Artificial intelligence is transforming historical research in profound ways, unlocking secrets contained in centuries-old documents and enabling technological and historical analysis at a level not previously thought possible.
The Great Isaiah Scroll
The Dead Sea Scrolls are considered the oldest manuscripts of the Hebrew Bible, dating from the 4th century BCE. The first discovery of the scrolls was in 1947 when a young Bedouin shepherd stumbled on them in the Qumran caves near the Dead Sea, now in the West Bank. Additional scrolls have continued to emerge, including some discovered just a few months ago. The scrolls are unsigned and undated, leading to understandable speculation and research as to their origins.
One of the longest (more than seven meters) and most complete scroll discovered in 1947 was the Great Isaiah Scroll. Visual analysis of the handwriting had almost universally concluded that it was the work of a single scribe, with any minute differences attributable to a tired hand or a change in pen. Now Dutch researchers from the University of Groninger, using computer techniques and Artificial Intelligence (AI), have concluded that it is probable that the Great Isaiah Scroll was penned by two different scribes. The changeover occurs in the middle of the scroll with a slight text break and a change in substantive content.
The study analyzed a single letter – aleph, the first letter of the Hebrew alphabet, which appears in the text more than 5,000 times. The process involved an artificial neural network trained using deep learning to extract details from the images.
The implications of this finding are significant:
- Because the handwriting is nearly identical throughout, it suggests that the two scribes received the same training or otherwise collaborated.
- Analysis of additional scrolls may elicit evidence of specific scribes and whether they belonged to different scribal communities with stylistic differences.
As explained in the abstract to the study:
Demonstrating that two main scribes, each showing different writing patterns, were responsible for the Great Isaiah Scroll, this study sheds new light on the Bible’s ancient scribal culture by providing new, tangible evidence that ancient biblical texts were not copied by a single scribe only but that multiple scribes, while carefully mirroring another scribe’s writing style, could closely collaborate on one particular manuscript.
Deciphering Lost Languages
AI has also been employed to decipher lost languages. Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have developed a system that can decipher a lost language without advanced knowledge of its relationship to known languages, an astounding advance since previous research has relied on mapping an unknown alphabet onto a known one. The system relies on a decipherment algorithm created from principles of historical linguistics to connect texts to related words in a known language.
This is not a process for the faint of heart. Because of the limited numbers of samples, there are no large datasets to generate an algorithm and it’s not possible to access lost languages that left behind no written documents. Moreover, from the practical point of view, there is not a huge demand to decipher these languages so research funding is limited.
Exploring Fragile Documents
AI is also allowing us to access ancient writings where the languages are decipherable, but the medium is so damaged that they cannot be read by humans. Machines are being taught, for example, to read stone tablets whose markings have been eroded over time.
In addition, there is “virtual unwrapping.” Where documents are so fragile that flattening or unfolding them would be destructive, they can now be digitally opened and revealed in their entirety. This technique has been developed over the years by Brent Seales, a professor of computer science at the University of Kentucky, who had the idea of adapting medical CT scanning to look inside documents. He used the process in 2009 to peer into scrolls from the library at Herculaneum, destroyed, along with neighboring Pompeii, by the eruption of Mount Vesuvius in 79 AD. The hope was that once the entirety of the scroll was visible, characters could be distinguished from background. For a variety of reasons relating to the small amount of trace metals and the inability to distinguish carbon in the papyrus from carbon in the ink, that did not turn out to be the case. However, in 2014 he was able to apply the process to the En-Gedi Scroll, a third or fourth century parchment scroll that turned out to contain the first two chapters of the Book of Leviticus, the earliest known Hebrew copy other than the Dead Sea Scrolls. Seales is using artificial intelligence to improve his software in the hopes that eventually it will be able to decipher the Herculaneum scrolls.
Bringing Historical Characters Back to Life
Another development, bound to be more controversial, is the use of AI to bring historical characters “back to life.” The Museum of Art and Photography in Bangalore, India (known as the Silicon Valley of India) has used facial recognition and other deep learning techniques to create a "conversational digital persona" of the painter M.F. Husain, who died in 2011. The digital twin can answer questions about Husain’s life and works. We can be confident that the responses of the digital Husain are based on competent research and used for educational purposes, but the AI that enabled them can be misused to distort history. In fact, we have previously examined the legal issues surrounding deepfakes that spread disinformation.
Like so many other AI applications, these technologies can be a double-edged sword, but the benefits for historians are potentially immeasurable.