Is AI going to rewrite history?
In 2023 we’ve already seen many predictions for how artificial intelligence will change the future. Breakthroughs in AI technology could potentially revolutionize entire industries by automating a wide range of tasks and creative processes in ways we may never have imagined.
But it may not just be the future that AI will change. AI tools have also demonstrated value for historians.
A deep neural network named Ithaca is helping historians fill in the gaps in historical knowledge, according to Oxford University. Over the passage of time, many ancient inscriptions and documents have been damaged and left barely legible, while others have been moved so often that historians struggle to determine their provenance. AI can aid in completing damaged inscriptions that have missing text, as well as determining likely dates and locations of historical documents.
Case in point: a series of important Athenian decrees was thought to have been written before 446/445 BCE — until new evidence led historians to believe 420–430 BCE was a more likely time period. When the historians turned to the AI Ithaca for a likely date, the answer it gave was 421 BCE. This showed that the software was in perfect alignment with the work of its human counterparts.
Much of this work has been in ancient Greece, but historians can also use Ithaca datasets for studying ancient writing systems that include Akkadian, Demotic, Hebrew, and Mayan languages. The team is working to develop versions of Ithaca trained on a wider range of ancient languages.
Google has partnered with DeepMind to provide a free interactive version of Ithaca for historians and researchers.
According to Active History, the future of historical research will see historians working together with chatbots, using them as essential tools in their everyday use of data. AI has the potential to assist with important processes that include finding, synthesizing, and communicating information. But the bots are less reliable when it comes to accuracy and discerning fact from fiction, so these skills are handled by humans.
In a recent project, Austrian Professor Johannes Preiser-Kapeller researched the decisions of the 14th-century Byzantine church. Network analysis software was used to create a database of individuals and then show their connections. This work revealed that the bishops that spoke the most in meetings were not the most influential. When the same technique was applied to the Byzantine elite, more was discovered concerning the hidden contributions of women.
The Venice Time Machine is a similar project that takes on Venetian State archives of 1,000 years of history on 80 kilometers of shelves. This huge digitization project aimed to make it possible to examine all the families living in each building at different times. But while this kind of large-scale project has helped make vast amounts of information more accessible, it has its limitations. Preiser-Kapeller suggests the AI is unable to recognize which connections are meaningful, which means humans will still need to analyze the data.
AI tools with optical character recognition (OCR) that enable them to transcribe and analyze databases of historical documents will lead to huge improvements in speed and efficiency, but it also calls for a higher level of collaboration and sharing. This type of work is the main objective of Canada Declassified, which is a collaborative effort open to historians, researchers, and other people digitizing documents to contribute declassified documents from the Cold War and other periods in recent history. The databases will then be made available to the public for free.
It looks like AI-based tools will do incredible things in historical research, and it won’t be too long before the technology will bring to light some amazing discoveries. But human historians still have an important place when it comes to noticing trends, making decisions, and interpreting findings.
