Voynich Handwriting: No Artificial Intelligence Can Decipher It
Contrary to popular belief, Wilfrid M. Voynich is not the author of the elegant letter manuscript that bears his name, but the bookseller who bought it in 1912. It is not yet known if the Voynich Manuscript is a witchcraft book., cooking recipes or a manual on medicinal herbs. With bizarre plant drawings, naked pregnant women bathing in green liquids and astrological signs, no one has ever deciphered it - not even artificial intelligence.
When it is applied to the text, research is usually rejected soon after because it makes no sense (or artificial intelligence is not as intelligent as we thought it was in translation).
One of the last attempts came in 2016, when computer science professor Greg Kondrak and student Bradley Hauer of the University of Alberta (Canada) computed the text, but only a few things, such as how often letters appear and what combinations are the most common. In theory, they could compare the manuscript to existing languages and thus find a match.
As a basis, they used the Universal Declaration of Human Rights, written in 380 languages. No neural networks or deep learning were used; the algorithms were based on the statistical analysis of modern languages, which has already been unusual since the manuscript was written in the 15th century.
The mystery remains
By publishing an article, the pair reported discovering that the Voynich Manuscript's original language was Hebrew, which was later coded. Many people wondered how they came to this conclusion, since AI had not been trained for this language in its archaic form.
Even though the algorithms pointed to a language, they could not assess the probability of each match. This means that it really is not known if the text was in Hebrew. Kondrak and Hauer assumed that the manuscript was an anagram, meaning that the letters in each word were shuffled, which would allow for greater freedom in interpreting the text - that is, it could be a witchcraft book, culinary recipes or a herbal handbook .
Using artificial intelligence to unravel the mystery of the Voynich Manuscript — and it fails miserably — raises another question: artificial intelligence has difficulty fully understanding the complexities of human language. For South Korea's artificial intelligence company chief scientist Mind Ai, John Doe, there is a big difference between natural language processing, a subset of artificial intelligence, and what he calls "natural language of reasoning."
According to him in Medium, natural language processing is based on training the computer to manipulate human language, and this depends on machine learning algorithms and neural networks. This happens, for example, when the mobile user talks to Alexa or Siri, or when he learns how to finish his sentences (like in Gmail).
While good at what they do every day, for John Doe AI “will never achieve mastery of language because algorithms don't know what they are doing.” That is, AI would need to have genuine human intelligence to understand the context, the tone and meaning of what is said, and this includes unraveling the mystery hidden in the pages of the Voynich Manuscript.