Another Crazy Day in AI: New Model Helps Decode Ancient Roman Inscriptions
- Wowza Team
- Jul 24
- 4 min read
Hello, AI Enthusiasts.
Week’s almost over, but AI’s timeline is jumping centuries.
A new model from Google DeepMind is helping scholars reconstruct fragments of Latin inscriptions. AI as historian? We’re there.
Amazon, on the other hand, is stepping back... closing its AI lab in Shanghai as global tensions tighten.
Meanwhile, President Trump is rolling out his administration’s AI “action plan,” including executive orders aimed at boosting innovation and tightening oversight to keep the U.S. ahead in the global tech race.
When one model restores the past, another shapes the future.
Here's another crazy day in AI:
Google trains model to restore fragmented Roman inscriptions
Amazon closes AI lab in Shanghai
Trump signs orders for AI oversight and innovation
Some AI tools to try out
TODAY'S FEATURED ITEM: Advanced Model Transforms Ancient Text Analysis
Image Credit: Wowza (created with Ideogram)
What stories are locked inside the thousands of weathered Roman inscriptions that historians encounter every day?
Inscriptions carved into stone, metal, and pottery once carried laws, messages, dedications, even casual notes across the ancient Roman world. Many survive, but often in damaged or partial form, eroded by time, or broken and scattered. For historians, piecing these texts back together and placing them in the right historical context can be incredibly difficult and time-consuming.
A new model from Google DeepMind, called Aeneas, was developed to support this work. Introduced in a paper published in Nature, Aeneas is a generative AI system built specifically to help historians interpret and restore fragmentary Latin inscriptions. It was developed in collaboration with researchers from the University of Nottingham, Warwick, Oxford, and others. By processing both the text and visual image of an inscription, it can retrieve parallels from a database of over 176,000 inscriptions—helping scholars identify patterns, estimate dates, and better understand the text’s origin and meaning. An interactive version is now available at predictingthepast.com, alongside open-source code and datasets.
Here's how it works:
Identifies textual parallels - Searches through extensive Latin inscription databases to find texts with similar language patterns, formulas, or historical contexts that would otherwise require lengthy manual research
Processes visual and textual data - Examines both the written content and physical appearance of inscriptions to determine their likely geographical origins across the Roman Empire
Reconstructs damaged text - Attempts to fill gaps in fragmentary inscriptions, including cases where historians don't know how much material is missing
Provides dating estimates - Offers chronological placement based on linguistic and contextual analysis, including probability ranges for contested examples
Enables collaborative research - Testing with practicing historians showed that combining human expertise with the model's suggestions improved overall results
Maintains accessibility - Made freely available to researchers and educators with supporting materials and open-source components
Shows broader potential - While focused on Latin inscriptions, the methods could potentially be applied to other ancient languages and writing materials
Working with ancient inscriptions has always been a specialized field that requires considerable expertise and patience. Historians spend years learning to recognize the subtle differences in language, formulaic expressions, and cultural contexts that can help date and locate a text. When faced with a damaged inscription, they must mentally compare it against hundreds or thousands of similar texts they've encountered, looking for patterns that might provide clues about its meaning and significance. This process, while fundamental to the discipline, can be enormously time-consuming.
Aeneas essentially automates part of this comparative work, allowing historians to quickly identify potential connections and parallels that they can then evaluate using their specialized knowledge. What's particularly interesting about the model's design is how it handles scholarly uncertainty. When the researchers tested it on Augustus's Res Gestae, one of the most famous and debated Roman inscriptions, Aeneas didn't produce a single definitive answer about its date. Instead, it generated a probability distribution that captured the ongoing academic discussion, with different peaks corresponding to the various theories historians have proposed over the years. This approach suggests that digital tools might be most valuable when they help organize and quantify scholarly debates rather than attempting to resolve them outright. The model's ability to work with both textual and visual information also opens new possibilities for understanding how inscriptions functioned in their original settings, potentially revealing connections that become apparent only when examining large collections of material together.
Read the full article here.
Read the paper here.
OTHER INTERESTING AI HIGHLIGHTS:
Amazon Closes AI Lab in Shanghai
/Ryan McMorrow, (Deputy Beijing Bureau Chief), and Zijing Wu, (Asia Tech Correspondent), on Financial Times
Amazon is closing its AI research lab in Shanghai, citing strategic shifts amid escalating US-China tensions. The move follows similar pullbacks from IBM and Microsoft as scrutiny grows over AI work in China. The lab, run by Amazon Web Services since 2018, contributed to open-source tools and generated nearly $1B in sales. Its closure underscores the impact of geopolitics on global AI R&D efforts.
Read more here.
Trump Signs Orders for AI Oversight and Innovation
/Spencer Kimball, (Energy Reporter), and Kevin Breuninger, (Politics Reporter), on CNBC
President Donald Trump announced a sweeping AI “action plan” that includes executive orders aimed at regulating bias and spurring innovation. The administration plans to require AI developers to prove ideological neutrality to qualify for federal contracts. Trump’s team is also seeking to revise procurement standards and reduce regulations it sees as roadblocks to AI growth. The announcement signals a stronger push to shape the U.S. AI landscape along policy and political lines.
Read more here.
SOME AI TOOLS TO TRY OUT:
That’s a wrap on today’s Almost Daily craziness.
Catch us almost every day—almost! 😉
EXCITING NEWS:
The Another Crazy Day in AI newsletter is on LinkedIn!!!
Leveraging AI for Enhanced Content: As part of our commitment to exploring new technologies, we used AI to help curate and refine our newsletters. This enriches our content and keeps us at the forefront of digital innovation, ensuring you stay informed with the latest trends and developments.
Comments