While teaching humans about their ancient civilizations may seem like an odd job for artificial intelligence, it has potential. Traditionally, archeological surveys and decipherment have been painstakingly tedious. This technology could automate or streamline much of the process, helping people uncover more about the past at an exponential rate.
Why AI Is Needed to Teach About Ancient Civilizations
Spoken language is more or less universal. Throughout history, written language has been far rarer. The earliest known writing system is cuneiform, which was invented around 3100 B.C. by the Sumerians. Preliterate carved images date back as far as 4400 B.C., so academics have thousands of years of records to pour through and translate.
There are also glyphs, pottery, graves, structures and statues, each with a unique story. For centuries, humans have painstakingly identified, deciphered and investigated these curios. Pursuit, discovery and success are rewarding — even thrilling. However, progress is slow. Sometimes, an exceedingly small number of subject matter exists, creating bottlenecks.
What if researchers didn’t have to wait? What if they could accelerate their progress tenfold? With AI, that might be possible. An advanced, purpose-built model could uncover secrets that have been hidden for thousands of years.
A machine learning model’s power lies in automation and evolution. Since it learns as it processes new information, it can evolve as research or archeological projects progress, effectively future-proofing itself. Moreover, it requires minimal human oversight and can act independently, enabling it to carry out complex multistep assignments on its own.
What Historians Have Learned About Premodern Cultures Using AI
While modern AI is relatively new, scientists and archeologists have already used it to learn more about where premodern people lived and how they communicated.
Words in Long-Dead Languages
One word can have countless meanings depending on the author’s intentions and the composition’s context. This complicates decipherment. Even simple, pointless phrases become complex puzzles. The joke “What does a clock do when it’s hungry? It goes back for seconds” is a great example because it is a play on words. In a different language, it may be meaningless.
In the past, computer programs stumbled over these nuances. Natural language processing technology uses part-of-speech tagging, tokenization and lemmatization to recognize individual morphemes. With this framework, an algorithm could grasp the intricacies of context and meaning, even in long-dead languages.
Typically, deciphering ancient languages manually has been a laborious, error-prone task. Now, a model with NLP capabilities could decode written language in a fraction of the time.
Take the figurative geoglyphs — pre-Columbian designs etched into desert sands — for instance. It took nearly one century to discover 430 Nazca geoglyphs around the Nazca Pampa. Using AI, a research team found 303 new ones, almost doubling the total known number within just six months of field surveying.
Where Archeological Sites Are
Recently, a research team from Khalifa University in Abu Dhabi used AI to identify signs of a 5,000-year-old civilization underneath the dunes of the Rub al-Khali, the world’s largest desert. Since it stretches over 250,000 square miles, it is notoriously difficult to study. Shifting sands and harsh conditions complicate archeological surveys.
The research team used high-resolution satellite imagery and synthetic aperture radar technology to detect buried artifacts from space. Those results were fed into a machine learning model for image processing and geospatial analysis, automating the investigation. This approach was accurate within 50 centimeters, demonstrating its potential.
Ways AI Is Improving Understanding of Bygone Eras
AI is also helping scientists understand more about how ancient civilizations functioned, giving them a clearer window into the past.
Simulating Ancient Cultural Attitudes
Michael Varnum, the social psychology area head and associate professor at Arizona State University, recently co-authored an opinion piece proposing using generative AI to simulate ancient cultural attitudes.
Existing methods struggle to uncover the mentality or behaviors of long-dead cultures. Varnum says people in his field usually use indirect proxies like archival data on crime levels or divorce rates to infer people’s values and feelings. However, this approach is indirect and inaccurate. His solution is to train an AI to analyze historical texts.
However, while AI could infer people’s opinions and emotions from written records, its insights will be skewed. Historically, the ability to read and write has been rare. Varmum admits any AI-generated insights would likely come from educated, upper-class individuals. Since social class affects psychology, the analysis would not provide a wholly accurate glimpse into the past.
Reconstructing Premodern Customs
Whenever archeologists recover objects from ancient burial sites or half-buried cities, guesswork is involved. Even if they know exactly what something was used for, they may be unable to determine how it works.
In the 1970s, researchers unearthed a grave in a Bronze Age cemetery in Iran. They found the oldest intact board game ever discovered, dating back 4,500 years. It consisted of 27 geometric pieces, 20 circular spaces and four dice. No rulebook was buried, so they could only guess how to play.
AI could recreate the rules, bringing back long-forgotten board games. The Digital Ludeme project is doing just that. Already, it has spanned three time periods and nine regions, making almost 1,000 games playable again. Today, these reconstructions are available online for anyone to play.
What More Can Be Learned From These Ancient Cultures?
There is still much more left to learn from AI. Cuneiform is one of the most interesting. Today, academics have access to around 5 million Sumerian words, millions more than Romans left in Latin. Many of the numerous clay tablets uncovered in the region have yet to be deciphered, and more are unearthed almost daily.
To streamline the process, the research team uses AI to join tablet fragments, piecing together parts to accelerate decipherment. They are also training it to decipher cuneiform, which only a handful of experts are capable of. The speed of algorithmic processing could make this technology infinitely faster than humans.
This new knowledge could fill gaps in history books. Even though humans have an expansive cultural history, many regions remain unexplored because they haven’t had the technology. With machine learning techniques and generative models, they can have a deeper understanding of the world, gaining a new perspective on history.
With AI’s help in uncovering archeological sites, deciphering long-dead languages and translating ancient texts, industry professionals could find new books, historical accounts, artworks and treasures. Those findings could be displayed in a museum or help descendants connect with their ancestors.
The Future Outlook of AI Solutions as Archeological Tools
AI can decipher long-dead languages, locate ancient burial grounds and simulate ancient practices. Its findings could end up in history books or museums. Of course, academics should tread carefully. While this technology is powerful, bias, inaccuracies and hallucinations are not uncommon. A human-in-the-loop approach could help them mitigate these issues.
