"When does it stop being AI generated?" Topic

Suppose you start a story with AI but tweak it: when is it no longer AI generated and how would anyone know?

link

Write a paragraph story about two adventures in a fantasy dungeon

In the depths of the mystical dungeon, two adventurers found themselves facing a daunting challenge. The first adventurer was a fierce warrior with a long, curved sword, while the second was a cunning rogue with a knack for traps and secrets. Together, they ventured deeper into the dungeon, fighting off hordes of goblins and dodging traps at every turn.

Experts in AI might very well have a way of detecting a known pattern of word choice. It's their job, after all.

For me, it's when a human being puts his own name to it and to multiple products as a sign that they're his work, and he takes both credit and responsibility. How the human author arrives at that point is immaterial. Some work alone. Some talk with friends. Some steal plots and characters. Some take entire passages. Doesn't matter to me. MY concerns are plot, setting, characterization, dialogue and prose.

Find me an author who can write like Avram Davidson, Leigh Brackett or Rudyard Kipling, and he can plug himself into his own CPU for all I care.

When it comes to determining whether a story is AI-generated or not, there is no clear threshold that can be used to draw a definitive line. This is because AI-generated stories can vary in their level of complexity and quality, depending on the underlying algorithms and data used to create them.

However, there are some indicators that can help you determine whether a story is likely to be AI-generated. For example, if the story contains an unusual or improbable sequence of events that seem disconnected or illogical, it may be a sign that it was generated by an AI model that lacks an understanding of narrative structure and human emotions.

Another indicator is the presence of errors in grammar or syntax that are typical of machine-generated text. AI models may struggle with issues such as tense agreement, word choice, and sentence structure, resulting in writing that feels stilted or unnatural.

Ultimately, the best way to determine whether a story is AI-generated or not is to look for signs of human authorship. This might include subtle nuances in the writing style, references to personal experiences or cultural knowledge, or an overall sense of coherence and emotional depth that is difficult for AI models to replicate.

However, as AI models become more sophisticated and better able to mimic human writing, it may become increasingly difficult to distinguish between human- and machine-generated stories. In such cases, the only way to know for certain may be to ask the author directly.

------
The text above was hallucinated by my AI Buddy.
I call him Chatty, by the way.

Regarding Paizo, in five years they will have to pay a premium for artists not using AI.
And customers to whom they can pass on their relatively increased costs.
If development continues at this rate, they'll be lucky if customers don't have their adventures written entirely by an AI.
Without publishing middlemen like Paizo.

Almost forgot to ask, where does Paizo have its products manufactured?

2023 was the year it finally dawned on most people that the human race is doomed. Not just a literary trope, a sci-fi scenario, or a facetious jest, our self-destruction is all but inevitable and starting to play out now. I knew it was coming, but I had hoped it would be after my time.

"The world of the future will be an ever more demanding struggle against the limitations of our intelligence, not a comfortable hammock in which we can lie down to be waited upon by our robot slaves." Norbert Wiener, God & Golem inc.

Humanity is not doomed. It's a chatbot. It's very good at producing well-formed English text. It's never thinking or understanding. Sometimes you ask it for code and it produces usable code, sometimes it produces gibberish and then insists that it's correct.

Look on boardgamegeek for a guy who tried to get it to play ASL with him. He spent most of the time trying convince the AI that it was capable of doing what he asked.

"General AI", the machine that actually thinks, is a long way off and may not happen.

The danger is when we believe we're talking to someone smart or someone that understands. Or someone at all. It's randomly spitting back stuff.

Say you wrote a story by starting with a deck of cards: there are themes, character archetypes, adjectives, locations. You shuffle, deal some of them out, you write a story. Did you write the story or did the deck of cards write the story? The AI is not qualitatively different, just better.

It's a tool, it's okay to use it. But all machines are amplifiers.

"if the story contains an unusual or improbable sequence of events that seem disconnected or illogical, it may be a sign that it was generated by an AI model that lacks an understanding of narrative structure and human emotions."

So Louise Penny and JD Robb are AIs? Probably Edgar Rice Burroughs, too? Certainly Michael Moorcock.

Feel free to add names. "Improbable sequence of events" is a pretty low bar, and the B&N is full of books which couldn't pass the "narrative structure of human emotions" test.

Andrew I find the story of the guy who tried to get the AI to play ASL hugely funny. I only got to play Squad Leader. Guess ASL is a bridge too far for our AL overlords. Cheers.

I still say the future of AI literature is in niche markets. Dozens of settings and story types are no longer commercially viable. It's go to AI or do without.

@Robert

I have passed on the objection.
Here is what Chatty replied.
You're lucky I didn't ask Sydney in [System] Mode.
:)

Back to Paizo: I wonder how they are going to get their employees, service providers, and so on not to use all these fancy text-to-image functions that are already or just being built into the tools of the trade for artists and designers like Photoshop, Illustrator, InDesign et al?
I mean, I was privately playing around with a Stable Diffusion addon for Blender recently to see if I could use it to improve or speed up the generation of models for 3d printing.
Not to speak of tools like Neuroflash and ResearchRabbit for business and academic work.
A lot of my posts on TMP are post-processed by Deepl and changed about 20% of the time.
Which part of it is mine and which is AI?

So, in my opinion this is a ridiculous publicity stunt.
As was mentioned above AI is a tool, it enhances productivity.
It will be used throughout the whole production process of products that can be digitalized.

Regarding ASL vs ChatGPT on Boardgamegeek, you find more relevant empirical research in that area here:

Mastering the game of Stratego with model-free multiagent reinforcement learning

Human-level play in the game of Diplomacy by combining language models with strategic reasoning

Actually, the fiddly boringness of ASL should be much easier to handle than Diplomacy.
Fortunately, you no longer have to force anyone to enter the rules into the system by hand, you only need to read the pdf into the system.
Maybe someone even created a database of the rules and put it on the web?

Just in case you prefer the discussion about Diplomacy vs AI over at Boardgamegeek:
link
Apparently, they didn't hear that Stratego has fallen too.
At least I couldn't find any proper discussion about it.
:)

I did ask Dall-E to make a "Frazetta style picture of a fighter and a thief in a dungeon"

It went well with my story and since I can't draw, so much the better.

Re Mr Elmo's anecdote:
(1) He didn't learn to draw, he tasked the drawing to an AI app that produced "good enough" results. HUMANS NOT LEARNING THINGS.

(2) He didn't pay another human to draw, he tasked an AI to do it for a fraction of the cost. HUMANS LOSING EMPLOYMENT TO AI.

(3) The AI produced "good enough" results, therefore pushing a future crowding out stories w/out art at all (so use of the tech will be mandated by the marketplace), and yet if the history of food, music, drama, etc are any guide drowning that marketplace in sheer volume of mediocre product. HUMANS LOSING THE OPTION TO KNOW AND EXPERIENCE GREATNESS, AS MARKETPLACE ECONOMICS DRIVE A RUSH TO THE AI-PRODUCABLE MUDDLING MIDDLE.

You folks that think this is "just a productivity enhancer" are failing to note the crucial difference between AI and technological advances of previous centuries. What's being replaced is not muscle-power but mind power, the very essence of what it is to be human. Within a few decades, 90+% of white-collar workers will lose their jobs, along with everyone who drives for a living and most of those who work with their hands.

But go on whistling past the graveyard…

What's being replaced is not muscle-power but mind power, the very essence of what it is to be human.

Oh yes, brothers and sisters, we've seen that with the invention of written language, books, computers, databases, and the internet.
I lively remember when sometime in the early nineties uncountable numbers of artists hanged themselves in the woods and jumped from the roof, if not the shark, because digital tools were invented.
None of them had the idea to use them instead and produce more, cheaper, better, more complex products and art…
And before that all those human computers, yeah they were called that because they … computed.
Long, tedious, and complicated calculations, like say the women that brought Man to the Moon.
With their heads, pen, and paper.
Their very essence.
All dead and gone.

And when I study the sports news, the weather and the stock reports, I remember with nostalgia that barely ten years ago they were still being written by humans and not by automated systems.
At least they are still as boring and meaningless as they were back then.

And yes, people don't learn anything anymore because it's made too easy for them.
I see this in myself every day.
Rather than analyzing tons of data in machine language or even better with pen and paper I write Python scripts with the help of Github Copilot…

When I think about it, I have much more and interesting work than ever before!

ugh, reading that AI-generated response was like fingernails-on-a-chalkboard for the eyes. It's just… wordy, like a term paper.
(If an AI copies it's responses from term papers, is it plagiarism…? ;-)