About That AI Assist In Writing

[Euphony]

I also had an 8" long orange Donkey Kong when I was a teenager.

I was, am, and will forever be...



Truly Outrageous.

 

[StillStunned]

Talk gets any hotter up in here and somebody owes somebody $4.99 a minute.

View attachment 2294348

Why don't we have a *drool* emoji?

 

[ElectricBlue]

Talk gets any hotter up in here and somebody owes somebody $4.99 a minute.

Whoa! You're a filthy fucker!

Does anyone remember Larry the Lounge Lizard (something like that) a little pixelated character you had to navigate through a hookers bar? We played it endlessly during lunch breaks, back in 1988.

 

[StillStunned]

Leisure Suit Larry! Quite raunchy for the mind-1980s.

 

[Euphony]

Whoa! You're a filthy fucker!

Does anyone remember Larry the Lounge Lizard (something like that) a little pixelated character you had to navigate through a hookers bar? We played it endlessly during lunch breaks, back in 1988.

Is that Leisure Suit Larry?

I'm not as decrepit so I'm more familiar with the Xbox reboot than the O.G.

 

[nice90sguy]

Ok, OLD MAN here: Back when I was at Berkeley, man, there wuz this arcade man, on Telegraph ave dude, where all the stoner computer dudes would try out their latest games.
I played a vector graphic game, aprecursor to asteroids, where you had to prevent spermatozoa fertilising an egg. Can't google it. And no, I wasn't just stoned and hallucianted the whole thing. That game did exist dammit, but my Google-foo isn't up to finding it. This was 1979.

 

[ElectricBlue]

Is that Leisure Suit Larry?

I'm not as decrepit so I'm more familiar with the Xbox reboot than the O.G.

View attachment 2294359

That's the one, but more like the original Pac-Man. No smooth renders at all. You could count the pixels. The only colour was in the characters, the rest of the screen was mostly black.

This was back when a 286 processor was considered state of the art.

 

[Bramblethorn]

No, my understanding is that ChatGPT doesn't do erotic or sexual at all.

It's not supposed to, but it can. ChatGPT is a combination of a relatively smart writing program which knows how to write erotica, and some relatively dumb filters meant to stop humans from accessing that capability.

https://forum.literotica.com/threads/story-sent-back-due-to-ai.1599503/page-2#post-97924729

 

[nice90sguy]

That makes me REALLY want to work at OpenAI now.

 

[MillieDynamite]

Oh, please, don't go there.

That makes me REALLY want to work at OpenAI now.

 

[nice90sguy]

Oh, please, don't go there.

I promise to lower their market value considerably if I do work there.

 

[EmilyMiller]

No, my understanding is that ChatGPT doesn't do erotic or sexual at all. It can be flirty but will skip over the event like old TV shows of the 50s and 60s. A few sites, with paid subscriptions might do sex, but they wouldn't do it well.

I asked it to write an erotic story in the style of Emily Miller of Literotica. It complained first about aping established authors (me? who knew?) and then about adult content.

Em

 

[NoTalentHack]

And this is why I ignore grammarly two thirds of the time. I have two rules that it breaks:

1. People DO NOT speak with perfect grammar. You can excuse it if they have a prepared speech before them, if they're reading from a prompt. When people talk with perfect grammar you feel unsettled because they feel less like people.
2. It can't get NUANCE... or the idea of other entities, it can't contain originality. I had to ignore its corrections quite often because it just did not get the message. It can't get that "The Genesis" does not refer to the biblical event, but to a SHIP, it can't get that a world is being used interchangeably with "planet", so when I say "on this world" it wants me to correct it to "in this world". Bruh, they ain't living underground.

But, often... it can give me some corrections on grammar. And much more often I tell it "you are wrong, robot".

Huh, and if what you say is correct, then I have been avoiding the AI detector because they would mark down my works with like 500 mistakes even though a person will see them as perfectly okay. If you OVERuse it and remove all of the "own", "actually" and take all its corrections... honestly, you deserve it. I have one GREAT example of why AI can be utterly wrong.

It wanted me to change those to "two fingers shoved into the oil". Truly, "they turned their thinking over to machines but that only allowed men with machines to rule over them".

Same. Over in the other AI thread (no, the other one. No, the other other one), I gave a pretty detailed rundown of how I use the tools, which largely ends up being "thanks for the advice, now STFU." You can see it in detail at https://forum.literotica.com/threads/ai-allegations-thread.1599778/post-97969898 if anyone's curious.

 

[Bramblethorn]

I've reviewed your post. It's a rather subtle description, and you managed to bypass the filtering by steering clear of explicit words that would raise alarms.

"He thrust his eggplant into her flower, the forceful penetration echoing in the room."

Not sure I'd call that "subtle description", even with the obvious word substitutions.

You also acknowledged flooding the machine with numerous prompts,

I didn't use the word "flooding". I'm not sure exactly what you mean by it, so I'm not sure whether it's accurate to say that I "acknowledge flooding the machine".

indicating it wasn't genuinely generated by AI.

This, I do very much dispute. Obviously it's produced by AI under human guidance, and is going to include some of that guidance, but a large amount of what it produced is coming from its training material and not from me.

Essentially, you used the AI as a keyboard to compose your content.

Nope. As I acknowledged in that post, it took some prompting to get GPT into a state where it'd write porn for me; in the first four examples, about 20% of the words in GPT's scenes are echoing things in prompts. But even there, the other 80% (including a good deal of the sexual content) is GPT's wording. Things like "thrusting" and "forceful penetration" are GPT's words, not mine; in other experiments it's produced things like "his eggplant, now softening, lay against his leg", "his eggplant released its essence in a torrent of warmth", "with a shudder, [male protag] spilled his essence", and so on. Those expressions aren't coming from my prompts, they're coming from the porn it's been trained on.

(GPT seems to like "essence". It also loves "dance of desire"/"ballet of passion", electric atmospheres, and dudes whose fingers leave trails down women's bodies.)

The first scene isn't really the point, though. Once it's written that scene, it then takes far less prompting to get it writing other explicit content, including elements not given in any of the prompts. For instance, I can tell it "include [male protag]'s sister in this scene" and it'll oblige, with a bunch of incest tropes that I didn't teach it.

I don't see that as merely "using the AI as a keyboard". Do you?

 

[Bramblethorn]

I have one GREAT example of why AI can be utterly wrong.

It wanted me to change those to "two fingers shoved into the oil". Truly, "they turned their thinking over to machines but that only allowed men with machines to rule over them".

The AI has a point here. Not that its proposed solution is right (it isn't) but it's flagging a weakness in the writing.

People parse as they read, starting to interpret each sentence well before they've reached its end. But sometimes the beginning of a sentence is misleading or ambiguous, and then the reader may head off on an incorrect interpretation before reaching something that tells them they've misunderstood. Then they have to go back, forget their ideas about what the sentence was saying, and try again. That kind of stumble is better averted.

In this case, because "oil" can be either a noun or a verb and because authors don't always get it right on things like "in to"/"into" or "alot"/"a lot", "two fingers shoved in to oil" can be ambiguous. It'd be better to avoid that potential stumbling stone by recasting the sentence.

It's not a big thing, but it's worth trying to avoid it when you can.

For instance, unless there's some reason for using passive voice here, it could be changed to "[PersonName] retrieved the bottle of lube from their nightstand to apply liberally on cock and asshole, and shoved in two fingers to oil the inside as well." Alternately, changing "oil" to "lubricate" or just something like "take care" would work here.

 

[MillieDynamite]

Should read, "shoved inside to oil as well." Or that's how I would've written it were I writing it. Which I didn't.

The AI has a point here. Not that its proposed solution is right (it isn't) but it's flagging a weakness in the writing.

People parse as they read, starting to interpret each sentence well before they've reached its end. But sometimes the beginning of a sentence is misleading or ambiguous, and then the reader may head off on an incorrect interpretation before reaching something that tells them they've misunderstood. Then they have to go back, forget their ideas about what the sentence was saying, and try again. That kind of stumble is better averted.

In this case, because "oil" can be either a noun or a verb and because authors don't always get it right on things like "in to"/"into" or "alot"/"a lot", "two fingers shoved in to oil" can be ambiguous. It'd be better to avoid that potential stumbling stone by recasting the sentence.

It's not a big thing, but it's worth trying to avoid it when you can.

For instance, unless there's some reason for using passive voice here, it could be changed to "[PersonName] retrieved the bottle of lube from their nightstand to apply liberally on cock and asshole, and shoved in two fingers to oil the inside as well." Alternately, changing "oil" to "lubricate" or just something like "take care" would work here.

 

[MillieDynamite]

Then why not write it yourself? That way, it's your writing and not just your prompting.

Of course, I do. It's as clear as daylight, but you're never wrong, are you?

The percentage of words is irrelevant. You don't credit the dictionary for the words you find in it. The real question is: Who's telling the story?

For instance, if you give a single prompt asking for a 1500 word short story about some topic, the output will be considered AI-generated. But it will likely be far from intelligent, so you'll have to prompt it again and again until it gets closer to what you want. I know how hard it is to coax the desired result from these stubborn machines, but you learn to tame them.

Somewhere after the 20th prompt, it becomes your content!

 

[MillieDynamite]

AI is never your work. It's work produced for you so you needn't bother with real thinking and being creative. AI doesn't write like humans. If your smart enough to figure out the prompts, write from then yourself. I can't stress this enough, you're not writing the words, it won't have a real personality to the stories. I have written with pens, pencils on paper. I sometimes still do. I write ideas down and stew on them. You're more of programmer than writer. And in case you haven't figured it out, you are not allowed to publish the type of work your suggesting on this site.

For the same reason, you don't write with a quill on papyrus paper and send your stories with a carrier pigeon. It's a new era, and we're still adjusting.

If you were to go back in time and offer Ovid the use of your iPhone for writing, he'd look at you like you've lost your mind.

 

[Paul_Chance]

Noam Chomsky called AI writing "high tech plagiarism".

AI can never be creative. It can never create anything new or original. That lack of creativity is inherent in its design and it's inseparable from it's design.

It can only generate works by reviewing already existing works, then reassembling them like Frankenstein's Monster, selecting bits and pieces from the graveyards of dead writers and resurrecting them in frightening conglomerations that imitate creativity, but are not.

 

[ronde]

This topic has been discussed to death, but unless I missed it, no one has ever said why they would even consider using AI to write a story here. Writing is a creative art, just like painting or sculpture and artists should take pride that the art they produce is a reflection of their imagination and talent. I've done a lot of programming over the years, and I can't believe anyone would consider what a computer program does as imaginative and talented. The program itself is the creative art of the programmer and is a reflection of his or her imagination and talent. The actions that program can do are not, not anymore than a cell phone picture of the Mona Lisa qualifies as art.

 

[Rustyoznail]

This is the original Larry

 

[Bramblethorn]

Of course, I do. It's as clear as daylight, but you're never wrong, are you?

When describing my own experiences? To a stranger on the internet who hasn't seen the prompts I used and is speculating blindly about how those prompts relate to the outputs I achieved? Yeah, I'm pretty comfortable in accepting my own assessment there, in the same kind of way that I'm going to trust myself as an authority on what I had for breakfast.

The percentage of words is irrelevant. You don't credit the dictionary for the words you find in it. The real question is: Who's telling the story?

We're straying some way off topic here. The point of the post you're arguing with was not to claim that GPT can write an entire story unassisted - that's not an argument I'd make - but to show that it's capable of producing explicit content which isn't merely echoing its inputs as a "keyboard" would do.

For instance, if you give a single prompt asking for a 1500 word short story about some topic, the output will be considered AI-generated. But it will likely be far from intelligent, so you'll have to prompt it again and again until it gets closer to what you want.

Somewhere after the 20th prompt, it becomes your content!

It only took me two or three prompts to get to the content I quoted...

 

[StillStunned]

This is the original Larry
View attachment 2294775

Is that what it looked like in colour? Back in the day we only had a monochrome monitor at home.

I never got very far with Larry. SpaceQuest, on the other hand, I played all the way through. Can't remember which number - SQ 4, I think.

 

[Emilymcplugger]

There was an experiment Patrick H Willems did in asking an AI model to write a film essay for his channel and, to be frank, it was shit.

Patrick described it as being like a kid’s homework who hadn’t studied anything and was just picking up random generic crap off the internet to fool their teachers.

No insight
No intelligence
It was to writing what Angel Delight is to desserts.

The only reason those with power want it is to feed the algorithm so they can replace creativity with machines, but machines aren’t people, you need to be able to feel to move others’ emotions.

Signed. Bryan Chatttering-Golf-Papa-Tango.

 

[MillieDynamite]

@Dybbuk seems to be Tilan reborn.