The gap in what we know about the ancient Romans can be met by AI

AI News


Pallab Ghosh profile images
Metropolitan Museum of Art This is an image of a rectangular bronze inscription broken in half. The text on the right is missing, but it is shown as being restored by AI in the predicted textMetropolitan Museum of Art

Artwork: New system predicts broken inscriptions, missing text from its date, original location

New AI tools could help you understand all human history, researchers say.

Artificial intelligence is already being used to fill the gaps in ancient Roman scrolls, but the new system is moving further.

It can be filled with missing words from ancient Roman inscriptions engraved on monuments and everyday objects, dated and geographically placed.

There is concern that AI often introduces errors into the analysis of simple modern texts, and that it may rely too much on this technology to distort rather than enhance understanding of history.

However, Professor Mary Beard, a historian at Cambridge University, explains that the technology could be “transformative” in the study of past events.

She said that the system, called Aeneas, after Greek and Roman mythical figures, can accelerate the speed at which historians piece together the past from ancient texts.

“This extremely challenging field breakthrough tended to rely on the memory, subjective judgments and reasoning of individual scholars supported by traditional encyclopedia databases. Eneas opens a whole new perspective.”

Atilimtunesbaydin is a large white stone monument, once a rectangular temple, but only three other broken, complete walls. atilimgramsbaydin

The inscriptions can be found on ancient monuments around the world. Located in the Temple of Augustus in Ankara, Turkey, this is known as the Queen of Inscriptions

Ancient inscriptions are usually incomplete, with unknown origin and date, and often all three. Historians and classicists try to fill in the blank by drawing texts that resemble words, grammar, appearance, and cultural settings known as “similarities.” Ancient inscriptions tend to be formulaic, so historians can often guess what the missing parts of the sentence say.

The process can be laborious and can take months and years, but it opens a new view of understanding the past, according to Dr. Thea Sommerschield, a historian at the University of Nottingham who co-led the research.

“The inscriptions are an early form of writing. They are invaluable to historians as they provide direct evidence of ancient history, language and society.

“But they deteriorate over the centuries, and interpreting them is like solving a huge jigsaw puzzle with tens of thousands of pieces, of which 90% is lost.”

Square stone plaque missing large chunks on the right side of Thea SommerschieldThea Sommerschield

The Roman bronze military diploma is in 94 AD. Like most inscriptions, it is incomplete

It is not the first time that AI has been used in Roman history to combine missing dots.

Earlier this year, another team of scientists used a combination of x-ray imaging and AI to digitally “package” a hellishly burnt scroll from the Roman town of Herculaneum, revealing lines and columns of text.

Dr. Sommerschield developed Aeneas with Dr. Yannis Assael, an AI specialist at Google Deepmind. Automate the process of contextualization based on gaze and gaze.

According to Dr. Aneas, Aeneas draws a vast database of 176,000 Roman inscriptions, including images, and uses a carefully designed AI system to raise broad historical similarities to support the work of historians.

“What historians can't do is assess these similarities in seconds across tens of thousands of inscriptions, which means that AI can come as assistants.”

The team tested the system when dating the famous Roman texts at the Temple of Augustus in Ankara, Turkey, known as the Queen of Inscriptions, because it is important for understanding Roman history. Res gestae Divi Augusti Composed by the first Roman Emperor Augustus, he described his life and achievements. The date is hotly contested among historians.

Aeneas was able to narrow the options down to two possible ranges. The most likely one is between 10-20 BC, unlikely to range from 10-1 BC. This demonstrated the accuracy of the system, as most historians agree that these two are the most likely possibilities.

An ancient Greek inscription with Google Deepmind Anion Blue Numbers is engraved to illustrate the application of AIGoogle DeepMind

Artwork: An earlier version of the system filled with missing words in ancient Greek. Eneas is much stronger

In testing the system with 23 historians, the team discovered that the historians working with Aeneas came up with more accurate results than Aeneas or historians themselves.

“The feedback was that Aeneas not only allowed historians to accelerate their work, but also revealed similarities that were previously unspecified,” says Dr. Sommershield.

“And that's the future value of this job, not only does it faster and better, it's also doing things we didn't think we'd do before.”

There is concern that even AI modern texts can have glitches, and that they may make mistakes. However, according to Dr. Assael, Aeneas is a tool to guide historians, not their place of replacement.

“We acknowledge that AI may not always be able to get everything right and don't think historians will work under that expectation,” he said.

He said it depends on the human historian to weigh Aeneus' predictions and determine what makes more sense.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *