015 and 2020 comes up in the 90% realm of some ai detectors, then others will say 0% so it's like 'you all suck'
 
Ireth313 said:
Nah, 10k is free chatgpt memory. I took a quick peep at the processing capacity of more advanced ones, and it seems that the specialised text-heavy llms can handle around 50-200k words as of right now without hallucinating but with some continuity errors. So if you, let's say, split the whole text into several big blocks and feed the AI a summary of the events + a snippet of text? Should be doable, though you'll have to check for continuity and repetitions. And the rewrite does have strange continuity\parallelism mishaps, e.g. two characters being described holding the glass in an almost same manner but a passage apart. Which could be a result of her writing a placeholder scene, or could be a result of such mitigation.

So just to show you what everyone means, I went to one model I remember having a somewhat similar lingo and basically told it: write me an intro for a reality-show bachelor style thing and listed like 5 qualities I remember about one of the dudes from the game. I think you can see how the style does resemble the IF. Note the Not an offering. A classification. bit in the 3rd passage and the em-dash love. I love my em-dashes because I am a foreigner and they are common in my mother tongue (as are colons and overly long sentences, tbh), but AI surely does love them more. But, yeah, ultimately we can only speculate.

View attachment 38154
Click to expand...
Although this prose seems flat compared to summer of Love, like summer of Love could be AI in theory but didn't read as flat. Overwritten, definitely, and a similar style, but I wouldn't say it was this flat. Also these em-dashes seem more correctly placed 😂
 
xeroxo said:
The rewrite wasn't needed though.
Click to expand...
It really did not need to happen. I wonder if this author saw other authors who were putting out impressive (sometimes over written) chapter sizes and thought they must be doing something wrong because they have smaller chapters. It reminds me of OLNF. Because while I love the amount of content she and her team are creating at GB Patch, its taking too long especially because of the writing style she chose for the game.
 
Grace5963 said:
Mood, most of my writing from between 2015 and 2020 comes up in the 90% realm of some ai detectors, then others will say 0% so it's like 'you all suck'
Click to expand...
I had a paper failed because a teacher ran it through a plagiarism detector back in 2016. And what was I to do? Older people than me who know nothing about technology just assume its always accurate so there was no fighting anything.

Grace5963 said:
Although this prose seems flat compared to summer of Love, like summer of Love could be AI in theory but didn't read as flat. Overwritten, definitely, and a similar style, but I wouldn't say it was this flat. Also these em-dashes seem more correctly placed 😂
Click to expand...
I do think the em dashes are overused and they have that many words because they felt filled page = great writing. Too many em dashes could just mean the author is either obsessed with them or, from what I've heard on tumblr it could be a form of protest as em dashes mean AI to millions of people now.
 
Last edited:
f8butterfly said:
It really did not need to happen. I wonder if this author saw other authors who were putting out impressive (sometimes over written) chapter sizes and thought they must be doing something wrong because they have smaller chapters. It reminds me of OLNF. Because while I love the amount of content she and her team are creating at GB Patch, its taking too long especially because of the writing style she chose for the game.
Click to expand...
The writing style is different though, it would be rather alarming for her to jump in where she left off with a new writing style, especially if she has more understanding and plans to take things in a different direction. I do agree it's overwritten and her reasons are unknown, but I don't think one chapter is enough to judge the reasoning of the rewrite from (but iirc, Izzie didn't intentionally couple with Theo platonically in the original, he just put her in the friendzine, whereas she did it intentionally here).
 
Grace5963 said:
Also these em-dashes seem more correctly placed 😂
Click to expand...
I asked it to be *mindful* of them (meaning: please, use less), but I don't think it worked.

All in all, whatever the reason for the style change, I don't think it was a good creative decision. I was kinda interested to see a slice-of-life-y IF, but I'll have to drop it now because it reminds me too much of the stuff I have to deal with daily, aka suffering in the creative industry (translation & localisation) where now every other client either wants to translate AI-generated slop or wants their human-written texts translated via AI and only mildly edited. And AI is subpar in both writing and translation. Gosh, I cannot wait for the bubble to burst...
 
You must log in or register to reply here.
Back
Top