#046 Saturday Feature: AI Is Transforming The Way We Study History - From Ancient Inscriptions To Medieval Times.
Fresh & Hot curated AI happenings in one snack. Never miss a byte 🍔
This snack byte will take approx 5 minutes to consume.
AI BYTE 📢 - AI Is Transforming The Way We Study History - From Ancient Inscriptions To Medieval Times
⭐History is often said to be written by the victors, but what if we could use artificial intelligence to uncover the hidden voices and perspectives of the past?
In this article, I will explore how historians are using machine learning, especially deep neural networks, to analyze historical documents and data in new and exciting ways.
Machine learning is a branch of artificial intelligence that uses algorithms to learn from data and make predictions or decisions. Deep neural networks are a type of machine learning that can process large and complex data sets by mimicking the structure and function of the human brain.
These networks can identify patterns, extract information, and generate outputs that would otherwise be impossible or impractical for humans to do.
One of the challenges that historians face is the sheer amount of historical data that exists, much of which has been digitized in recent years. For example, the Venetian state archives contain 1,000 years of history spread across 80 kilometers of shelves, many of which have never been examined by modern historians.
To make sense of this data, researchers have developed computational tools that can help them parse complexity, such as network analysis, natural language processing, and computer vision.
However, these tools have limitations when applied to historical data, which often vary in quality, format, language, and style. For instance, early modern print shops developed unique typefaces for their books, making it difficult for natural language processing models to read the text.
Similarly, computer vision models trained on contemporary images struggle to recognize objects or features in historical images.
This is where deep learning comes in. Deep learning models can overcome some of these limitations by learning from large amounts of data and being able to abstract and generalize.
For example, researchers at the Max Planck Institute for the History of Science used deep learning to analyze a collection of 359 astronomy textbooks published between 1472 and 1650.
They were able to detect, classify, and cluster illustrations and tables based on their content and similarity, revealing patterns of knowledge transmission and evolution across Europe.
Another example is Ithaca, a deep learning model developed by DeepMind that can reconstruct missing portions of ancient inscriptions and attribute dates and locations to them.
The model was trained on a data set of more than 78,000 inscriptions and can produce a range of hypotheses ranked by probability. The model was able to align with the most recent dating breakthroughs for some inscriptions of decrees from classical Athens, showing how machine learning can contribute to debates around one of the most significant moments in Greek history.
Machine learning can also be used to create new forms of historical data or simulations. For instance, the Venice Time Machine project aims to digitize the Venetian state archives and use deep learning to reconstruct historical social networks.
The project hopes to create a digital simulation of medieval Venice down to the neighborhood level. Another example is generative AI, which can create texts or images that look convincingly like historical records or events.
However, machine learning also poses risks and challenges for historical research. One of them is the possibility of creating false or misleading history.
For example, deepfakes are videos or images that use deep learning to manipulate or replace faces or voices. These can be used to create fake speeches or events that never happened or alter existing ones. Another example is generative AI, which can create texts that sound plausible but are not based on evidence or context.
Another challenge is the lack of transparency and interpretability of some machine learning models. These models are often referred to as black boxes because it is not clear how they arrive at their outputs or why they make certain decisions.
This can lead to errors, biases, or misunderstandings that may affect historical analysis or conclusions.
Therefore, historians need to be aware of the strengths and limitations of machine learning and use it with critical detachment.
Machine learning is a useful tool, but not a substitute for human judgment or expertise. Historians also need to collaborate with computer scientists and other disciplines to ensure the ethical and responsible use of machine learning for historical research.
Machine learning is transforming the way we study history by opening up new possibilities and perspectives. It can help us uncover hidden patterns, connections, and voices in historical data that would otherwise be inaccessible or overlooked.
It can also help us visualize and simulate historical scenarios that can enhance our understanding and appreciation of the past.
However, machine learning also comes with challenges and risks that require caution and collaboration. As historians of tomorrow, we need to embrace machine learning as an ally, but not a master.
