Quote from: TDoS on Apr 13, 2025, 08:05 AM

Quote from: K-Dog on Apr 13, 2025, 07:50 AMI wrote it.  I used AI to generate the pics.

For hitting all the hot buttons from a given perspective it was pretty good. But the odds of it being a real event was unlikely, but most people don't calculate the probabilities of a given event based on the underlying odds of occurrence of each component like I do though. But take out that personal quirk and it was well designed to make a point.

Can AI write something similar if given the proper inputs?

The answer to your question is a qualified yes.  AI has a huge store of knowledge.  But for now it takes a human to put realism and a correct timeline into a story like this one.  I had to research dates and events to make the timeline correct.  Given the length of my story, writing proper inputs for AI would have taken longer than writing the story itself.

I started with a plot which I turned into a list and from that I wrote the story paragraphs with care to make sure the timeline comported with reality and with the sliver of true-ness I wanted to present.

A lot of people worry about AI and perhaps soon enough their worries will be valid, but for now AI does not replace coders, it just makes better ones.  This is because AI is not yet goal oriented.  Writers can use it to be better writers, but writers are not replaced because AI is not yet goal oriented. 

I wrote this story with specific intent.  Communicating my intent in this story to a llm (large language model) would require very specific prompts that essentially would write the story.  In that case AI I would only be doing a translation because the plot is not mainstream and everything would have to be specified in detail.

The nature of censorship in AI is such that AI would have given the story a happy ending.  Publicly available llm models have beaucoup censorship built into them which guide them in certain directions.  But you would not know that if you did not start tinkering under the hood as I obviously am.

Try asking a photo generator (also an llm) for a picture of a Tesla with a broken window.  Good fucking luck. 


I wasted an hour doing that yesterday.  You can ask current llm models for a story, but not like this one of mine.  Public AI will not generate politically sensitive material.  If you trick it into doing so, as I have, a built in bot detects it and clears the screen.  It is impressive.  Like Whoah! Then you are in the predicament of realizing you just told a machine to fuck itself.

Quotethe odds of it being a real event was unlikely

America happens to have millions of people in case you did not know.  Somewhere my 'mud thrown at the wall' found a place to stick.  In fact I suspect my mud stuck in a few dozen places. That was my intent.

* It is an interesting and true fact that Diner Text is being collected to be used to train AI right now.  This very post will be sanitized and fed into a training set.  Bots are allays collecting content here.  Bots look for and collect changes.  Somewhere sometime - someone may be accused of using AI, and the fact will be, that AI has stolen from me.