Google DeepMind’s AI Historian Aeneas Rewrites Ancient History Research

For decades, historians have painstakingly reconstructed fragmented inscriptions like detectives solving millennia-old cold cases. Now, an artificial intelligence named Aeneas—developed by Google DeepMind—is accelerating breakthroughs in decoding ancient texts. In a landmark study published in Nature (October 2023), this “AI historian” accurately dated Emperor Augustus’s Res Gestae Divi Augusti to approximately 15 CE, soon after his death in 14 CE—a task requiring years of specialized human analysis.

The AI Historian Revolutionizing Ancient Text Analysis

Aeneas mimics historians’ methodology but processes centuries of data in seconds. By scanning global archives of Greek and Latin inscriptions, it identifies linguistic parallels, historical contexts, and textual patterns invisible to the human eye. As Warwick University classics professor Alison Cooley explains, “State-of-the-art generative models are turning epigraphy from a specialist discipline into a cutting-edge field of historical enquiry.”

Historians traditionally rely on physical fragments scattered across museums worldwide. “Studying history through inscriptions is like solving a gigantic jigsaw puzzle,” notes University of Warwick historian Thea Sommerschield. Aeneas accelerates this by cross-referencing damaged texts with similar wording, scripts, or historical periods—a process that previously demanded decades of training.

Human-AI Collaboration: 90% Accuracy Breakthrough

In tests across Greek and Latin inscriptions, Aeneas provided useful research starting points in 90% of cases, boosting historians’ confidence by 44%. Crucially, human-AI teams outperformed either working alone. When analyzing Augustus’s Res Gestae, Aeneas didn’t force a single date but highlighted two probable ranges—mirroring scholarly debates. “It exactly reflected current differences in scholars’ opinions,” said Cooley, praising its nuanced approach.

This addresses a critical AI weakness: hallucination. Unlike chatbots inventing false details, Aeneas assigns probability scores to predictions. Researchers credit its transparency to training on the I.PHI database—500,000+ translated ancient texts compiled by universities including Oxford and Heidelberg.

The Future of Historical Discovery

While not replacing historians, Aeneas acts as a tireless research assistant. Teams at DeepMind and European universities are expanding its capabilities to analyze coinage, artwork, and archaeological site data. As Cooley emphasizes, “This tool excels when guided by experts who ask precise questions.”

The fusion of artificial intelligence and human expertise isn’t just accelerating historical research—it’s revealing lost chapters of human civilization. As Aeneas analyzes previously indecipherable texts, our understanding of ancient politics, economies, and daily life transforms. Historians worldwide are now partnering with AI to resurrect voices silenced for millennia. Explore how technology continues to reshape archaeology in our analysis of AI-powered Pompeii discoveries.

Must Know

Q: How does the Aeneas AI historian work?
A: Trained on 500,000+ ancient Greek and Latin inscriptions, it identifies linguistic patterns, contextual clues, and historical parallels. It cross-references fragmented texts against its database to propose dates, origins, and missing text reconstructions.

Q: Can AI historians replace human researchers?
A: No. As demonstrated in the Nature study, Aeneas augments human expertise. Historians using the tool achieved superior results by interpreting its probabilistic outputs and validating findings against physical evidence.

Q: What safeguards prevent AI hallucinations in historical analysis?
A: Aeneas generates probability scores for each prediction and cites analogous inscriptions. Its training data—curated by universities—excludes unverified sources, reducing fictional outputs.

Q: Which institutions contributed to this AI historian project?
A: Google DeepMind collaborated with Warwick University, Oxford University, Ca’ Foscari University of Venice, and the University of Athens. The I.PHI database provided foundational epigraphic material.

Q: Will Aeneas analyze non-Roman historical periods?
A: Researchers confirmed plans to expand its capabilities to Egyptian hieroglyphs, Mayan scripts, and cuneiform tablets pending funding and data partnerships with global museums.