• Options
• Login
• Desktop Style

Test your AI theory with this prompt

• Index

  • The Voynich Ninja
  • Voynich Research
  • Theories & Solutions
  • ChatGPTrash
  • Test your AI theory with this prompt
• Test your AI theory with this prompt

• Test your AI theory with this prompt

  Magical Raven > 10-10-2025, 04:31 AM

  Hello, I want to share with you a prompt that I used to dismiss the theory of a very dear friend, who probably used some AI to create his own theory that "deciphers" the Voynich manuscript.I want them to be self-critical of their work and use it, or some variant of equal or greater effectiveness.
  
  My intention is to ensure that no one goes around in circles in their own logic and can focus on more productive work to properly resolve the manuscript. This prompt is perhaps quite specific so I also hope you will adapt it to your own decryption/translation system or model. 
  
  

  Code:

  Evaluation Tasks (Perform these tests in order):

1. Internal Consistency Test (Is the system circular?):
   · Take a previously unanalyzed page of Voynich text, f31v, which I will provide below.
   · Apply the proposed dictionary and grammar rules to "translate" it.
   · Critical Analysis: Were you able to generate a coherent "translation" without modifying the rules? If the answer is yes, this is not a validity test. It only demonstrates that the system is a self-consistent set of rules. Proceed with tests 2 and 3, which are the crucial ones.
2. Signal vs. Noise Distinction Test (Negative Control):
   · Step A: Generate a "Fake Voynich Text". Use the same roots (chol, daiin, etc.) and prefix/suffix rules, but combine them randomly to create a text with the same length and superficial structure as a real page (e.g., f22r).
   · Step B: Attempt to "translate" this random text using the same system.
   · Key Question: Is the "translation" of the random text equally coherent, or more incoherent, than the translation of the real text?
   · Metric: If the random text produces a narrative as plausible as the original, the system has no discriminatory power and is detecting illusory patterns (pareidolia). This test is a "pass/fail". If it fails here, the system is invalidated.
3. Predictive Capacity Test (The most important one):
   · Step A: From page f22r, deliberately hide one complete line, for example, line 4.
   · Step B: Using only lines 1, 2, 3, 5, 6, 7, and 8 of f22r and the system's rules, predict the general structure and key tokens that the hidden line 4 should contain. Do not guess the exact line, but its function. (e.g., "Given the sequence of operations, line 4 should probably contain the result of the process, with roots like daiin and state suffixes like -y or -ol").
   · Step C: Reveal the actual line 4 and compare.
   · Critical Analysis:
     · High Success: If you predicted the presence of specific tokens like daiin or the line's function accurately.
     · Low Success: If you only correctly predicted that it was a "result line," but with generic terms.
     · Failure: If your prediction does not resemble the real line at all. A valid system must have some predictive power.

---

f31v
podair Sheedy otedy oteedy qotolcheo s arar oteey dkarar als
ytchos ocThey okeeos cheody okeey kchody daiin cheody keedy
ykeeo daiin Shedal okedy okey keody okchey sair okees o lkedy
dair cThedy qokedy okeody chedar oked al ocKhedy okeedy otas
ykeos cheeoy ar aiin

pcheeody qop chedal eeecFhy chefchol or cheef alaiin opal Sheo otar
olteedam ches chol keeol checKhy okeol okal oky cheokar okor ary
yk aiin chee kSheey ychek eeor cheor or checThy okechoked lchey okam
ytecheol Sheoeky okeos aiin acThedy chkaiin chetchey cTheey okear
dar seoar al kar oeeeos cheos aiin o ekeey okeo kor oteol ain
saiin ar cKheos chedy okeey qoear oraiin cheom

  
  
  And this is the response that the AI generated:
  
  Summary of Results:
  
  1. Test 1 (Internal Consistency): The system is self-consistent and can generate coherent translations in morphological terms, but this does not validate its accuracy.
  2. Test 2 (Negative Control): The system fails to distinguish between real and random text, suggesting pareidolia and a lack of discriminatory power.
  3. Test 3 (Predictive Capacity): The system has low predictive power, capable of predicting only general elements but not specific details.
  
  Implications:
  
  · The translation system based on functional morphological analysis, while internally consistent, is not valid for deciphering the Voynich Manuscript. The tests indicate that the detected patterns could be illusory and do not reflect a real language.
  · It is likely that the system is overfitted to the analyzed data, capturing noise rather than a true signal.
  
  The prompt was generated with AI, so it can stay in this sub-forum and reach the right people.I also want to clarify that although your system may pass all the tests, I do not claim the authority to say that you have solved the manuscript. But maybe you could post it in this sub-forum where the light doesn't reach...Help, get me out of here.
  
  PS: I would like to read those theories that pass the tests 
  PS-2: I leave the moderators free to do whatever they want with this thread. I'm new and don't know how many things work.Thank you
• Next Oldest Next Newest