AI and the Ancient World: Unveiling Secrets Through Technology
The video opens by emphasizing that for centuries, numerous ancient languages remained enigmatic, their meanings lost to time. Now, AI is cracking codes that baffled scholars for generations. The focus then shifts to how AI is revolutionizing our ability to interpret ancient texts. We're introduced to 18 ancient languages decoded by AI, with the first being the Herculanium scrolls. These fire-charred relics, buried by Mount Vesuvius in 79 AD, held an extensive collection, with scholars struggling for centuries to read them without destroying them. However, thanks to the development of advanced imaging techniques and machine learning, in 2023, the "Vesuvius Challenge" was launched, a global competition aimed at decoding these ancient manuscripts using cutting-edge AI technology.
- AI has enabled the virtual unrolling of scrolls from the Bodleian Library at Oxford, revealing multiple columns of ancient Greek text for the first time without physically opening the fragile papyrus.
- The decoded passages are believed to be the musings of Philodemus, an Epicurean philosopher, offering insights into ancient perspectives on pleasure and human experience.
- This demonstrates the profound synergy between technology and the humanities, which is essential for further exploration of ancient civilizations.
Shifting gears, we journey to Central Asia, where archaeological digs in the 1950s unearthed inscriptions in a previously unknown script. Fast forward to 2022, and a linguistic breakthrough occurred in the Almosi Gorge of Tajikistan. Here, a bilingual inscription, featuring both the mysterious script and Bactrean – a language we *do* understand – was discovered. This find served as the Rosetta Stone for the Kusan script, allowing researchers to correlate symbols with sounds. This led to identifying the name Vimma Taktu, a Kusan emperor, and the title "King of Kings" in both languages. Deciphering this script revealed a Middle Iranian language, shedding light on the Kusan Empire, a key player in spreading Buddhism and shaping trade routes. However, the translations hint at unsettling practices. The Kusons might have engaged in rituals that challenge the peaceful image often associated with them.
- This breakthrough offers fresh insights into cultural exchanges in ancient times.
- AI and collaborative research efforts revealed previously unknown aspects of ancient societies, leading to a paradigm shift in historical interpretation.
- Apply this to your own research: look for bilingual texts to help identify previously indecipherable scripts or languages.
Here are the specific applications of AI mentioned in the transcript for understanding ancient languages:
-
Virtual unrolling of scrolls - AI combined with advanced imaging techniques to read carbonized Herculaneum scrolls without physically opening them
-
Handwriting analysis - AI identifying microscopic differences in handwriting on the Dead Sea Scrolls, revealing multiple scribes worked on what was thought to be a single-scribe document
-
Pattern recognition - AI analyzing patterns in scripts like the Tangut, Olmec glyphs, and Rongorongo to identify linguistic structures
-
Language identification - Algorithms analyzing the Voynich Manuscript against 400 different languages to identify its likely source language (Hebrew)
-
Text restoration - DeepMind's "Ithaca" system restoring damaged Greek inscriptions with 62% accuracy (increasing to 72% when combined with human expertise)
-
Character prediction - "Pythia" AI model predicting missing characters in ancient texts, outperforming human experts with 30% fewer errors
-
Multilingual comparison - AI processing bilingual inscriptions (like the Kushan script found with Bactrian text) to establish translation references
-
Multispectral imaging analysis - AI combined with multispectral imaging to reveal erased texts in palimpsests without damaging manuscripts
-
Segmentation and reconstruction - AI used to analyze and segment Maya glyphs, reconstruct damaged text, and predict missing symbols
-
Statistical analysis - AI identifying the Copial Cipher as a masked German script by analyzing patterns within the text
The artifact contains a more comprehensive breakdown of these applications organized by technique.
Here are the ancient languages discussed in the video:
- 1. Herculanium scrolls
- 2. Voynich Manuscript
- 3. Copial Cipher
- 4. Linear Elamite
- 5. Tingua bamboo slips
- 6. Dead Sea Scrolls
- 7. Maya script
- 8. Etruscan
- 9. Linear A (Minoan)
- 10. Rongorongo
- 11. Kusan script
- 12. Olmec glyphs
- 13. Isthmian script
- 14. Kitan small script
- 15. Tangut script
- 16. Ancient Greek texts
- 17. Hidden texts in palimpsests
- 18. Indus Valley script
Detailed Discussions
1. Herculaneum Scrolls
Archaeological Context
The Herculaneum scrolls represent one of the most significant literary finds from the ancient world - the only surviving complete library from antiquity. These papyri were discovered in the Villa of the Papyri in Herculaneum, Italy, which was buried during the eruption of Mount Vesuvius in 79 CE. The intense heat carbonized the scrolls, preserving but also making them extremely fragile. The library is believed to have belonged to Lucius Calpurnius Piso Caesoninus, Julius Caesar's father-in-law (Smithsonian, 2024).
AI Application and Research
The breakthrough in reading these scrolls came through the Vesuvius Challenge, launched in March 2023 by computer scientist Brent Seales, along with entrepreneurs Nat Friedman and Daniel Gross. This $1 million competition employed AI and advanced imaging techniques to virtually unroll and read the carbonized scrolls without physically opening them.
Key developments included:
- High-resolution CT scans of the scrolls were taken at the Diamond Light Source particle accelerator in the UK
- Machine learning algorithms were trained to detect carbon-based ink on carbonized papyrus
- A virtual unwrapping technique was developed to digitally flatten the scrolls
In February 2024, a team of three students (Youssef Nader, Luke Farritor, and Julian Schilliger) won the $700,000 grand prize by deciphering over 2,000 Greek characters from the first scroll, revealing text believed to be by Epicurean philosopher Philodemus about the nature of pleasure (National Endowment for the Humanities, 2024).
The current goal for 2024 is to reveal 90% of the four scrolls that have been fully scanned, potentially recovering previously lost works from antiquity (Time, 2023).
Sources
- National Endowment for the Humanities. (2024). Students Decipher 2,000-Year-Old Herculaneum Scrolls
- Smithsonian Magazine. (2024). Three Students Decipher First Passages of 2,000-Year-Old Scroll Burned in Vesuvius' Eruption
- The Conversation. (2024). The Vesuvius Challenge is using AI to virtually unroll Pompeii's ancient scrolls
- Time. (2023). Inside the AI-Powered Race to Decode Ancient Roman Scrolls
- Wikipedia. (2025). Herculaneum papyri
2. Voynich Manuscript
Archaeological Context
The Voynich Manuscript is a mysterious illustrated codex hand-written in an unknown script, with its vellum carbon-dated to the early 15th century (1404-1438). The manuscript contains approximately 240 pages featuring illustrations of unidentified plants, astronomical symbols, and human figures. It was acquired by Polish book dealer Wilfrid Voynich in 1912 and is currently housed in Yale University's Beinecke Rare Book and Manuscript Library (Wikipedia, 2025).
AI Application and Research
In 2018, computer scientists Greg Kondrak and Bradley Hauer from the University of Alberta applied artificial intelligence to analyze the manuscript's language patterns. Their approach included:
- Developing algorithms to analyze the frequency patterns of letters and words
- Training AI on 400 different language samples from the "Universal Declaration of Human Rights"
- Initially hypothesizing Arabic origin, but concluding Hebrew was most likely the base language
- Proposing that the text was created using alphagrams (words with letters arranged alphabetically)
Their AI analysis suggested that over 80% of the words matched Hebrew dictionary entries. After some spelling corrections, the first line was tentatively translated as "She made recommendations to the priest, man of the house and me and people" (University of Alberta, 2018).
However, medieval scholars and Voynich manuscript experts have expressed skepticism about these findings. Lisa Fagin Davis, executive director of the Medieval Academy of America, noted that an algorithm trained on modern languages cannot reliably identify a 15th-century text, especially given differences in grammar, spelling, and vocabulary (Artnet News, 2018).
Sources
- University of Alberta. (2018). Using AI to uncover the mystery of Voynich manuscript
- Artnet News. (2018). Did Artificial Intelligence Really Decode the Voynich Manuscript? Some Leading Scholars Doubt It
- Smithsonian Magazine. (2018). Artificial Intelligence Takes a Crack at Decoding the Mysterious Voynich Manuscript
- Wikipedia. (2025). Voynich manuscript
3. Copiale Cipher
Archaeological Context
The Copiale Cipher is an encrypted manuscript consisting of 75,000 handwritten characters filling 105 pages in a bound volume. Dating from the 1730s, it remained undeciphered for over 260 years until 2011. The manuscript was bound in gold and green brocade paper, with the only plaintext being "Copiales 3" at the end and "Philipp 1866" on the flyleaf (Wikipedia, 2025).
AI Application and Research
In 2011, an international team led by Kevin Knight of the University of Southern California's Information Sciences Institute successfully decrypted the Copiale Cipher using computational linguistics approaches:
- The researchers transcribed a machine-readable version of the text
- They developed pattern recognition algorithms to analyze the cipher's structure
- Their analysis revealed that unaccented Roman letters in the text served only as spaces
- The actual content was encoded by accented Roman letters, Greek letters, and abstract symbols
The deciphered text revealed the rituals and political leanings of an 18th-century German secret society called the "Oculists" (Hocherleuchtete Oculisten-Orden) of Wolfenbüttel. The manuscript describes initiation ceremonies that symbolically focus on vision and eye operations, including a ritual where a candidate is asked to read a blank paper, given eyeglasses, and has a single eyebrow hair plucked (USC Today, 2023).
The breakthrough demonstrated how AI-assisted cryptographic techniques could unlock historical secrets that had resisted traditional decipherment methods for centuries.
Sources
- Wikipedia. (2025). Copiale cipher
- USC Today. (2023). USC Scientist Cracks Mysterious "Copiale Cipher"
- CS Monitor. (2011). Copiale Cipher: How a secret society's code was finally cracked
- Culver City Foshay Lodge № 467. The Oculists and the Copiale Cipher
4. Dead Sea Scrolls
Archaeological Context
The Dead Sea Scrolls represent the oldest known biblical manuscripts, discovered in caves near Qumran in the Judaean Desert beginning in 1947. Dating from the 3rd century BCE to the 1st century CE, they include biblical texts, religious writings, and community rules. The Great Isaiah Scroll (1QIsaa) contains the entire Book of Isaiah and predates other Hebrew manuscripts of Isaiah by over 1,000 years (Smithsonian Magazine, 2021).
AI Application and Research
In 2021, researchers from the University of Groningen's Qumran Institute published a groundbreaking study in the journal PLOS ONE using AI to analyze the handwriting in the Great Isaiah Scroll:
- They trained an artificial neural network to distinguish ink from the parchment background
- The AI identified subtle variations in handwriting that were imperceptible to the human eye
- Pattern recognition techniques analyzed specific letter shapes and features across the manuscript
- Statistical analysis revealed that the manuscript was likely written by two different scribes with similar styles
- A clear transition point was identified after column 27 (halfway through the scroll)
The research demonstrated that what appeared to be a uniform handwriting style actually contained subtle differences that only AI could detect. This finding supports the theory that the Dead Sea Scrolls were produced by teams of scribes who were trained to write in a standardized style, with apprentices possibly learning to mimic the writing of senior scribes (Live Science, 2021).
Sources
- PLOS ONE. (2021). Artificial intelligence based writer identification generates new evidence for the unknown scribes of the Dead Sea Scrolls exemplified by the Great Isaiah Scroll (1QIsaa)
- Live Science. (2021). Mysterious second writer of Dead Sea Scroll uncovered by AI
- Smithsonian Magazine. (2021). How A.I. Is Helping Scholars Unlock the Secrets of the Dead Sea Scrolls
- The Conversation. (2024). Dead Sea Scrolls: two scribes probably wrote one of the manuscripts
5. Indus Valley Script
Archaeological Context
The Indus Valley script is one of the most enigmatic ancient writing systems, used by the Indus Valley Civilization that flourished between 2600 and 1900 BCE in what is now Pakistan and northwest India. The script appears on thousands of small seals, tablets, and amulets, with typical inscriptions containing 5-6 symbols from a set of approximately 400-700 unique characters. This civilization was highly advanced, with sophisticated urban planning, drainage systems, and long-distance trade networks (Rest of World, 2023).
AI Application and Research
Multiple research teams have been applying AI and computational methods to analyze the Indus Valley script:
- In 2009, a team led by Rajesh Rao at the University of Washington used pattern recognition algorithms to analyze symbol placement sequences, creating a statistical model for the unknown language (University of Washington, 2009)
- AI methods have been used to fill in missing symbols on damaged artifacts, increasing the pool of data available for decipherment
- Computer scientists have trained machine learning models to identify linguistic patterns and potential relationships to known ancient languages like Proto-Dravidian
- In January 2025, the government of Tamil Nadu in India offered a $1 million prize for anyone who can decipher the script, stimulating renewed interest in AI approaches (Archaeology Magazine, 2025)
Despite these efforts, the Indus script remains undeciphered due to several significant challenges: the brevity of inscriptions (most under 5-6 characters), the absence of a bilingual text like the Rosetta Stone, and disagreement about whether it represents a true writing system or merely a collection of religious or political symbols (Language Log, 2022).
Sources
- University of Washington. (2009). Computers unlock more secrets of the mysterious Indus Valley script
- ResearchGate. (2025). Deciphering the Indus Valley Script: New Approaches and AI-Driven Insights
- Discover Magazine. (2023). Could AI Language Models Like ChatGPT Unlock Mysterious Ancient Texts?
- Rest of World. (2023). An ancient language has defied decryption for 100 years. Can AI crack the code?
- Archaeology Magazine. (2025). $1 million prize offered to decipher 5,300-year-old Indus Valley script
6. Rongorongo Script of Easter Island
Archaeological Context
Rongorongo is a unique writing system discovered in the 19th century on Easter Island (Rapa Nui). The script consists of glyphs carved into wooden tablets, staffs, and other artifacts using shark teeth or obsidian flakes. Only about two dozen objects bearing rongorongo inscriptions have survived, all collected in the late 19th century and now scattered in museums worldwide. The system is read in a reverse boustrophedon pattern (alternating direction with each line), starting from the bottom left corner (Wikipedia, 2025).
AI Application and Research
While there haven't been widely publicized AI breakthroughs with Rongorongo specifically, computational approaches are being developed:
- Researchers have used 3D modeling and photogrammetry to create digital models of the tablets, revealing glyphs that were previously invisible to the naked eye (HeritageDaily, 2021)
- On GitHub, a project called "rongopy" uses computational approaches to test decipherment hypotheses, including frequency analysis of glyphs and comparison with syllable patterns in Rapa Nui language (GitHub, 2022)
- Statistical analyses have been used to determine if Rongorongo follows linguistic patterns that would indicate it is a true writing system rather than proto-writing or a mnemonic device
The debate continues about whether Rongorongo represents true writing or is a form of proto-writing (a mnemonic device for genealogy, choreography, navigation, or religious recitation). If it does prove to be an independently invented writing system, it would be one of very few such inventions in human history. While there has been some progress in identifying portions of text that may represent lunar calendars or genealogical information, the script as a whole remains undeciphered (Wikipedia, 2025).
Sources
- Wikipedia. (2025). Rongorongo
- Wikipedia. (2025). Decipherment of rongorongo
- GitHub. (2022). jgregoriods/rongopy: Ideas for the decipherment of Easter Island's rongorongo writing
- HeritageDaily. (2021). 3D technology reveals the mysterious Rongorongo language of Easter Island
- Big Think. (2024). Mysterious writing system from Easter Island may be completely unique
7. Maya Hieroglyphs
Archaeological Context
Maya hieroglyphs represent one of the most sophisticated writing systems developed in ancient Mesoamerica. This script was used by the Maya civilization from at least the 3rd century BCE until the Spanish conquest in the 16th-17th centuries. The writing appears on stone monuments, pottery, murals, and the few surviving Maya codices. The Maya script is logosyllabic, using both logograms (representing whole words) and syllabic signs that function somewhat similarly to modern Japanese writing (Wikipedia, 2025).
AI Application and Research
Several research initiatives have applied computational and AI methods to Maya hieroglyphs:
- Researchers at the University of Groningen have developed AI models to segment and identify individual glyphs on Maya ceramic vessels, using fine-tuned foundation models like the Segment Anything Model (SAM) (arXiv, 2025)
- Computer scientists have created digital shape descriptors (HOOSC - Histogram of Orientations - Shape Context) to analyze and compare the forms of Maya glyphs for categorization and identification (Oxford Academic, 2017)
- Interactive visualization tools have been developed that allow both expert epigraphers and novices to explore and learn about Maya writing, incorporating glyph co-occurrence information to assist in identifying partially damaged glyphs
A significant portion of Maya writing has been deciphered through traditional scholarly methods, with approximately 85% of known glyphs now understood. The decipherment accelerated in the 1950s through the work of Yuri Knorozov, who demonstrated the writing system was partially phonetic, and Tatiana Proskouriakoff, who showed that the inscriptions recorded historical events rather than merely astronomical information. AI and computational methods are now helping to advance this work further by analyzing patterns, restoring damaged texts, and making the corpus more accessible to researchers.
Sources
- arXiv. (2025). Segmentation of Maya hieroglyphs through fine-tuned foundation models
- Oxford Academic. (2017). Analyzing and visualizing ancient Maya hieroglyphics using shape: From computer vision to Digital Humanities
- World History Encyclopedia. (2015). How to Read a Maya Glyph
- ushistory.org. Deciphering Maya Glyphs
- Wikipedia. (2025). Maya script
8. Linear B
Archaeological Context
Linear B is a syllabic script used for writing Mycenaean Greek, the earliest attested form of the Greek language, dating from around 1450 BCE. The script was primarily found on clay tablets discovered at Knossos, Pylos, Thebes, and other Mycenaean palace archives. It disappeared with the fall of Mycenaean civilization during the Late Bronze Age collapse (around 1200 BCE), leading to the Greek Dark Ages during which there is no evidence of writing (Wikipedia, 2025).
AI Application and Research
Linear B was famously deciphered in 1952 by Michael Ventris, an English architect and self-taught linguist, making it the only fully deciphered Bronze Age Aegean script. Modern AI and computational methods are now being applied to the script in several ways:
- Researchers have developed a generative neural language model (Bidirectional Recurrent Neural Network) to capture the statistical structure of Mycenaean documents written in Linear B (ACM, 2023)
- This AI model is being used to supplement damaged parts of Mycenaean texts by predicting missing or incomplete words on partially damaged clay tablets
- Machine learning approaches have been used to automatically translate Linear B texts, with researchers from MIT and Google's AI lab reporting in 2019 that they were able to correctly translate 67.3% of Linear B cognates into their Greek equivalents using their neural decipherment system (MIT Technology Review, 2019)
These computational approaches to Linear B are not only helping to fill gaps in our understanding of Mycenaean texts but are also serving as test cases for techniques that might be applied to still-undeciphered scripts like Linear A.
Sources
- ACM. (2023). A Generative Model for the Mycenaean Linear B Script and Its Application in Infilling Text from Ancient Tablets
- MIT Technology Review. (2019). Machine learning has been used to automatically translate long-lost languages
- Cambridge University. The Decipherment of Linear B: Introduction
- Wikipedia. (2025). Linear B
- Britannica. Linear A and Linear B | Mycenaean, Minoan & Decipherment
9. Linear A
Archaeological Context
Linear A was used by the Minoan civilization on Crete between approximately 1800 and 1450 BCE, predating Linear B. It was discovered by British archaeologist Sir Arthur Evans during excavations at Knossos and other Minoan sites. The script has been found primarily on clay tablets, seals, and stone offering tables. Although Linear A and Linear B share many similar signs, Linear A remains undeciphered because the language it records is unknown (Wikipedia, 2025).
AI Application and Research
Researchers are increasingly turning to AI and computational methods to tackle the challenges of Linear A:
- Computational pattern recognition techniques are being applied to analyze the symbols of Linear A, using natural language processing and data mining methods to identify statistical and information-theoretic patterns (HAL Science, 2021)
- Researchers have employed machine learning to compare consonant clusters in Linear A (using phonetic values from Linear B as a provisional assignment) with consonant clusters from various language families to test potential connections (MDPI, 2023)
- AI approaches are being used to compare Linear A with other writing systems visually, using feature-based similarity measures to develop new potential phonetic grids for the script
- A "brute force" method has been proposed using computational technology to systematically test different decipherment hypotheses against language families from the Mediterranean and Black Sea regions (Ancient Origins, 2024)
Despite these technological advances, Linear A remains resistant to decipherment due to fundamental challenges: the limited corpus (only about 1,400 inscriptions), the lack of a bilingual text, and uncertainty about the language being recorded. Some researchers believe the corpus is simply too small to ever achieve a proven decipherment unless new texts or a Rosetta Stone-type artifact is discovered (Language Log, 2023).
Sources
- HAL Science. (2021). Computational Pattern Recognition in Linear A
- MDPI. (2023). Minoan Cryptanalysis: Computational Approaches to Deciphering Linear A
- Ancient Origins. (2024). Linear A and The Machine: a Brute Force Attack to Decrypt the Minoan Code
- Greek City Times. (2023). Cracking the Code: Advancements in AI Bring New Hope for Deciphering Linear A
- Wikipedia. (2025). Linear A
10. Linear Elamite
Archaeological Context
Linear Elamite was a writing system used in Elam (modern-day southwestern Iran) during the Bronze Age between approximately 2300 and 1850 BCE. It appears mainly on monumental inscriptions, including ceremonial objects and architectural elements. The script was discovered during French excavations at Susa beginning in 1903. Linear Elamite was used contemporaneously with Elamite cuneiform to record the Elamite language, a language isolate unrelated to neighboring Semitic or Indo-European languages (Wikipedia, 2025).
AI Application and Research
While AI hasn't played a central role in Linear Elamite research until recently, a major breakthrough in decipherment occurred in 2022 through traditional linguistic methods:
- French archaeologist François Desset and colleagues published a near-complete decipherment of Linear Elamite in 2022 after years of study
- Their work began in 2017 and was aided by the discovery of bilingual inscriptions on silver beakers that provided crucial comparative material
- The team identified 72 distinct signs in the Linear Elamite syllabary and determined that the script was primarily syllabic in nature
- While AI wasn't central to this breakthrough, computational methods are now being used to validate the proposed decipherment and apply it to newly discovered inscriptions
This decipherment has significant implications for understanding the linguistic and cultural history of the Iranian plateau and surrounding regions. The Elamite language, previously known primarily through cuneiform sources, now has an expanded corpus of texts that may reveal new insights into its grammar, vocabulary, and development over time.
Sources
- Wikipedia. (2025). Linear Elamite
- Language Log. (2022). Decipherment of Linear Elamite
- ResearchGate. (2022). The Decipherment of Linear Elamite Writing
- Smithsonian Magazine. (2023). Have Scholars Finally Deciphered Linear Elamite, a Mysterious Ancient Script?
- Ancient Near East Today. (2024). Breaking the Code: Ancient Iran's Linear Elamite Script Deciphered
11. Olmec Script
Archaeological Context
The Olmec civilization flourished in the tropical lowlands of Mexico's Gulf Coast region between 1500 and 400 BCE and is considered the earliest major Mesoamerican civilization. Evidence of Olmec writing includes symbols found on monuments, stone tablets, celts, and other artifacts. The most significant discovery came in 2006 with the Cascajal Block, a serpentine tablet containing 62 symbols unlike any previously known Mesoamerican script, dated to approximately 900 BCE, making it potentially the oldest writing system in the Americas (Wikipedia, 2024).
AI Application and Research
Research on Olmec hieroglyphs is still in its early stages, with AI and computational methods beginning to be applied:
- Researchers are developing pattern recognition algorithms to analyze Olmec symbols across different sites and artifacts to identify consistent usage patterns
- Statistical modeling is being used to determine if the symbols represent a true writing system or are more limited symbolic or iconic representations
- Comparative analyses between Olmec symbols and later Mesoamerican scripts (such as Maya, Zapotec, and Isthmian) are being conducted using computational methods to identify potential evolutionary relationships
- Digital imaging and 3D modeling technologies are being used to create more accurate records of weathered or damaged Olmec inscriptions
There remains significant debate about whether Olmec hieroglyphs constitute a true writing system or are more limited in scope. Some researchers see them as precursors to later Mesoamerican writing systems, while others argue they may have functioned as mnemonic devices or emblems rather than encoding language. The limited corpus of Olmec inscriptions, coupled with their antiquity and the absence of bilingual texts, makes decipherment particularly challenging (Mesoamerican writing systems, 2025).
Sources
- Wikipedia. (2024). Olmec hieroglyphs
- PubMed. (1993). A decipherment of epi-olmec hieroglyphic writing
- Science. A Decipherment of Epi-Olmec Hieroglyphic Writing
- Ancient Origins. (2020). Hidden in the Glyphs: Deciphering Bilingual Mayan-Olmec Text
- Wikipedia. (2025). Mesoamerican writing systems
No comments:
Post a Comment