AI Just Reexamined the Rosetta Stone — What It Found Changes Everything

追加: 2026-05-14

[00:00:00]onto a story that comes from Egypt. It's about a clear message that the country is sending out regarding the return of artifacts held in Britain's museums that it is not looking for the return of these artifacts, but it's also very clear that it wants the return of the Rosetta Stone.

[00:00:17]>> For 200 years, every Egyptologist on Earth agreed on what the Rosetta Stone said. Then a research team fed the triilingual inscription into an AI model and asked it to do something nobody had thought to do before.

[00:00:30]>> I think all nations um have the right to ask back for their um heritage in the 19th century.

[00:00:37]>> What came back stopped them cold. The algorithm flagged a pattern buried in the carved surface that two centuries of human scholarship had walked straight past. And the moment you understand what it found, you cannot look at this stone the same way again. It is sitting in the British Museum right now. And it is no longer saying what we thought it said.

[00:00:58]The stone that rewrote history. More than 2,000 years ago, somewhere in ancient Egypt, a group of highly trained scribes carved a message into a slab of dark granodorite. They were carving for kings, for priests, for an audience they could not imagine. What they left behind would sit buried for nearly 15 centuries until July of 1799 when a French soldier spade struck something hard in the mud near the Egyptian port town of Rosetta.

[00:01:27]Napoleon Bonapart's army had swept into Egypt the previous year alongside a team of scholars cataloging Egypt's past.

[00:01:35]>> Today we are in the 21st century and we have to correct the mistakes of the past. When that spade hit stone near the town known today as Rashid, no one understood what had just been uncovered.

[00:01:47]Stay with this. He did not find gold. He did not find a tomb. He found a broken slab covered in three different kinds of writing. It looked like nothing. It was the key. The British seized the stone from Napoleon's forces under the 1801 Treaty of Alexandria. It has been displayed continuously at the British Museum since 1802. For two centuries, everyone agreed on what it was. A key that had already given up everything it had to give. That was the consensus. And the algorithm sitting on Pirelli's desk has just put a print out through it. A language the world had forgotten. Most people think the Rosetta Stone was a translation puzzle one Frenchman cracked. They have no idea how dark it had gotten before he got there.

[00:02:31]>> We can uh uh crack the code to the secrets of Egypt. The [music] last person known to have read Egyptian hieroglyphics in the ancient world died around 394 AD. After that, the knowledge simply ceased to exist. The Roman world took over. Christianity spread. The old Egyptian temples were closed, repurposed, or left to crumble. 5,000 years of recorded history, theology, and royal account became unreadable. For over a millennium, brilliant minds tried to crack it and produced nothing usable.

[00:03:04]The persistent mistake was assuming hieroglyphics were purely symbolic. That each image stood for a concept rather than a sound. A hawk meant the soul. A circle meant eternity. It felt profound.

[00:03:17]It was a complete dead end. The truth was far more complex. Hieroglyphics combined phonetic signs with ideographic elements and semantic classifiers.

[00:03:28]Without a key to a known language, there was no way through. For 15 centuries, the wall held firm. Then a French soldier's spade cracked it open. Just enough. But the strangest part of this story is not what the men of 1822 found inside the crack. It is what they missed. Three scripts, one assumption.

[00:03:49]When French scholars first examined the stone, what struck them was the structure. Not one script, three. At the top, hieroglyphics, the formal sacred script. In the middle, deodic, the fluid form of Egyptian writing used for contracts and personal letters. At the bottom, ancient Greek, the administrative language of TMIC Egypt, when the pharaohs were Greekeaking [music] descendants of one of Alexander the Great's generals. And here is the assumption every scholar made instantly >> after working from the known Greek language to the unknown languages for some 20 years. It's three languages, one message, a faithful translation chiseled into rock. That assumption is what the AI on Pirelli's desk has just refused to make. The Greek was a decree issued in 196 BC by a council of Egyptian priests honoring the young pharaoh the fifth Epiphanies on the first anniversary of his coronation. tax reductions, temple donations, military victories, bureaucratic text. Because it was readable, for the first time in 1500 years, someone had a map. Swedish diplomat Johan Ukerblad made early progress identifying proper names in the demonic. The first decisive breakthrough belonged to British physicist Thomas Young in London in 1814. Young noticed that one group of hieroglyphic symbols kept appearing inside an oval loop, a cartou. Wherever the Greek used the name tammy, the symbols inside were not ideas. They were sounds. That was the crack, the complete framework came from French linguist Jean Francois Champolon.

[00:05:29]He had mastered Coptic, the final surviving form of the Egyptian language preserved in Egyptian Christianity's liturgy. Coptic kept the actual sounds of ancient Egyptian alive in a Greek alphabet. Champolon used it as an acoustic bridge to 15 centuries of silence. When the last piece locked into place in September of 1822, Champolon stumbled out of his Paris study. Grabbed his brother Jacqu Joseph and said four words. Jatimmon affair.

[00:05:58]I've got it. Then he collapsed. He was unconscious for five full days. Hero falls. Code is broken. Every textbook stops the story right there. But Champolon's collapse is not the climax of this story. It is the doorway.

[00:06:13]>> He came upon this this rather large rock.

[00:06:16]>> The assumption that let him crack the script that the three texts were saying the same thing is the exact assumption the algorithm at University College London has just refused to make. If you want to keep following stories like this, where the ancient world and the newest technology on Earth collide and the answers stop being simple, subscribe right now and hit the bell. What comes next is the part Champolon could not have seen, Young could not have seen, and two centuries of scholars after them never thought to ask. It took something that does [music] not think like a human to even ask the question. The AI enters the room. Marco Pirelli is back at his desk. Different day, same chair. For weeks, he and his team at University College London have been doing something nobody had thought to do with the Rosetta Stone before, not translating it. Two centuries of scholarship had already done that. Instead, they had instructed the model to treat the inscription as a data structure, to map statistical relationships between the three language versions simultaneously, to treat the hieroglyphic, the deodic, and the Greek as three independent witnesses to the same event. That is the part you need to hold on to. Three witnesses, not one document in three voices. Three witnesses who might or might not agree. Every human scholar for 200 years had assumed they agreed. The algorithm did not assume. The algorithm asked. That is a small distinction in language and an enormous distinction in method. For two centuries, scholars had been answering the question. Nobody had thought to first verify the question was the right one. The first time it returned its results, the team was disoriented. They had expected nothing, a confirmation the existing translation was complete, or scattered statistical noise. What the model produced instead was something with a shape. The flag divergences were not random. They formed a pattern visible immediately once the output was visualized. Pirelli printed it out, held it at arms length, said three words. Run it again. It ran again.

[00:08:21]Same places, same category, same direction. Run it again. Same places.

[00:08:27]That is when he sat down. That is when his colleague brought him coffee he did not drink. That is the print out you watched him stare at on a desk in Bloomsberry at the top of this video. He has read this stone. He has read every published scholarly paper on this stone.

[00:08:42]He thought he knew this stone. The algorithm just told him he did not. Now you need to see what the algorithm actually flagged. Stay with this. It is concrete. In section 14 of the decree, the passage describing the king's donations to the temples. The hieroglyphic version uses a specific formulation that appears three times in close proximity. In standard Egyptian ceremonial language of the tameic period, that formulation appears once per passage. Jean Francois Jean Polon both these men knew that cracking hieroglyphs >> repetition of that construction in priestly administrative texts signals a legal binding clause, a formal commitment carrying contractual weight.

[00:09:23]The Greek text at the same location uses a single honorific phrase, gracious, routine. The kind of sentence a speech writer produces on autopilot. Here is the gap. Where the Greek says the king gave generously to the temples, the hieroglyphic section appears to say, "The king is bound to give to the temples permanently in a way that carries legal force under priestly law."

[00:09:47]That is not a translation difference.

[00:09:49]That is not grammatical variance. That is a different message delivered in the same document at the same point to two different audiences. The hieroglyphic text and the Greek text are not saying the same thing. Let that sink in. For 200 years, every textbook said they were. Every museum label said they were.

[00:10:09]Every translation said they were. The algorithm asked the one question no human scholar had ever thought to ask first. Do they actually agree? And the answer that came back was no. Not random. No. Not scattered. No.

[00:10:22]Patterned. No. 17 instances. Same category, same direction. The divergences cluster in a single category of passage. Every instance where the decree touches the relationship between royal authority and priestly economic rights. In passages about military campaigns and public construction, the three language versions align almost perfectly. The moment the decree discusses temple authority, priestly land rights, and the limits of royal power over sacred property, that is where the algorithm consistently identified structural divergences with no linguistic explanation.

[00:11:00]17 instances, one category, one direction, and we are not done. what the pattern means. In ancient Egypt, layered communication in official texts would not have been unusual if it was intentional. Egyptian scribes mastered the art of what scholars call dlossia, the ability to use language differently for different audiences within the same document. Nobody had ever looked at the Rosetta Stone through that lens of what is known today as the Rosetta Stone, which is housed today in the British Museum in [music] London. The universal assumption was that the three scripts were faithful translations of each other, equivalents. The algorithm refused to make that assumption. It treated the three versions as three independent witnesses and asked one simple question. Do they agree? They did not. And not in the places that mattered most politically. Look at what that means. The TMIC pharaohs were navigating an extraordinarily delicate situation.

[00:12:01]foreign rulers, Greekeaking descendants of a Macedonian general trying to maintain legitimacy with an ancient deeply conservative Egyptian priestly class that held enormous institutional power. The priests controlled the temples. The temples controlled vast agricultural lands, workshops, and economic infrastructure. The pharaohs needed the priests public endorsement.

[00:12:24]The priests needed guarantees. Here is where it gets uncomfortable. Those guarantees may have been written directly into the Rosetta Stone, not in the Greek text the public read, not in the demonic [music] everyday people read in the hieroglyphic layer only the priestly class could fully interpret.

[00:12:40]The stone may not have been one decree.

[00:12:42]It may have been two different deals carved into the same slab for two different rooms, the same monument, two different contracts, two different audiences, and the only people who could read both layers were the people who wrote them. The algorithm did not arrive at this story by intuition. It arrived at it by refusing to assume what every human reader assumed first. That is the part that should make you stop scrolling. You have to ask yourself who that benefits. The objection from Durham.

[00:13:12]Not everyone is convinced and the push back deserves serious weight. 300 m north of Perili's office at Durham University, Dr. Penelopey Wilson has spent decades inside tmic priestly texts. Her bookshelves are walls of comparable decrees, photographs of carved cartes, transliterated demonic margins. She is one of the few specialists in the world fluent in the literary conventions of exactly the period the Rosetta Stone comes from. And she is not buying it. Her objection lands in a specific place. Phrase repetition in hieroglyphic ceremonial language, she argues, is far more stylistically conventional than the algorithm appears to account for.

[00:13:54]Egyptian scribes used formulaic repetition as a rhetorical device, not necessarily as a legal or coded signal.

[00:14:01]Without a much larger corpus of comparable toic era, priestly decrees analyzed under the same conditions. She argues the statistical baseline is too narrow to support the divergence claims.

[00:14:14]Translation: The algorithm is finding a pattern. But a pattern is not the same as a signal. You have to take that seriously. Wilson is not arguing from gut feel. She is arguing from a career of reading exactly these kinds of decrees in their exact ceremonial context. [music] And then there is the deeper objection. Algorithms trained to find patterns will find patterns.

[00:14:35]Whether those patterns mean anything or not. The question is not whether anomalies exist. They always do. The question is whether they are intentional and whether intent can be inferred [music] by an algorithm at working across a cultural gap of 2,000 years.

[00:14:50]The model's training data also matters.

[00:14:53]Modern large language models learn from existing scholarship. Scholarship already mediated through centuries of human interpretation. If those texts [music] were themselves imperfectly understood, the AI's comparison may rest on a shakier foundation than it first appears. But here is what even Wilson conceds. The algorithm identified [music] specific phrase clusters and cross- language divergences in passages about temple authority and priestly economic rights that the existing scholarly literature has never adequately explained. The features are present. The debate is what caused them.

[00:15:28]And here is what changes everything if Pirelli is right. What it would mean.

[00:15:33]Pirelli is back at his desk. Different week, same chair. There is a stack of comparable toic decrees on his right.

[00:15:41]There is a phone [music] face down that he has stopped answering. He has not slept well in days. The printout is still there. If careful future scholarship confirms what the algorithm appears to be telling him, the implications stretch far beyond a single [music] artifact in London. Every major inscription from the Tomic period would need to be subjected to the same computational scrutiny. Hundreds of established [music] translations would be open to revision. The model is not just suggesting we misread one decree.

[00:16:10]It is suggesting we may have misread an entire category of ancient political communication. That is the part Pirelli says when he says anything that keeps him in his chair [music] past midnight.

[00:16:20]It would force a reassessment of the ancient Egyptian priestly class. The idea that religious officials [music] engineered formal royal proclamations to serve their own interests while appearing simply to honor their king suggests strategic intelligence traditional histories have consistently underestimated. These were architects of political language sophisticated enough to fool two centuries of modern scholarship. And here is where it gets really uncomfortable. It raises a question about the TMIC pharaohs themselves. If they participated in the layered construction of their own official documents, that is sophisticated political theater performed in three languages simultaneously for three different audiences on one piece of stone. The famous cultural synthesis of the TMIC period may not have been merely performance. It may have been negotiation encoded directly into the monuments they left behind. The algorithm did not invent any of that.

[00:17:15]Pirelli did not invent it either. The stone was always carrying it. For 2,000 years, it was simply waiting for a reader who would not start with the assumption that finished it. And there is one more thing the algorithm noticed about the stone. Something the scribes themselves may have been counting on for 2,000 years. The stone still speaks. It sits today in the British Museum in London inside a climate controlled [music] case under precise conservation lighting. a broken slab of dark grano diorite, roughly the size of a large gravestone, covered in three bands of ancient text. Thousands of visitors pause in front of it every day. Most look for about 45 seconds. They read the label. They take a photo. They move on.

[00:17:59]They think they know what it is. For two centuries, the Rosetta Stone has been celebrated as the key that unlocked the ancient Egyptian written world. That story is true. Champolon's decipherment remains one of history's most remarkable intellectual achievements. Now the key itself is under examination. The scribes who carved it were professionals of extraordinary skill, operating in a world where language was the primary technology of power, where writing [music] was controlled by a tiny educated elite, where the distance between what a document said and what it meant could be precisely, deliberately, and invisibly engineered. They understood that most people would read the surface. [music] They may have been counting on it. For 2,000 years, the layer they built beneath the official record sat untouched, waiting for something that could read it without two centuries of human consensus, telling it what the answer was supposed to be. What the AI cannot tell us is whether those scribes had any idea their work would outlast everything. whether they could have imagined their inscription being interrogated by a form of intelligence they had no way to conceive. One that would look at their carefully constructed document and immediately ask the one question no human scholar had thought to ask first. Are all three versions actually saying the same thing?

[00:19:17]What the algorithm has already told us is that the translation we have been pointing to for 200 years. As proof, we crack the code may be only the most visible layer of what is actually carved into that stone. The layer designed to be read by anyone. There may be one or two more layers underneath it. Layers carved for audiences we are only now beginning to identify, by people who knew exactly what they [music] were doing. Stay with that for a second. The Rosetta Stone is not just a translation key. It may be a diplomatic transcript, a financial agreement, and a public relations document all at once, all on one slab. And the only reason we can begin to see those layers now is that something looked at the stone without human assumption and asked. The ancient world left us a stone. And a question it was never fully asked to answer. Now that we have recovered the language to read the surface, the algorithm is asking the harder question. What if you only ever read the version they wanted most people [music] to see? If this story made you think differently about what we know and what we only think we know, subscribe and hit the bell. Every week we go [music] exactly here. The place where the ancient world and the newest technology on Earth collide and the answers stop being simple. The Rosetta Stone [music] has been waiting 2,000 years to be properly understood.

[00:20:34]Something tells us we are really only just getting started.