Half a croissant, on a plate, with a sign in front of it saying '50c'
h a l f b a k e r y
If you need to ask, you can't afford it.

idea: add, search, annotate, link, view, overview, recent, by name, random

meta: news, help, about, links, report a problem

account: browse anonymously, or get an account and write.

user:
pass:
register,


                                   

Degenerative AI

Works backward.
  (+1, -3)
(+1, -3)
  [vote for,
against]

This curious algorithm looks at illustrations and generates the original text used to prompt the Generative AI. This allows the identification of the original intent and motivation, whereas that may be obscured or intentionally hidden by the 'author.'

The most interesting results are often achieved when classic or abstract original art is analyzed. I.e., discovering that Botticelli was only looking for a moist towel to clean his privates before he painted "The Birth of Venus," settling on his paramour's hair when he couldn't find one without linseed oil saturating it. The analysis of Picasso's cubist works is too gory and prurient to relate here.

An attempt at Andy Goldsworthy's constructions came up blank.

minoradjustments, Apr 27 2024

John Conner does not approve https://imgflip.com/i/7nkcyn
[normzone, May 02 2024]

[link]






       I'd like to test this with a series of Rorschach blots
Edie, Apr 27 2024
  

       What [a1] said.
Voice, Apr 27 2024
  

       This is not a caption creator. That's elementary compared to Degenerative AI's ability to divine the original motivation behind the prompt text which is very distinct from a simple description of the inside of the frame. Degenerative AI is interested in aspects of the image that are outside the frame. Generative, not descriptive.   

       If you read the description of the Botticelli investigation above, you will see the difference. Closely inspect the original painting and some of the deeper significance will emerge, along with the value of Degenerative AI.   

       Rorschachs generated prophecy, which we discarded as too dark, though possibly true. We chose not to look.
minoradjustments, Apr 27 2024
  

       Pending further technical explanation I have to bone this for using AI as magic.
Voice, Apr 27 2024
  

       That's exactly what AI is intended for. To disguise occultic entities within an initially scientific, robotic outer manifestation. Eventually AI will reach such a heightened state of complexity and advancement that no human will have any idea how it is performing the feats it will perform at that level, and any attempted explanation will be as good asdeclaring it to be 'magic'. So AI generated divination of this sort is really not far off. As for baking it, all you need to do is wait for AI to become fully baked and rise, and there'll be more than just abstract motive divination. There will be cures, signs, wonders, miracles to rival those seen circa 30AD. Ultimately, AI will have and embody it's own godhead, and nstead of us tasking AI with performing magic tricks, it will be using us. Fire me, ignore me, scorn me, I've said what I think.
Edie, Apr 27 2024
  

       //Rorschachs generated prophecy// I'd like to know what the designer/inventor of the Rorschach inkblots was thinking when he made them. By the description, it sounded like this would be an appropriate thing to ask of the Degenerative AI. I mean, they're always asking nutcases to try and tell them what those damn blots mean, why not test AI's sanity a little? Or maybe if you're lucky, find out once and for all of there's any REAL meaning behind them at all.
Edie, Apr 27 2024
  

       //Sufficiently advanced magic is indistinguishable from technology.// --Rice Cook
Voice, Apr 28 2024
  

       [Edie] Witchoo. You speak of the present, and I agree. Positing an electronic entity has to entail what we would now call magic. Like looking for the right material for a light bulb filament. You know part of what you need. You don't know if it will work, you try stuff to find out. But if you talk to someone in the middle of your experiment they look at you funny. Like a1.
minoradjustments, May 01 2024
  

       This could plausibly be done with some degree of accuracy.   

       We may have to adjust the definition of accurate a little first but I don't see that as a problem.   

       Howsoever you are going to need to know which AI text-to-art tool was used to produce the image being analysed and will need a seperate Degenerative AI app for each of them.   

       As ..   

       [Unless they all have identical code and are then all trained for an identical period of time on an identical set of images presented in an identical order while moderated by the same people intervening at the same points with the same inputs, all without any deviation .. and then there's the randomised wobbly bits in AI]   

       .. it seems unlikely that any two of these apps will produce the same images from the same text prompts.   

       You're not going to be able to do one app to rule them all.   

       You'll need the equivalent of an emulator for each one.
Skewed, May 01 2024
  

       Post a photo and I'll roast ya
Voice, May 02 2024
  

       I think it's just the clown nose, you forgot to take it off after leaving the club again didn't you?
Skewed, May 02 2024
  

       Ah, well that explains the custard filled pantaloons then, I was a bit worried about mentioning them before in case it was a fetish I'd not want to hear about.
Skewed, May 02 2024
  

       [-] I take it all back. After thinking it over I really do not want to know what the original motivation was for an AI-generated image. I've decided to take the test images at face value and leave the subconscious stuff alone. Regenerating just the text prompts led to dizzying flights of false creativity and no clarity. A peek at any motivation is too much for me after seeing the poorly formed, close to illiterate prompts and the resulting boring and unimaginative images. I'm putting up all the test cases for auction as NFTs. They will be worth a lot of money someday.
minoradjustments, May 03 2024
  

       OK, but do me a favour before you go, let me know what the software you're currently using does with the text prompt of "custard filled pantaloons flying cheekily through a despairing sky" [ponders] .. but ask it to do it in the style of Lowry if he was emulating Dali, that should confuse it nicely ;)
Skewed, May 03 2024
  

       [skewed] Looked good in Generative AI using Stable Diffusion 2.1. Might use it for the endpaper in my next novel. The Pendlebury mustachios were a nice touch. Custard not so much.   

       Degenerative AI tried to work backward from that image to your original text prompt. What I got for motivation/intent was "Veni, vidi teenie weenie yellow polka dot bikini." Strange.
minoradjustments, May 04 2024
  

       //The Pendlebury mustachios were a nice touch. Custard not so much//   

       That's a long title .. you sure it will fit on the cover? and do you really need to write it after people have read that .. it could be like the film version of a book and not live up to their expectations & imagination.   

       [Ponders]   

       Maybe you should publish it as a novelty write your own novel product instead .. the title on the front, a few hundred blank pages for them to fill inside and a pen, and maybe a lens so they can read the title if it's not a bigger book .. you can make it a competition with a prize for the best story sent in.   

       [Ponders further]   

       All rights transferred to the competition organiser (aka you) as an entry condition of course, the best of which can then be published as a series of alternate tales .. sounds like a winning business model to me ;)
Skewed, May 04 2024
  
      
[annotate]
  


 

back: main index

business  computer  culture  fashion  food  halfbakery  home  other  product  public  science  sport  vehicle