| Welcome, Guest |
You have to register before you can post on our site.
|
| Online Users |
There are currently 876 online users.
» 2 Member(s) | 871 Guest(s)
Applebot, Bing, Google, Clauraslya
|
| Latest Threads |
Would a NEW Voynich Manus...
Forum: Provenance & history
Last Post: Bluetoes101
5 hours ago
» Replies: 9
» Views: 178
|
Just a hoax?
Forum: Theories & Solutions
Last Post: ReneZ
5 hours ago
» Replies: 35
» Views: 825
|
Elephant in the Room Solu...
Forum: Theories & Solutions
Last Post: MHTamdgidi_(Behrooz)
11 hours ago
» Replies: 80
» Views: 4,605
|
“The Library of Babel” by...
Forum: Fiction, Comics, Films & Videos, Games & other Media
Last Post: Koen G
Yesterday, 11:30 AM
» Replies: 1
» Views: 102
|
Has anyone actually made ...
Forum: Analysis of the text
Last Post: Mauro
Yesterday, 09:58 AM
» Replies: 17
» Views: 577
|
Could incantations explai...
Forum: Theories & Solutions
Last Post: JoJo_Jost
Yesterday, 09:42 AM
» Replies: 42
» Views: 2,422
|
Trip to Egypt
Forum: Voynich Talk
Last Post: Koen G
Yesterday, 08:44 AM
» Replies: 1
» Views: 115
|
"Abnormal" words
Forum: Analysis of the text
Last Post: Jorge_Stolfi
Yesterday, 08:28 AM
» Replies: 20
» Views: 2,812
|
[split] Retracer Thread: ...
Forum: Voynich Talk
Last Post: Jorge_Stolfi
Yesterday, 08:21 AM
» Replies: 306
» Views: 40,737
|
THE MASTER REFERENCE GUID...
Forum: The Slop Bucket
Last Post: Ronald Pittman
Yesterday, 02:17 AM
» Replies: 0
» Views: 235
|
|
|
| SIGBOVIC 2023 paper |
|
Posted by: hatoncat - 17-04-2023, 04:32 PM - Forum: Voynich Talk
- Replies (1)
|
 |
My Voynich Cryptography paper was accepted by the annual SIGBOVIC conference. Note: SIGBOVIC covers the "three neglected quadrants of research: joke realizations of joke ideas, joke realizations of serious ideas, and serious realizations of joke ideas."
You are not allowed to view links. Register or Login to view. p.238/378 (234 on the page itself)
You are not allowed to view links. Register or Login to view.
|
|
|
| Bing AI Generating Continuations of Voynich Text |
|
Posted by: Psillycyber - 10-04-2023, 11:04 PM - Forum: Analysis of the text
- Replies (4)
|
|
I wrote an article relating to the Voynich Manuscript for LessWrong, a website dedicated to rationality and AI Safety research.
You are not allowed to view links. Register or Login to view.
The article is meant partly as an introduction to the topic of the VMS for that crowd, many of whom may not be familiar with the VMS. So, you might skip the first few paragraphs.
The meat of the article is the question of whether Bing AI (or, I suppose, one of the other AI large language models, or LLMs, like ChatGPT) can generate plausible continuations of Voynichese text.
Why might we in the Voynich community care? Well, if an AI LLM can generate high-fidelity continuations of Voynichese text that reproduce all of the statistical regularities of the original VMS, then that suggests that, somewhere inside the "Giant Inscrutable Matrices" of these LLMs, these LLMs are able to model how Voynichese is produced, which might eventually give us some clues as to how Voynichese was produced. I say "eventually" because, even if we could prove that this current generation of LLMs can generate high-fidelity continuations of Voynichese, this current generation of LLMs won't have the "insight" to know how they are doing that. We would need a later, more powerful generation of LLMs to inspect the "Giant Inscrutable Matrices" of these earlier LLMs to eventually make the rules explicit to us.
Why might we suspect that Bing AI could possibly generate high-fidelity continuations of Voynichese text, i.e., "speak Voynichese"? It's because that's basically the sort of task that these LLMs are tailor-made for: predicting the next token across essentially all of the languages and domains of the Internet. We also know that the most recent generation of LLMs can essentially create and decompress their own languages to themselves across sessions. Even though the size of the VMS corpus is relatively small compared to the size of all of the text on the Internet on which they were trained, if the rules behind predicting the next Voynichese token end up being simpler than the rule of predict the next token across any domain you might see on the Internet, then there is reason to believe that these LLMs might know Voynichese.
I explain the basic idea here in the article:
Quote:Humanity is essentially in the same relationship to the VMS as AI large language models (LLMs) are to the entire textual output of humans on the Internet. The entire Internet is the LLM's Voynich Manuscript. This might help give people some intuition as to what exactly LLMs are doing.
The LLM starts off with no clue about human concepts or what our words mean. All it can observe is statistical relationships. It creates models for creating that text that allows it to predict/generate plausible continuations to starting text prompts. In theory, with sufficient statistical mastery of the text in the VMS, humans should be able to simulate a process by which to generate increasingly-plausible-sounding continuations of "Voynichese" in the same way that AI LLMs generate plausible-sounding continuations of English or Japanese, even if humans never "understand" a single "vord" of Voynichese. As our process becomes increasingly-good at generating continuations of Voynichese that obey all of the statistical properties of the original distribution, we might say that humans would be asymptotically approaching a high-fidelity simulation of the process (whatever that was) that originally created the Voynichese.
So, in this article, I take my first stab at getting Bing AI to generate a continuation of Voynichese text. I also (futilely) try to get Bing AI to explain its method. Unfortunately, I don't think this will work as a backdoor method to find the way Voynichese was created because no current LLM has that much insight into how it decides to do the things it does. It would require a later, more powerful LLM to go back and analyze what Bing AI was doing here.
But before we get there, I wrote in the comments under the article some suggestions for, if someone wanted to continue this project to really rigorously find out how well Bing AI can generate Voynichese, how I would do it:
1. Either use an existing VMS transcription or prepare a slightly-modified VMS transcription that ignores all standalone label vords and inserts a single token such as a comma [,] to denote line breaks and a [>] to denote section breaks. There are pros and cons each way. The latter option would have the disadvantage of being slightly less familiar to Bing AI compared to what is in its training data, but it would have the advantage of representing line and section breaks, which may be important if you want to investigate whether Bing AI can reproduce statistical phenomena like the "Line as a Functional Unit" or gallows characters appearing more frequently at the start of sections.
2. Feed existing strings of Voynich text into Bing AI (or some other LLM) systematically starting from the beginning of the VMS to the end in chunks that are as big as the context window can allow. Record what Bing AI puts out.
3. Compile Bing AI's outputs into a 2nd master transcription. Analyze Bing AI's compendium for things like: Zipf's Law, 1st order entropy, 2nd order entropy, curve/line "vowel" juxtaposition frequences (a la Brian Cham), "Grove Word" frequences, probabilities of finding certain bigrams at the beginnings or endings of words, ditto with lines, etc. (The more statistical attacks, the better).
4. See how well these analyses match when applied to the original VMS.
5. Compile a second Bing AI-generated Voynich compendium, and a third, and a fourth, and a fifth, and see if the statistical attacks come up the same way again.
There are probably ways to automate this that people smarter than me could figure out.
|
|
|
|
| Abbreviations, anyone? |
|
Posted by: R. Sale - 04-04-2023, 06:43 PM - Forum: Analysis of the text
- Replies (12)
|
|
I know that the potential use of abbreviations has been an ongoing topic. Here is a manuscript with a particular part of the text noted for the use of abbreviations. Perhaps it is useful. Better to be redundant than silent.
Beromünster, Stiftskirche St. Michael, Ms. C 14
Paper · 265 ff. · 21 x 15 cm · Beromünster · second half of the 14th century / 15th century
Compendium morale de avibus / de quadrupedibus – Heinrich von Langenstein, De discretione spiritum – Johannes Gerson, Opus tripartitum de praeceptis Decalogi, de confessione, et de arte moriendi – Bonaventura, De praeparatione ad missam – Moralitates super evangelium sancti Lucae – Jacobus de Cessolis, De ludo scachorum (excerpts) – Sermon on Mary
Composite manuscript of catechetical-ascetic content, in quarto format on paper. Three fascicles of various strengths. The oldest is from the second half of the 14th century; it is written by Albert von Münnerstadt, Conventual from the Commandry of the Teutonic Knights of Hitzkirch, and contains Moralitates super evangelium sancti Lucae. In the second half of the 15th century, probably in Beromünster, this was bound together with two natural science Compendia moralia (excerpts from Thomas of Cantimpré's encyclopedia) and with catechetical treatises by Heinrich von Langenstein, Johannes Gerson and Bonaventure. Scholarly manuscript for regular use in the area of pastoral care (hasty hand with numerous abbreviations, especially in the third fascicle). (luz)
You are not allowed to view links. Register or Login to view.
|
|
|
|
| What initially sparked your interest in Voynich research? |
|
Posted by: Mark Knowles - 27-03-2023, 07:58 AM - Forum: Voynich Talk
- Replies (23)
|
|
As an example: Obviously I was curious about the unknown script and its meaning. However my interest was really sparked by seeing the Rosettes folio and particularly the "castle" and also to some extent the other buildings and geographical features that appeared to be illustrated on that page. The page looked like a fantasy map out of Tolkien or some other fantasy literature, however it seemed more likely that it was a real map than something dreamt up by modern fantasy fiction. I found the idea that the page represented a map of some unidentified, but real geographical region, a fascinating one and that by identifying the geographical area one might be able to say something about where the manuscript came from enticing. So that is what drew me initially to be interested in this manuscript.
However, I am curious what really attracted others to this manuscript amongst the uncountable other topics one could devote one's time to.
|
|
|
|
| Yet another solution: The Voynich Silenen Comedy |
|
Posted by: lelle - 26-03-2023, 07:14 PM - Forum: News
- Replies (33)
|
|
You are not allowed to view links. Register or Login to view.
There's a PDF available for download in English.
I couldn't find any description of the methodology in it though.
Since there is a connection to the Heiligenkreuz University, I had high hopes for this decipherment, but from a quick look it seems I got my hopes up in vain.
What do you think?
|
|
|
|
| Mapping Voynichese Text to Music |
|
Posted by: pfeaster - 22-03-2023, 07:16 PM - Forum: Analysis of the text
- Replies (3)
|
 |
The recent discussion You are not allowed to view links. Register or Login to view. drew my attention back to Stephen Gorbos's musical work "Such sphinxes as these obey no one but their master," which is supposed to have drawn somehow on the Voynich Manuscript. I wasn't able to find any other reference to Gorbos on this forum, so I'm guessing his composition hasn't been much discussed here.
But it doesn't seem to have been much discussed anywhere else, either. Gorbos's You are not allowed to view links. Register or Login to view. has the audio and a text "about" the composition, but the latter has nothing to say about any particular process by which the Voynich Manuscript fed into it. An You are not allowed to view links. Register or Login to view. on the same site states that the work was "inspired" by the manuscript, but again without providing any specifics about the connection. Finally, an You are not allowed to view links. Register or Login to view. on the Beinecke's own site states that Gorbos "drew from the text," and that the composition was one of several in which composers had "used" Beinecke mansucripts as "inspiration"; but again, no specifics.
That's all the information I was able to find. Does anyone here know anything more?
Based on the sources I've mentioned, my impression is that Gorbos probably used the Voynich Manuscript as "inspiration" in only a very loose sense, and that his composition probably doesn't map content from the Voynich Manuscript to music in any consistent or replicable way. There wouldn't be anything wrong with that; indeed, he never claimed to have done any such thing, as far as I can see. But the fact that his work exists might have discouraged others from playing around with more direct mappings of text to music by creating a false impression that this has already been tried.
The idea has sometimes been put forward that Voynichese is actually a musical notation. If it is, "deciphering" it correctly ought to produce something that sounds conventionally musical, like this audio pulled algorithmically from some plates depicting programs for automatic organ barrels in Athanasius Kircher's Musurgia Universalis (which we can thereby recognize as containing "real" music):
You are not allowed to view links. Register or Login to view.
On the other hand, if a mapping of text to music sounds thoroughly weird and atonal, I'd consider that the equivalent of a "word salad," and strong evidence of a "wrong" solution. For example, I was never able to get any plausible-sounding music out of this plate in Robert Fludd's Utriusque Cosmi Maioris Scilicet et Minoris Metaphysica, Physica atque Technica Historia:
You are not allowed to view links. Register or Login to view.
I don't know whether this means that I haven't hit on the right mode of decipherment yet, or whether the plate is just a mock-up of a medium for automatic music without any "real" musical content in it.
Now, in retrospect, I guess I'd always assumed that Gorbos had done something similar to what I'd done some years back when I tried to play that Robert Fludd plate: choosing a method for mapping an inscription to music and implementing it without worrying about whether the result sounded conventionally musical or not -- or maybe even hoping it wouldn't sound conventionally musical, since that would be boring from an experimental music standpoint. That is, I supposed he'd come up with a musical equivalent of a "word salad" and reveled in it as a musical equivalent to Dadaist poetry.
But now I suspect Gorbos didn't do anything like that after all. So I'm halfway tempted to try it myself.
What I have in mind is an algorithm that would take a standard EVA transcription as input and convert it automatically into music (output in MIDI), ideally handling Voynichese paragraph, line, and word structures in such a way that they're still recognizable as structures. That is, I'd want the mapping to be aurally intelligible in the same way that EVA is visually intelligible (and "almost pronounceable"): listening to the music ought to help us latch onto patterns that are "really there," as an alternative way of experiencing them and puzzling over them.
Some crude initial ideas:
(1) Interpret each Voynichese word as either a single note or a chord, with a particular duration -- but what word elements should represent what? Is there any way to arrange this such that every word would be practically "playable"?
(2) Interpret gallows as clefs, key signatures, or accidentals?
(3) Interpret each paragraph as a separate "piece" of music?
(4) Interpret each line as a stanza-like unit within the "piece" (but through-composed and non-repeating)?
(5) Map the most common word structures to the most "ordinary" musical features -- but how?
|
|
|
|
|
