Navigating history with AI: The promise and peril of digital time travel

Will AI improve our understanding of human history, or will it be a tool that will end our history? Right now, there is evidence to support both perspectives.

AI unlocks the past

This week, a major prize was awarded to three computer science students who used AI, specifically machine learning (ML) and neural networks, to successfully read and translate an ancient Roman scroll recovered from Herculaneum, near Pompeii.

The document is part of a collection known as the Herculaneum scrolls, which were recovered from what was thought to have been a large library in a villa belonging to Julius Caesar’s father-in-law. Between 1752 and 1754, some 1,800 fragments of ancient papyrus scrolls representing perhaps 1,000 original volumes were excavated from the villa. The scrolls were severely damaged during the eruption of Mt. Vesuvius in A.D. 79 and have been described as looking like leftover logs in a campfire.

As noted by the Herculaneum Society, these blackened, carbonized, incredibly fragile scrolls constitute the largest surviving library from classical antiquity, potentially holding lost philosophical dialogues and insights that could reshape our understanding of the ancient world. Attempts to read these documents have been unsuccessful until now, as the scrolls would disintegrate if handled.

Reading the papyri was achieved by three contestants in the Vesuvius Challenge, a competition launched in March 2023 in which people around the world raced to read the ancient Herculaneum papyri. The winners were able to unspool it virtually and read more than 2,000 Greek letters. As reported by The Economist, the text is thought to be a previously unknown work on pleasure by Philodemus, an Epicurean philosopher who lived in Herculaneum. Now, it is likely that the other scrolls can be similarly read.

Robert Fowler, a renowned classicist and the chair of the Herculaneum Society, told Bloomberg: “Some of these texts could completely rewrite the history of key periods of the ancient world. This is the society from which the modern western world is descended.”

This discovery is one example of how AI is unlocking the past, adding depth of knowledge and deepening our understanding of the world. In addition to these scrolls, deep learning techniques are being applied in archeology.

These applications are helping to classify pottery fragments and locating shipwrecks in sonar images, thus “opening new windows on to the past,” according to The New York Times. Another project using AI and robotics aims to reconstruct artworks in Pompei. The first targets of the “RePAIR” project — an acronym for Reconstructing the Past: Artificial Intelligence and Robotics meet Cultural Heritage — are a pair of 2,000-year-old frescoes. 

AI: A threat to history?

These use cases show that the potential uses for AI in historical research are vast and varied. Yet, there are others who view AI as potentially leading to a fundamental change in human history and culture, possibly not for the better. This proves the dual-edged nature of AI, indeed any advanced technology, to have both positive and negative consequences.

As reported by Fortune, “Sapiens” author and historian Yuval Noah Harari told an audience last fall “It’s very likely that in the next few years, [AI] will eat up all of human culture, [everything we’ve achieved] since the Stone Age, and start spewing out a new culture coming from an alien intelligence.” Harari warned: “Potentially we are talking about the end of human history — the end of the period dominated by human beings.”

This suggests that AI models could gain an ability to reconstruct history according to their own analyses, potentially overriding existing consensus when interpreting the lessons of the past. If true, AI could become a dominant force in shaping the world, and humans will have to adapt to its influence, altering human views of the past and changing cultural identity.

A need for digital integrity

One idea to prevent the pollution of the historical record is to better preserve what currently exists to serve as a reference point. As the examples of the Herculaneum scrolls show, the storage medium is critical to preservation success. It is not enough to store content digitally, as this too can degrade.

One possibility comes from Microsoft in the form of Project Silica. Using glass as the medium and laser options for writing and reading the data, the result is a storage medium that can potentially last thousands of years without degradation. This could be an excellent archival system to capture a pre-AI historical record.

Beyond the potential threat from an alien intelligence is one that springs from human minds. A recent New York Times op-ed warns that: “History can be a powerful tool for manipulation and malfeasance. The same generative AI that can fake current events can also fake past ones.”

Which is another way of saying that it is increasingly difficult to distinguish what is real from what is fake, and this could apply equally to the historical record. The authors note that “false documents are a key part of many efforts to rewrite the historical record.” 

For example, viewers on YouTube can watch Richard Nixon make a speech that had been written in case the 1969 moon landing ended in disaster. Fortunately, it never needed to be delivered. However, researchers at MIT created this deepfake to show how AI could manipulate our shared sense of history. 

MIT’s Francesca Panetta said of the deepfake: “This alternative history shows how new technologies can obfuscate the truth around us, encouraging our audience to think carefully about the media they encounter daily.”

The potential for deepfakes to destroy trust in our present reality, as well as fundamentally distort the historical record, only amplifies the need for strict standards and regulations on AI-enabled content creation and dissemination.

If there is an opportunity for us to shape the path of AI and its impact on history, we must do it soon, because the current conditions will set the long-term trend. As we are still at the start of this new era, it is crucial that we navigate these advancements responsibly, ensuring that AI serves as a tool for enlightenment, not confusion or deception.

Instead of AI marking an endpoint for human history, we still can make it a springboard for enlightenment.

Gary Grossman is EVP of technology practice at Edelman and global lead of the Edelman AI Center of Excellence.