The origins of ancient writing systems have long fascinated historians and linguists alike, but a groundbreaking new study using artificial intelligence has revealed surprising structural connections between scripts from Africa and the Caucasus. Researchers from San Diego State University (SDSU) have found that the Armenian alphabet shares striking structural similarities with the ancient Ethiopic writing system, also known as Ge'ez, challenging long-held assumptions about the independent development of these ancient scripts.
The study, published in Digital Scholarship in the Humanities, employed a deep convolutional neural network named "FeedelLigence" to objectively measure structural similarities across thousands of characters. By moving beyond the subjective visual impressions that have guided historical linguistics for decades, scientists were able to provide quantitative evidence for connections that scholars had long suspected but could never definitively prove.
A Computational Approach to Ancient Scripts
For many years, scholars noticed visual resemblances between the Ethiopic script, developed in the Horn of Africa more than 1,600 years ago, and the alphabets of Armenia, Georgia, and Caucasian Albania. However, these observations were difficult to quantify and replicate. To address this gap, the SDSU team, led by Professor Sam Kassegne of the Department of Mechanical Engineering, trained an AI model on over 28,000 images of Ethiopic characters. The computer learned to recognize fundamental shapes, curves, straight lines, and angles without any contextual data regarding history, religion, or geography.
"Our aim was to move beyond visual impressions that are difficult to test or replicate," explained Kassegne. "By making our criteria explicit and mathematical, we introduced an objective computational approach that is easily reproducible. We believe that this reproducibility is the key contribution of our method."
When the AI compared the Armenian letters to the Ethiopic/Ge'ez writing, the results were striking. Armenian showed the strongest structural similarity, with a mutual information score of 0.7428 bits and a low cosine distance of 0.0774. The Caucasian Albanian script demonstrated moderate similarity, while the Georgian alphabet showed less consistent connections. As a control, the Latin alphabet was tested and exhibited significantly lower structural resemblance, confirming that the observed patterns were not random.
- The Armenian Alphabet, A Vision From God
- The Kingdom of Axum: Facts and Legends of a First Millennium Empire
Mesrop Mashtots monument in Armenia (AKB7 / CC BY-SA 4.0)
Converging Timelines and Cultural Exchange
One of the most compelling aspects of this research is how precisely the computational data aligns with known historical timelines. The Armenian alphabet was created around 405 AD by the scholar-monk Mesrop Mashtots. During this same period, the Ethiopic writing system of the Kingdom of Aksum was expanding its reach and influence across the region. The AI model found that the Armenian alphabet appeared almost as similar to Ethiopic as Ethiopic is to its own earlier form, Proto-Ge'ez, a finding that suggests the resemblance is unlikely to be coincidental.
Historical records confirm that people from Ethiopia traveled extensively through the Middle East during the 4th and 5th centuries, including to Jerusalem, Egypt, and Syria. Mesrop Mashtots himself journeyed through these same regions while developing his script. While the study does not definitively prove direct copying or borrowing, it strongly supports the possibility of cultural contact and the sharing of ideas between the Horn of Africa and the Caucasus.
- The AI-driven Evolution of Archaeological Exploration
- New AI Algorithm is Cracking Undeciphered Languages
Daniel Zemene, the study's first author and an AI and machine learning researcher at SDSU's NanoFAB Lab, emphasized the broader significance of the convergence between computational and historical evidence. "What makes the research significant is that computational geometry and historical scholarship converge on the same scripts and time period," said Zemene.
"The model had no access to historical records, yet it learned purely from visual and structural data and identified Armenian as the closest structural match to Ethiopic within the very timeframe historians have long debated. That convergence between computation and history is powerful."
Old Armenian Manuscript (Definitions of Philosophy, Ms. 1280) (walter callens / CC BY 2.0)
The Future of AI in Historical Linguistics
This study represents a significant advancement in how researchers approach ancient alphabet origins. While AI deciphering ancient languages is becoming increasingly common - from reading charred scrolls to identifying the earliest alphabets - using machine learning to map structural similarities between writing systems provides historians with a powerful new analytical tool. The technology allows scholars to bypass human bias and visual limitations, offering mathematical evidence for theories that were previously based only on observation.
The researchers are careful to note that structural similarity does not automatically mean direct borrowing. Throughout history, societies have shared ideas, including writing systems, through trade, religion, and diplomacy. Greek, Roman, Persian, and Arabic civilizations all influenced one another, and this new research suggests Ethiopia's ancient writing culture may also have played a meaningful role in the exchange of ideas across regions. The findings open new avenues for future research into the interconnected world of the ancient Middle East and its surrounding civilizations.
A Broader Picture of Ancient Connectivity
The implications of this research extend well beyond the specific alphabets studied. If the Armenian and Ethiopic writing systems were indeed influenced by shared contact, it suggests a degree of cultural and intellectual exchange between Africa and the Caucasus that historians have not fully appreciated. Throughout history, the regions of the Middle East (particularly Jerusalem, the Sinai, and the Levant) served as crossroads for diverse religious communities, including Ethiopian and Armenian groups, who may have interacted and exchanged knowledge.
The researchers emphasize that this study is a beginning rather than a conclusion. The mathematical framework developed by the SDSU team, using mutual information scores and cosine distance measures, can now be applied to other writing systems around the world. As AI continues to unlock secrets of the ancients, the intersection of computer science and the humanities is reshaping our understanding of how ancient civilizations communicated, connected, and influenced one another across vast distances and centuries.
Top image: From left, characters in the Ethiopic (portions only), Armenian, Georgian and Caucasian Albanian alphabets. Source: San Diego State University
By Gary Manners
References
Avitabile, R. and Glotzer, Z. 2026. Ancient alphabets, new insights: Researchers uncover hidden links among the letters. San Diego State University NewsCenter. Available at: https://www.sdsu.edu/news/2026/03/ancient-alphabets-new-insignts
Zemene, D., Zemene, E., Sankpal, A., Sahle, E., Keshavamurthy, V.A.B. and Kassegne, S.K. 2026. Machine learning techniques for exploring influence, commonalities, and shared origin of scripts: cases of Ethiopic, Armenian, Georgian, and Caucasian Albanian scripts. Digital Scholarship in the Humanities. Oxford University Press. Available at: https://academic.oup.com/dsh/advance-article/doi/10.1093/llc/fqag029/8539597
Zahid, N. 2026. Armenian and Georgian Languages Show Links to Ethiopic Script, Study Finds. GreekReporter. Available at: https://greekreporter.com/2026/04/01/armenian-georgian-languages-ethiopic-script-study/
