Skip to content Skip to footer

This artificial intelligence can complete ancient inscriptions that were damaged by time

The study of history is one of the most complicated fields that we can find. They should not only prepare themselves professionally to study cultures and languages ​​that are often no longer traced; rather, many of the fonts they need are damaged beyond repair . However, a new and powerful tool has emerged in Ithaca, and it is an artificial intelligence with the power of DeepMind that arrives to change the course of history.

As described by TechCrunch, this AI is capable of guessing words lost or difficult to understand due to wear and tear with surprising precision. In addition, you can also interpret the location and approximate date of the text. Of course, it is a use that is at least different from the one usually given to artificial intelligences; but it shows how useful this technology can be in whatever field we prepare it for.

The problem of wear in ancient texts affects many disciplines that study historical elements. Many of the experts and academics work with materials that have been degraded over time. Among them we have stone, clay or papyrus. Also, these writings can be in Akkadian, Ancient Greek, Linear A, and can talk about anything; from a hero's journey to a daily report. However, they all have in common a great damage accumulated by the passage of millennia.

What does this artificial intelligence do

Artificial intelligence, now called Ithaca, has shown surprisingly high accuracy. Allows you to fill gaps in worn or missing text. According to comments from the aforementioned medium, these spaces “can be as short as a lost letter or as long as a chapter.”

Egyptian hieroglyphs Gaps where text is worn or torn are often called lacunae, and they can be as short as a missing letter or as long as a chapter, or even an entire story. Filling them in can be trivial or impossible, but you have to start somewhere, and that's where Ithaca aims to help.

TechCrunch Ithaca has been trained with an entire library of ancient Greek texts. In addition, its creators have named it after the home island of Odysseus. Ithaca has the ability to recover lost words or phrases as they were probably written. It is so efficient that it can even figure out when said stories were written, and even where. Of course, it is not going to fill in entire lost pages, but it is a great tool that expands the possibilities of the research teams that work with these very old texts.

Imagen de DeepMind

Ithaca, the future of historical studies

Ithaca illustrates the potential contribution of natural language processing and machine learning in the humanities.

We need more projects like Ithaca to continue showing this potential, but also adequate courses and teaching material to train future researchers, who will have a better joint understanding of the humanities and AI methods.

Ion Androutsopoulos, professor from the University of Athens The specialized journal Nature has published the results of its effectiveness. To do this, have used as an example decrees of periclite Athens. Until then, it was thought that these texts had been created on dates around the year 445 BC; however, the suggestions of Ithaca show that they were written in the year 420 BC approximately. This result is the same as the most recent tests carried out by experts.

Of course, this artificial intelligence still has some way to go before it is totally reliable. According to the experts in the study, Ítaca obtained a total of 25% reliability in the first scan. However, matching AI efforts with human insights, academics obtained up to 72% accuracy. Of course, this is one of those situations where human perception is essential to the process; but without a doubt Ítaca has been a great help when it comes to speeding up processes, eliminating loose ends and suggesting starting points.

If you want to try Ithaca, the team has uploaded a limited version of the tool to the internet. You can enter and complete your own Greek texts that contain gaps. On the other hand, if your texts have more than 10 words, you can test them in this Colab.

In case you are simply interested in seeing how it works, you can use the provided examples to see how artificial intelligence is able to fill in gaps in ancient writings. The AI ​​code is also available on GitHub.