A new hypothesis: The Voynich Manuscript as a Slavic-based macaronic mixed language

[tavie]

[Edit -wrote this before seeing above post]

Hi BadBigX.  Welcome to the forum.  Thanks for acknowledging the use of AI and asking for feedback.  Please read this You are not allowed to view links. Register or Login to view. about why AI is currently producing garbage in relation to the Voynich Manuscript, and what our forum policy is regarding this.   

ChatGPT does not scan folios, assess images, and compare words on a folio to that image.  Instead it makes stuff up to keep the user engaged. So far, we've seen two key types of hallucination: 

1. it either tells a user that their brilliant idea leads to the solution of the manuscript and then creates fake Voynichese to "prove" it
   
2. or it tells the user that their brilliant idea has revealed undiscovered patterns in the Voynich manuscript.  These patterns are made up.  Neither the AI nor the user is usually able to describe them properly, let alone prove their existence.
   

Your patterns are at least describable.  And indeed falsifiable.  We can look at each herbal folio ourselves and judge whether the assignments of leaf-dominant, flower dominant, etc, are utterly incorrect or too subjective to support the theory.  

We can also check if the words you have given actually line up with the category of folio.  Unfortunately:

• folio 1r is not a plant page.  This is becoming a signature of AI slop. 
  
• The three words you associate with"leaf dominant" folios are in none of the folios you assign to be leaf dominant.  
  
• For the three words you associate with "root dominant" folios, two of them do not feature in any of the folios you assign to be root dominant.  The only one that does is /otol/ and this features once in two folios each. 
  
• For the three words you associate with "flower dominant" folios, two of them do not feature in any of the folios you assign to be flower dominant.  The only one that does is /okar/, and that is once in one folio.
  

You have a bit better luck with the two folios assigned as mixed, but we are still talking only one or two mentions when these particular clusters like /otal/ occur 300 times in the manuscript.

That's enough to be indicating that this is ChatGPT producing slop. You seem really open to feedback and don't seem wedded to this idea (often we get people arriving insisting they are 100% correct and only want validation) so I've kept this open temporarily for the moment.  But it will have to be moved to the Chat GPT prison and locked unless you're able to show me why I'm wrong about the above.

Please do keep contributing but I'd suggest either staying away from ChatGPT, or if you must use AI, try to test what it is producing by comparing its claims to the manuscript's text.