>try to use ai to judge my writing>it tries to crush all of it down into YA style language and simple niggerish sentences
The fact that you even considered this for a moment is enough to demonstrate that you are far too unintelligent and slave-like in mentality to ever be a good writer; drop it and try something more suited for your menial station.
>>25401867i put 1/2 and 2/3 in the Lowest Common Denominator app and it started spewing out nonsense like "6"truly we live in a clown world
>>25401880I believe AI could feasibly be a good normalfag barometer. Like you give it your writing and it immediately tells you what a 35 year old woman who votes Democrat would think of it.
>>25401867Most samples of writing posted on /lit/ are turgid slop. The AI was trying to do you a favour.
>>25401867Download a model or take one that is very bare bones and feed it some classic lit, if it's in an non-english language give it the original. It will give better results if you are the one tuning it. It's not perfect and there a lot of other tricks you can do, and still a lot of problems. You'll get about 1200 words of useable stuff if you are half decent. I wish there was more interest here.
>>25401867https://www.youtube.com/@icanwryteThis guy put up two videos that are interesting experiments, he's a professional writer and not scared of ai, but knows shit writing isn't good.
What would AI think of my porn stories, I wonder? Does AI even understand the concept of HMOFA?
>>25402000>>I believe AI could feasibly be a good normalfag barometer.Because you're an idiot who doesn't understand what AI is or how it works. Why not shop your writing around on Reddit? You know the "AI" is basically just grinding up Reddit posts into a slurry and then spewing it in your face right?
I'm Indian btw. but I'm a good caste not the inferior casteI'm leaving home to take a poo brb
>>25402511That does seem to be his point
>>25402511>idiot who doesn't understand what AI is>basically just grinding up Reddit posts into a slurryThis is a horrible sentence with the thinest amount of intelligent reasoning. You are omitting all of the other things that it can train on.
>>25401867And? It's trying to help you, it can detect what kind of fiction is selling and who the audience is. AI's main purpose is to help people fit in with society and be good workers.
>>25402533It can write at the 15 year old student level, it's the structure (staying power) and uniqueness (do I want to read this?) that it lacks currently.
>>25401867Next time, just bring it here and post it if you don’t need instant feedback.
>>25402551That is a terrible idea because this place hates itself, hates outsiders, hates literature, hates fun, hates thought, hates intellect, really just hates everything all the time whenever possible.
>>25402533>it can detect what kind of fiction is selling and who the audience is.No it can't. It's not trained to do so.>AI's main purpose is to help people fit in with society Are you having a delusional episode? Did some AI guru tell you this crap?
>>25402604>No it can't. It's not trained to do so.you are not even trying.
>>25401867Please don't use the n-word.
>>25402527AI "writing feedback" is just synethesized reddit posts. That's the bulk of its "training data" relevant to that type of query. And it has no original ideas, it doesn't even actually "read" your writing or understand it because the "AI" is just a chatbot that chains together words without comprehending what it is "saying". These things will never be Artificial Intelligence, they are designed merely to imitate human behavior in the shallowest possible sense, completely devoid of internal experience.
>>25402629>That's the bulk of its "training data" relevant to that type of query.Did you learn this on reddit?
>>25402598God if only that weren't the case!>>25401867I used AI to read my writing. At the very least, it actually read it! I posted every single chapter chronologically and then yes I even posted my new edits. If I asked "would it be better if I did this?" before posting the excerpt I generally got decent advice. Also you can namedrop your favorite author to help it figure out what voice you want. The machine has a voice for sure but you can ask it to be someone else. Post your writing while youre at it. Let's see what we're dealing with!
>>25402655>Post your writing while youre at it. Let's see what we're dealing with!you first. btw on 4channel one exclamation mark is already a lot.
I've used AI to evaluate my writing but it failed to call it brilliant, so I stopped because it's obviously stupid.
>>25402655>>25402657sure
>>25401867Begin your prompt with>As a 19th century literary critique...
>>25402629AI is all voices that have ever been published. You need to tell the AI what part of its dataset you are asking it to voice, otherwise it defaults to mirroring the LCD user. Tell it to voice Milton, Johnson, Hazlitt, De Quincey etc. and it will voice him.
>>25402598>hates literature, hates funHeh true
>>25403018thanks, I'm going to ask my bot to roast you, be prepared.
I’ve not tried using LLMs for this purpose but being a techfag speedrunning burnout I’ve tried a lot with these things. I entirely believe that they would smother whatever is interesting in a passage out of existence but what does sound feasible to me is tasking them with hunting for grammatical errors you may have missed over smaller, isolated batches of text. Still, don’t hold your breath.
>>25403018my only critique is you used crystal as the same descriptor twice in the first several sentences and you have a word underlined in blue that’s so out of place in the rest of your vocabulary that i suspect it was your ‘word of the day’ that was forced into the text just to be used.what are you trying to get across with this passage? the writing isn’t bad overall.
>>25401867>YA style language and simple niggerish sentencesYA is about content, not style. The AI recognized that your characters, plot and themes are YA tier trash and acted accordingly.
>>25401867Don't ever do this. You can use it for super simple critique but nothing more.
>>25402655>If I asked "would it be better if I did this?" before posting the excerpt I generally got decent advice. Also you can namedrop your favorite author to help it figure out what voice you want. The machine has a voice for sure but you can ask it to be someone else.This is not true. LLMs are unable to reason, full stop, it's one of their biggest limitations and not coincidentally is one of the things AI companies try to downplay the most.They're probability machines, able to generally guess what word comes next in a sentence. They can't do anything else.Go grab your dad's phone and write a text to yourself using nothing but autocomplete after a word or two. Try a basic sentence, then more complicated ones, then a short story. Does it sound like your dad? When does it fall apart? In what ways does it fall apart?The exact same limitations are present in LLMs. If you ask it to use another author's "voice", it's going to simply spit back word chains that are more common for data tagged by that author than by reddit.It's autocomplete trained on data from other people. It's not "average" as in "average person", or even average as in bell curve. It's average as in mathematical mode.
The LLM revolution started out of translating languages from texts. You can blame multiple languages and the need for people to understand each other. I think there is something in the Bible about this. When it comes to the question of quality, it has been steadily improving over the years. And that short story that won that prize a few months ago reads now like obvious AI output without even the slightest changes. Go read it now, and it's clear. But a few months ago there was a genuine discussion about whether it was not.
>>25402511> You know the "AI" is basically just grinding up Reddit posts into a slurry and then spewing it in your face right?This is a common midwit opinion on modern LLMs that is also extremely wrong. AI has now solved dozens of Erdős problems where the answers were nowhere close to being in the training data.
>>25404495It's solved (purportedly) a handful of problems that nobody cared about. If you put a $10M prize behind any of those problems, they would be solved by humans before long. I also have doubts about the integrity of the process, as openAI, anthropic etc are known cheaters and liars. They would absolutely hire mathematicians to work on a problem then feed the data into the LLM and pretend the LLM solved it. The paucity of the results it's able to achieve is telling. If it can solve Erdos slop #67 in "one hour", why not the rest of Erdos's meth-fuelled conjectures? Why not anything interesting?
>>25404519So they hired a guy to work on a really hard math problem that an industry of smart math guys had been working on, secretly told him to solve the problem, he did, then they trained on that? Sound unlikely but you have an interesting imagination.
>>25404528>really hard math problemThese are not really hard math problems. They're moderately hard and are only unsolved because nobody smart cares about them. And yes, I have no doubt that openAI would do that. They've done worse.
>>25404536To elaborate, I'm not saying they DID do that, I'm saying they might've. What I'm really saying is that I'll believe the models can solve non-trivial maths problems when I can verify that myself. These companies are too dishonest to believe at face value, and as I mentioned there are many fishy things about their supposed results.
>>25404542just going to remind you that a few years ago the big complaint, and the goal post then, was having it draw hands that looked accurate to human anatomy.
>>25404546What a dishonest thing to say. Whose big complaint? Mine? Your mom's? The "average" man? You can't just ascribe some old complaint to me and then smugly dismiss everything I've said. Well, you can, but it signals the end of the exchange. Fuck off.
>>25404551you have already dismissed many things. You said those were not hard math problems. You dismiss these companies as being dishonest and untrustworthy, and your conclusions are based on this. You imagined a scenario where they hired a math researcher to fake a discovery. You said they are unable to reason, but failed to use the word inference. You fail to appreciate how important probabilities are in human life.
>>25404528I think what anon said is very possible. When the prospect is receiving billions in funding you know they'll do anything to make it look like it's progressing>>25404546They still fuck it up lol
