4 ms·

Are there transcriptions available of voynich script into ascii? Someone with access to GPT3 should try having it translate some bits into English. Would be th

Are there transcriptions available of voynich script into ascii? Someone with access to GPT3 should try having it translate some bits into English. Would be the most shocking 'look what GPT can do' moment, to me!

dane-pgp 6y ago
Alternatively, there is an AI technique for translating between languages without needing a bilingual dictionary or parallel texts[0]. There may not be enough data in the Voynich manuscript to generate an accurate translation of any word in it, but the model generated by the AI might be enough to decide which language is closest to Voynichese, or which transcriptions of the manuscript are the least "noisy". [0] https://www.technologyreview.com/2017/11/29/241639/artificial-intelligence-can-translate-languages-without-a-dictionary/ https://www.technologyreview.com/2017/11/29/241639/artificia...
- DonaldFisk 6y ago
  For this to succeed, the text would have to be meaningful in some language, which is by no means certain. All languages have grammars, but nobody has yet discovered any grammar for the Voynich Manuscript operating outside of words.
  - dane-pgp 6y ago
    If it is not meaningful in any language, then we would not expect it to obey Zipf's Law, however it has been shown that it does (as mentioned by an earlier commenter). https://skeptics.stackexchange.com/questions/6085/has-it-been-mathematicaly-determined-that-the-voynich-manuscript-is-not-gibberis/16748#16748 https://skeptics.stackexchange.com/questions/6085/has-it-bee...
    - DonaldFisk 6y ago
      If it didn't obey Zipf's Law, we could safely conclude it wasn't in an unencrypted language, but you can't make the reverse inference, as lots of phenomena obey Zipf's Law which are nothing to do with language. I've generated a fake Voynich Manuscript which I can assure you is entirely meaningless, but it obey's Zipf's Law. Here it is: http://www.fmjlang.co.uk/voynich/generated-voynich-manuscript.html http://www.fmjlang.co.uk/voynich/generated-voynich-manuscrip... And here's the proof it obey's Zipf's Law: http://www.fmjlang.co.uk/voynich/VoynichZipf.html http://www.fmjlang.co.uk/voynich/VoynichZipf.html Two other researchers (Gordon Rugg and Torsten Timm) have also generated fake Voynich Manuscripts, and theirs obey Zipf's Law too.
      - Daniel_sk 6y ago
        “ Zipf’s law was discovered centuries after the accepted date of creation of the Voynich text. Thus, proposed solutions like the use of sixteenth-century cipher methods, although not impossible, can hardly account for the presence of Zipf’s law in the Voynich text.”
        DonaldFisk 6y ago
        The constraint the author of the Voynich Manuscript would have been working to is that it had to look at least superficially like a natural language. The bare minimum for that is that the choice of current glyphs depends on the previous glyph, and I think that would have been obvious even at the time the Voynich Manuscript was written. I think that's a necessary (but not a sufficient) property of all alphabetic languages, and suspect it's enough to make the distribution Zipfian. It would be nice to have a mathematical proof or an experimental verification that it's always or nearly always Zipfian. Generating text in that way doesn't require a prior knowledge of Zipf's Law, and I don't think it was unthinkable in the 15th Century.
        dane-pgp 6y ago
        It's not unthinkable that it could be the result of a hoax, and I admit that there aren't a lot of examples of documents even remotely like the VM, but I still think that an Occam's Razor approach would rule out a hoax. How likely is it that someone (or even some group) trying to perpetrate a hoax would come up with a method which was not only undetectable by analytical tools available at the time but also by tools that wouldn't be invented even centuries later? I suppose we should compare it to the invention of the Vigenère cipher, in the 16th century, which was only broken in the 19th century. Conversely, we have examples of texts written in forgotten languages which were later translated / remain untranslated, and we don't / didn't assume them to be hoaxes (assuming they obey Zipf's Law). We also have examples of people inventing new written languages for spoken languages that didn't have a (known) written form, so I think Occam's Razor would support the hypothesis that this is what happened with the VM.

dane-pgp 6y ago

Alternatively, there is an AI technique for translating between languages without needing a bilingual dictionary or parallel texts[0]. There may not be enough data in the Voynich manuscript to generate an accurate translation of any word in it, but the model generated by the AI might be enough to decide which language is closest to Voynichese, or which transcriptions of the manuscript are the least "noisy". [0] https://www.technologyreview.com/2017/11/29/241639/artificial-intelligence-can-translate-languages-without-a-dictionary/ https://www.technologyreview.com/2017/11/29/241639/artificia...

DonaldFisk 6y ago
For this to succeed, the text would have to be meaningful in some language, which is by no means certain. All languages have grammars, but nobody has yet discovered any grammar for the Voynich Manuscript operating outside of words.
- dane-pgp 6y ago
  If it is not meaningful in any language, then we would not expect it to obey Zipf's Law, however it has been shown that it does (as mentioned by an earlier commenter). https://skeptics.stackexchange.com/questions/6085/has-it-been-mathematicaly-determined-that-the-voynich-manuscript-is-not-gibberis/16748#16748 https://skeptics.stackexchange.com/questions/6085/has-it-bee...
  - DonaldFisk 6y ago
    If it didn't obey Zipf's Law, we could safely conclude it wasn't in an unencrypted language, but you can't make the reverse inference, as lots of phenomena obey Zipf's Law which are nothing to do with language. I've generated a fake Voynich Manuscript which I can assure you is entirely meaningless, but it obey's Zipf's Law. Here it is: http://www.fmjlang.co.uk/voynich/generated-voynich-manuscript.html http://www.fmjlang.co.uk/voynich/generated-voynich-manuscrip... And here's the proof it obey's Zipf's Law: http://www.fmjlang.co.uk/voynich/VoynichZipf.html http://www.fmjlang.co.uk/voynich/VoynichZipf.html Two other researchers (Gordon Rugg and Torsten Timm) have also generated fake Voynich Manuscripts, and theirs obey Zipf's Law too.
    - Daniel_sk 6y ago
      “ Zipf’s law was discovered centuries after the accepted date of creation of the Voynich text. Thus, proposed solutions like the use of sixteenth-century cipher methods, although not impossible, can hardly account for the presence of Zipf’s law in the Voynich text.”
      - DonaldFisk 6y ago
        The constraint the author of the Voynich Manuscript would have been working to is that it had to look at least superficially like a natural language. The bare minimum for that is that the choice of current glyphs depends on the previous glyph, and I think that would have been obvious even at the time the Voynich Manuscript was written. I think that's a necessary (but not a sufficient) property of all alphabetic languages, and suspect it's enough to make the distribution Zipfian. It would be nice to have a mathematical proof or an experimental verification that it's always or nearly always Zipfian. Generating text in that way doesn't require a prior knowledge of Zipf's Law, and I don't think it was unthinkable in the 15th Century.
        dane-pgp 6y ago
        It's not unthinkable that it could be the result of a hoax, and I admit that there aren't a lot of examples of documents even remotely like the VM, but I still think that an Occam's Razor approach would rule out a hoax. How likely is it that someone (or even some group) trying to perpetrate a hoax would come up with a method which was not only undetectable by analytical tools available at the time but also by tools that wouldn't be invented even centuries later? I suppose we should compare it to the invention of the Vigenère cipher, in the 16th century, which was only broken in the 19th century. Conversely, we have examples of texts written in forgotten languages which were later translated / remain untranslated, and we don't / didn't assume them to be hoaxes (assuming they obey Zipf's Law). We also have examples of people inventing new written languages for spoken languages that didn't have a (known) written form, so I think Occam's Razor would support the hypothesis that this is what happened with the VM.