Europe

How to Make History Come Alive With AI

Harrison May, creator of the Instagram account Legends of Amsterdam, uses AI to generate photorealistic images of historical Amsterdam. By working with Midjourney, he adds a human touch to historical storytelling, making the past more relatable. While early versions of Midjourney were unsuitable for the purpose, the technology is rapidly improving, and May now counts 18,500 followers.
By
man1

Via @legendsofamsterdam on Instagram.

September 07, 2024 04:28 EDT
Print

2024 marks 400 years since the founding of New Amsterdam (now New York) by the Dutch colonists. American-born Harrison May now calls “Old” Amsterdam his home. He is the person behind the popular Instagram account Legends of Amsterdam.

May uses artificial intelligence (AI) to bring historical Amsterdam to life. What drives the creator of these photorealistic historical images, and how does he go about making them?

Making history come to life

May enters the historic building where the editorial office of Dutch historical magazine Ons Amsterdam is located. The first thing he does is ask me how old the small drawing of a boat on the facade of the building is. “Maybe it’s ancient graffiti. I’m going to put it online and ask my followers!” This enthusiasm is exemplary of how May absorbs the city.

“I had been to Europe before moving to Amsterdam,” May starts the conversation. “In Europe, you experience history much more than in the US, where I grew up.” In his homeland, he did not study history, but film. “During my studies in Germany — where I had to get my so-called Abitur diploma with an emphasis on history before applying to a university — I found out that although I found history extremely interesting, I was mostly interested in small details that colored it. The personal stories described in memoirs fascinated me much more than an essay on the economy of the Roman Empire,” May explains with a telling smile.

Unfortunately, he was not admitted to university in Germany, so he returned to his homeland. Between 2013 and 2015, May studied film at Northern Arizona Unversity and moved to Amsterdam after his graduation.

Making use of new technology

In Amsterdam, May worked as a film director and photographer. In 2021, when AI as a technology became accessible to private creators with the launch of programs like Midjourney, May immediately recognized that this might have been the technological leap he was waiting for.

“In the beginning, the AI technology that allowed you to generate images was very rudimentary, because it was untrained,” he says. “The images it generated were terrible; they almost gave me nightmares.” But the technology also allowed one to cinematically time-travel without investing tons of money in elaborate sets and period costumes.

Six months later, Midjourney was so well developed that May attempted to conjure beautiful images from it again: “When I noticed that Midjouney had leapt in its development, I started working with it intensively. The images it generated were now so good that I started sharing them on the Instagram account @legendsofamsterdam in June 2022.”

For anyone running to their computer to get started with Midjourney, there’s more to it than just entering a few requests (“prompts” in AI jargon) using text or a pre-existing image, and, like magic, seeing a photorealistic image from the 17th century pop up on the screen. May happily explains: “I use my experience from film and photography to write prompts to, produce the right first image. If I don’t quite like the result, I will adjust the prompt and go back and forth until I’m happy with it. Then I use Photoshop to edit the image further. With Midjourney, I can now tell stories that I used to tell through film. My biggest challenge now is approaching reality.”

Can AI-generated images be historically accurate?

And with approaching reality comes historical accuracy, May says: “For me, it is essential that the images are historically accurate as far as that can be validated, of course. For example, there is a little tower in the first AI versions of the Palace on Dam Square. That, of course, is incorrect, and by now, the program has learned that the turret does not belong there. Also, some generated images show large ships in the city. However, these did not enter the city, I recently learned.”

But what can AI do for the visual historiography of Amsterdam? “AI can be used as a tool to create just about anything you can imagine digitally. With Legends of Amsterdam, I want to tell stories with few visual references and bring a new perspective of history, we may not have seen before. That way, we can fill in visual gaps from the past to create a world we could only, until now, imagine.”

The reliability of AI is something the medium has struggled with from its inception, as the debacle of Microsoft’s chatbot Tay illustrates. Tay had to be taken down a mere sixteen hours after its launch when it started delivering racist messages.

The question of how AI can paint a reliable picture of history is a relevant one for visual historiography. May links this to “old-fashioned” historiography: “I think it’s important for people to remember that history has always been subject to interpretation. Even before the advent of AI, ‘reliable’ historical information was mostly subjective. Even the visual references from different historical periods are subject to artistic freedom and interpretation. AI makes no ethical or moral decisions about the reliability or historical accuracy of what it creates. So, just as in the good old days before AI, the best way to test whether information is reliable is to spend time researching and working with experts and historians who know what they’re talking about.”

Midjourney updates appear weekly; the program keeps getting better and more precise. Staying up-to-date is vital to May, as is feedback from his now more than ten thousand followers on Instagram. “I used a type of window in the beginning that didn’t match the era. That’s the nice thing about social media; you get told quickly when something is off. So, feedback, especially from experts, is essential for me to get as close to the historical truth as possible.”

Telling stories to human beings

Asked about his primary motivation for continuing to polish the images generated by the program in Midjourney, May harkens back to his passion for storytelling: “By working with AI, I add the human touch to the history of the city of Amsterdam. As a visual storyteller, I see this as an indispensable ingredient to invite people into my stories and make them think about the history we are telling. I can share historical narratives with a direct connection to the present for people to converse about them. Because if we forget history, chances are it will repeat itself. I could tell stories about colonization, slavery and war in a way that engages people today and makes them think. But of course, I would also like to tell beautiful stories because light and dark have an essential place in our history.”

“For the near future, collaborations are important to me,” May explains. “I would like to collaborate with institutions, artists and people with great historical knowledge to tell stories that are as historically accurate as possible. This is part of my vision for the future, so I hereby call for these professionals to approach me. In addition, my interest lies in telling unknown stories, the legends, folklore, the stories where the truth has been subject to erosion over the years.” 

Of course, the most pressing question arises when it comes to the use of artificial intelligence: How great is the danger of AI? What will be the impact of this technology, which is only in its infancy? Harrison May doesn’t know the answer. But he will be happy to discuss it with you.

The views expressed in this article are the author’s own and do not necessarily reflect Fair Observer’s editorial policy.

Comment

Only Fair Observer members can comment. Please login to comment.

Leave a comment

Support Fair Observer

We rely on your support for our independence, diversity and quality.

For more than 10 years, Fair Observer has been free, fair and independent. No billionaire owns us, no advertisers control us. We are a reader-supported nonprofit. Unlike many other publications, we keep our content free for readers regardless of where they live or whether they can afford to pay. We have no paywalls and no ads.

In the post-truth era of fake news, echo chambers and filter bubbles, we publish a plurality of perspectives from around the world. Anyone can publish with us, but everyone goes through a rigorous editorial process. So, you get fact-checked, well-reasoned content instead of noise.

We publish 2,500+ voices from 90+ countries. We also conduct education and training programs on subjects ranging from digital media and journalism to writing and critical thinking. This doesn’t come cheap. Servers, editors, trainers and web developers cost money.
Please consider supporting us on a regular basis as a recurring donor or a sustaining member.

Will you support FO’s journalism?

We rely on your support for our independence, diversity and quality.

Donation Cycle

Donation Amount

The IRS recognizes Fair Observer as a section 501(c)(3) registered public charity (EIN: 46-4070943), enabling you to claim a tax deduction.

Make Sense of the World

Unique Insights from 2,500+ Contributors in 90+ Countries

Support Fair Observer

Support Fair Observer by becoming a sustaining member

Become a Member