Humanity has always had a desire to learn about our past. It helps us to understand where we came from and how our future could unfold. We have discovered our ancestors were much more intelligent and inventive than we could have imagined but learning about them requires being able to read what they have left behind for us. In some cases that is easy as Ancient Greek and Latin have been translated and are generally understood by scholars and archaeologists. But what do you do when you don’t understand the language?

The Rosetta Stone

The Rosetta Stone was one of the greatest archaeological finds of all time. When it was found outside of Memphis, Egypt in 1799 it had three languages on it. The first was Egyptian hieroglyphs, the middle was demotic (the language used by the common people) Egyptian and the bottom was Ancient Greek. At the time understanding Egyptian hieroglyphics was next to impossible but by comparing the hieroglyph to other languages archaeologists and scholars were able to gain a basic understanding, which has expanded over time. 

The problem is that the languages that have been lost to time and are only recorded on pieces of stone do not have their own Rosetta Stone. Just figuring out where to start with these ancient languages is a daunting task and actually deciphering them can prove to be impossible. There are a lot of them out there and they were found all over the world. Scholars and archaeologists may have a new ally to enlist in their quest: AI.

Machine Learning To The Rescue

Researchers from MIT and Google have teamed up to develop a machine learning system that is capable of translating lost languages. The machine requires large amounts of text, from which compares words to determine any patterns in an attempt to map the language. Commonly found words may be nouns, like the name of a city or a person and the rest of the profile can be built around that. 

Linear A And Linear B

The researchers put their algorithm to the test. They used a language known as Linear B, which was discovered in 1886 by British archaeologist Arthur Evans on the island of Crete. The stone that the language was found on was determined to have two different languages on it, with Linear A being the other. Linear A was determined to be dated from around the time of the Minoan civilization or around 1800-1400 BCE (making it one of the earliest forms of writing known to mankind) while Linear B dated from around the time of the Myceneanas, who occupied the island after 1400 BCE.

Evans tried to decipher the languages unsuccessfully throughout his life. It took until 1953 for Linear B to be cracked when an amateur linguist pieced it together. That person made two assumptions, one) that commonly used words in the text were actually the names of places on Crete and two) that it was an early form of Ancient Greek. Those assumptions were both correct and he was able to decipher it, proving that Ancient Greek was in use far earlier than had been known. Linear A though has never been deciphered.

First Test

This new AI system may be the best bet we have on this planet to translate that, and other lost languages. Linear B and another lost language called Ugaritic, an early form of Hebrew, was run through the algorithm. Linear B was translated with nearly 67% accuracy by the algorithm. Linear A was not tested as a part of this simulation.

The algorithm relies on the fact that while languages do change over time they do not change that much, think of the small differences between French and Quebecois. Some symbols may change, names will change but those symbols will only appear with other symbols. If this is the case, and as long as the modern language that it became is known (or can be guessed), deciphering the language is actually fairly easy. 

Why Is Linear A So Difficult?

That is part of the problem with Linear A as it shares many of the same symbols with Linear B but it could not be deciphered into Ancient Greek. It appears that Minoan is completely different than Greek. The Minoan civilization was devastated by the eruption of Santorini (which baked the clay tablets that Evans found thereby preserving them) and lead the Mycenaeans to conquer the island. They appear to have adapted much of the Minoan culture into their own so it could be possible that they adapted the symbols that made the Minoan alphabet into their own language as a written form of it. There is of course no one alive today who speaks Minoan or Linear A and no Rosetta Stone exists for it, or at least has been found. What Evans found may be as close as it comes.

AI will most certainly be put to work at trying to decipher Linear A. It would be one of the great archaeological and scholarly triumphs of the 21st century if it would be successful. Considering mankind has had more than a millennia to understand it and a translation has eluded us AI may be our only hope to understand what these texts say. It would also help us understand a civilization that we only know from the remains of their buildings and from the art that has survived. 

There are of course other lost languages out there as well. AI will make those deciphering easier as well as it will be able to handle many of the most tedious and arduous tasks. While no one will make movies about AI in an Indiana Jones-esque role, it may play an even greater role than the fedora-wearing bullwhip toting hero of the silver screen could have ever hoped for.

Comments are closed.

Scroll to Top