CLOSE

Researchers Claim to Crack the Voynich Manuscript Using AI, But Experts Are Skeptical

Computing scientists at the University of Alberta recently made a bold claim: They say they’ve identified the source language of the baffling Voynich Manuscript, and they did so using artificial intelligence.

Their study, published in Transactions of the Association of Computational Linguistics [PDF], basically states that an AI algorithm trained to recognize hundreds of languages determined the Voynich Manuscript to be encoded Hebrew. On the surface, this looks like a huge breakthrough: Since it was rediscovered a century ago, the Voynich Manuscript’s indecipherable text has stumped everyone from World War II codebreakers to computer programmers. But experts are hesitant to give credence to the news. “I have very little faith in it,” cryptographer Elonka Dunin tells Mental Floss. “Hebrew, and dozens of other languages have been identified before. Everyone sees what they want to see.”

Anyone who’s familiar with the Voynich Manuscript should understand the skepticism. The book, which contains 246 pages of illustrations and apparent words written in an unknown script, is obscured by mystery. It’s named for Wilfrid Voynich, the Polish book dealer who purchased it in 1912, but experts believe it was written 600 years ago. Nothing is known about the person who authored it or the book’s purpose.

Many cryptologists suspect the text is a cipher, or a coded pattern of letters that must be unscrambled to make sense. But no code has been identified even after decades of the world’s best cryptographers testing countless combinations. With their study, the researchers at the University of Alberta claim to have done something different. Instead of relying on human linguists and codebreakers, they developed an AI program capable of identifying the source languages of text. They fed the technology 380 versions of the Universal Declaration of Human Rights, each one translated into a different language and enciphered. After learning to recognize codes in various languages, the AI was given some pages of the Voynich Manuscript. Based on what it had seen already, it named Hebrew as the book’s original language—a surprise to the researchers, who were expecting Arabic.

The researchers then devised an algorithm that rearranged the letters into real words. They were able to make actual Hebrew out of 80 percent of the encoded words in the manuscript. Next, they needed to find an ancient Hebrew scholar to look at the words and determine if they fit together coherently.

But the researchers claim they were unable to get in touch with any scholars, and instead used Google Translate to make sense of the first sentence of the manuscript. In English, the decoded words they came up with read, “She made recommendations to the priest, man of the house and me and people." Study co-author Greg Kondrak said in a release, “It’s a kind of strange sentence to start a manuscript but it definitely makes sense.”

Dunin is less optimistic. According to her, naming a possible cipher and source language without actually translating more of the text is no cause for celebration. “They identify a method without decrypting a paragraph,” she says. Even their method is questionable. Dunin points out the AI program was trained using ciphers that the researchers themselves wrote, not ciphers from real life. “They scrambled the texts using their own system, then they used their own software to de-scramble those. Then they used it on the manuscript and said, ‘Oh look, it’s Hebrew!’ So it’s a big, big leap.”

The University of Alberta researchers aren’t the first to claim they’ve identified the language of the Voynich Manuscript, and they won’t be the last. But unless they’re able the decode the full text into a meaningful language, the manuscript remains as mysterious today as it did 100 years ago. And if you agree with cryptographers like Dunin who think the book might be a constructed language, a detailed hoax, or even a product of mental illness, it’s a mystery without a satisfying explanation.

nextArticle.image_alt|e
NOAA, Wikimedia Commons // Public Domain
arrow
Animals
Watch the First-Ever Footage of a Baby Dumbo Octopus
NOAA, Wikimedia Commons // Public Domain
NOAA, Wikimedia Commons // Public Domain

Dumbo octopuses are named for the elephant-ear-like fins they use to navigate the deep sea, but until recently, when and how they developed those floppy appendages were a mystery. Now, for the first time, researchers have caught a newborn Dumbo octopus on tape. As reported in the journal Current Biology, they discovered that the creatures are equipped with the fins from the moment they hatch.

Study co-author Tim Shank, a researcher at the Woods Hole Oceanographic Institution in Massachusetts, spotted the octopus in 2005. During a research expedition in the North Atlantic, one of the remotely operated vehicles he was working with collected several coral branches with something strange attached to them. It looked like a bunch of sandy-colored golf balls at first, but then he realized it was an egg sac.

He and his fellow researchers eventually classified the hatchling that emerged as a member of the genus Grimpoteuthis. In other words, it was a Dumbo octopus, though they couldn't determine the exact species. But you wouldn't need a biology degree to spot its resemblance to Disney's famous elephant, as you can see in the video below.

The octopus hatched with a set of functional fins that allowed it to swim around and hunt right away, and an MRI scan revealed fully-developed internal organs and a complex nervous system. As the researchers wrote in their study, Dumbo octopuses enter the world as "competent juveniles" ready to jump straight into adult life.

Grimpoteuthis spends its life in the deep ocean, which makes it difficult to study. Scientists hope the newly-reported findings will make it easier to identify Grimpoteuthis eggs and hatchlings for future research.

nextArticle.image_alt|e
Penn Vet Working Dog Center
arrow
Stones, Bones, and Wrecks
New Program Trains Dogs to Sniff Out Art Smugglers
Penn Vet Working Dog Center
Penn Vet Working Dog Center

Soon, the dogs you see sniffing out contraband at airports may not be searching for drugs or smuggled Spanish ham. They might be looking for stolen treasures.

K-9 Artifact Finders, a new collaboration between New Hampshire-based cultural heritage law firm Red Arch and the University of Pennsylvania, is training dogs to root out stolen antiquities looted from archaeological sites and museums. The dogs would be stopping them at borders before the items can be sold elsewhere on the black market.

The illegal antiquities trade nets more than $3 billion per year around the world, and trafficking hits countries dealing with ongoing conflict, like Syria and Iraq today, particularly hard. By one estimate, around half a million artifacts were stolen from museums and archaeological sites throughout Iraq between 2003 and 2005 alone. (Famously, the craft-supply chain Hobby Lobby was fined $3 million in 2017 for buying thousands of ancient artifacts looted from Iraq.) In Syria, the Islamic State has been known to loot and sell ancient artifacts including statues, jewelry, and art to fund its operations.

But the problem spans across the world. Between 2007 and 2016, U.S. Customs and Border Control discovered more than 7800 cultural artifacts in the U.S. looted from 30 different countries.

A yellow Lab sniffs a metal cage designed to train dogs on scent detection.
Penn Vet Working Dog Center

K-9 Artifact Finders is the brainchild of Rick St. Hilaire, the executive director of Red Arch. His non-profit firm researches cultural heritage property law and preservation policy, including studying archaeological site looting and antiquities trafficking. Back in 2015, St. Hilaire was reading an article about a working dog trained to sniff out electronics that was able to find USB drives, SD cards, and other data storage devices. He wondered, if dogs could be trained to identify the scents of inorganic materials that make up electronics, could they be trained to sniff out ancient pottery?

To find out, St. Hilaire tells Mental Floss, he contacted the Penn Vet Working Dog Center, a research and training center for detection dogs. In December 2017, Red Arch, the Working Dog Center, and the Penn Museum (which is providing the artifacts to train the dogs) launched K-9 Artifact Finders, and in late January 2018, the five dogs selected for the project began their training, starting with learning the distinct smell of ancient pottery.

“Our theory is, it is a porous material that’s going to have a lot more odor than, say, a metal,” says Cindy Otto, the executive director of the Penn Vet Working Dog Center and the project’s principal investigator.

As you might imagine, museum curators may not be keen on exposing fragile ancient materials to four Labrador retrievers and a German shepherd, and the Working Dog Center didn’t want to take any risks with the Penn Museum’s priceless artifacts. So instead of letting the dogs have free rein to sniff the materials themselves, the project is using cotton balls. The researchers seal the artifacts (broken shards of Syrian pottery) in airtight bags with a cotton ball for 72 hours, then ask the dogs to find the cotton balls in the lab. They’re being trained to disregard the smell of the cotton ball itself, the smell of the bag it was stored in, and ideally, the smell of modern-day pottery, eventually being able to zero in on the smell that distinguishes ancient pottery specifically.

A dog looks out over the metal "pinhweel" training mechanism.
Penn Vet Working Dog Center

“The dogs are responding well,” Otto tells Mental Floss, explaining that the training program is at the stage of "exposing them to the odor and having them recognize it.”

The dogs involved in the project were chosen for their calm-but-curious demeanors and sensitive noses (one also works as a drug-detection dog when she’s not training on pottery). They had to be motivated enough to want to hunt down the cotton balls, but not aggressive or easily distracted.

Right now, the dogs train three days a week, and will continue to work on their pottery-detection skills for the first stage of the project, which the researchers expect will last for the next nine months. Depending on how the first phase of the training goes, the researchers hope to be able to then take the dogs out into the field to see if they can find the odor of ancient pottery in real-life situations, like in suitcases, rather than in a laboratory setting. Eventually, they also hope to train the dogs on other types of objects, and perhaps even pinpoint the chemical signatures that make artifacts smell distinct.

Pottery-sniffing dogs won’t be showing up at airport customs or on shipping docks soon, but one day, they could be as common as drug-sniffing canines. If dogs can detect low blood sugar or find a tiny USB drive hidden in a house, surely they can figure out if you’re smuggling a sculpture made thousands of years ago in your suitcase.

SECTIONS

arrow
LIVE SMARTER
More from mental floss studios