Google AI Deciphers Ancient Roman Inscriptions

A groundbreaking AI tool named Aeneas, developed by Google DeepMind in collaboration with researchers, is now helping historians reconstruct, date, and locate damaged Latin inscriptions from Ancient Rome with extraordinary precision.

Decoding the Past with the Help of AI

Each year, archaeologists unearth around 1,500 new Latin inscriptions—etched into stone, metal, or pottery. These ancient texts vary from imperial edicts and personal epitaphs to poetic verses. Unlike traditional historical records, many of these inscriptions were written by ordinary people, giving researchers a raw and authentic glimpse into Roman daily life.

But there’s a challenge: Most inscriptions are incomplete, eroded, or broken, making interpretation extremely difficult and time-consuming—even for experienced epigraphists.

To solve this historical puzzle, a team led by Yannis Assael from DeepMind has introduced Aeneas, a neural network trained on nearly 200,000 inscriptions—equivalent to 16 million Latin characters. This vast dataset enables the model to deliver high-confidence reconstructions and contextual understanding.

What Aeneas Can Do

This AI model goes far beyond traditional keyword recognition. Aeneas performs three core tasks with impressive accuracy:

Text Reconstruction: It intelligently fills in missing or damaged parts of Latin inscriptions using linguistic context and historical patterns.

Dating: Aeneas estimates the decade of origin with a 13-year margin of error.

Geolocation: It predicts the original Roman province—among 62 possible regions—where the inscription likely originated.

In essence, Aeneas serves as a smart assistant for historians, capable of detecting subtle linguistic nuances and drawing connections across space and time.

Outperforming Expectations in Real-World Tests

In one key test involving the famous inscription Res Gestae Divi Augusti, the model successfully identified specific archaic spellings and language patterns, narrowing down its dating to two possible decades—mirroring the ongoing scholarly debate.

In another challenge, Aeneas analyzed an inscription found in modern-day Germany. By detecting similarities in language and structure, it suggested the inscription was inspired by an older altar inscription found in the same area—something human historians had only recently proposed.

A Tool, Not a Replacement

Despite its intelligence, Aeneas isn't meant to replace expert historians. Instead, it’s designed to complement their work. In a blind test with 23 professional historians, the model provided a helpful starting point in 90% of cases. Researchers emphasize that the most accurate insights emerge when human intuition and AI analysis work together.

Redefining the Future of Digital Archaeology
With the launch of Aeneas, Google DeepMind is not just showcasing the power of AI in science—it's redefining the role of technology in historical discovery. As more inscriptions are uncovered every year, tools like Aeneas will play a crucial role in preserving, interpreting, and sharing the hidden stories of our ancestors.

For those passionate about ancient civilizations, this represents a massive leap forward in archaeological research—powered by artificial intelligence.

smsonline