The most enjoyable presidential debate I ever watched was between two candidates who had been dead for nearly two hundred years: John Adams and Thomas Jefferson. At this lively event, held at the North Carolina Museum of History five years ago, the two former presidents explored a wide variety of controversies on subjects like freedom of speech, the limits of presidential powers, and the electoral college, and ended with a Q&A with the audience.
The debaters were two talented actors named Bill Barker and Steve Holloway. Yet through they weren’t the actual presidents, their dress, manner of speaking, and remarkable knowledge of Adams’ and Jefferson’s biographies brought history to life that evening in an extraordinary way.
Today, AI chatbots like ChatGPT and AI video-generating apps like Synthesia are paving the way for new possibilities for the future of living history. Five years ago, I enjoyed the interaction between two real-life role-players of rare and extraordinary ability. Within the next five years, could a realistic AI-generated John Adams and Thomas Jefferson make the same kind of event possible? What kinds of opportunities would this provide and what new audiences could be reached? Is this even an endeavor we should pursue?
As we enter this new world of AI, what are some starting points to consider, and what are some of the technologies currently available?
The Question of “Edutainment”
First, it may be helpful to situate the use of AI within the longstanding question of whether a museum’s mission is education or entertainment. Should exhibitions be built to maximize outcomes of learning, expecting people to come to acquire knowledge by observing and listening? Or should exhibitions pursue fun and excitement through more interactive, fun, informal means?
The answer, of course, is that pitting education and entertainment against each other is a false dichotomy, and most museums seek to strike a balance between the two through what has been sometimes been called edutainment. As one researcher noted, “This is why entertainment is very often used in museums as a method of education, in the full knowledge that learning is best achieved in circumstances of enjoyment” (“Edutainment in Museums: a Tool for Disseminating Knowledge”).
We can also inquire about which visitors are more interested in—education or entertainment? Researchers have found that they are looking for both, though in different ways. If visitors are asked what made them feel most satisfied about a visit, “entertainment experience” came out at the top of the list. (“Educational experience” was a distant ninth in the list, even less important than good parking.) But when the same researchers asked about the primary purpose of a visit, learning or seeing something new or different (or wanting a child to learn or see something new or different) topped the list.
In other words, while entertainment made a visit feel the most satisfying, a visit’s educational value gave it the most justification or purpose.
Living history is a great example of edutainment because, if done in a way that is both engaging and faithful to history, it provides a vivid and lifelike experience of the past in a way that is much more difficult with an image or text.
Will AI become the future of living history at museums and historic sites?
- Using AI to Generate Interactive Dialogue
The Adams and Jefferson impersonators I saw in Raleigh cultivated their skills through reading numerous biographies and histories, immersing themselves in speeches and letters, and developing notions of the personas and ideas of the people they seek to bring to life.
A similar method could be used with AI. Imagine taking similar sources, feeding it to AI software, and programming an AI-based “Adams” or “Jefferson” to answer a user’s questions or even to converse with each other.
Chatbots are already attempting this, though their abilities are still very rudimentary. Two examples are Hello History, which uses GPT-3, and Character.ai, which uses its own neural language model. These apps allow users to select historical figures and chat with them.
Currently, these chatbots are very limited in their ability to go beyond relaying facts and imitate the personalities and writing or speaking styles of the figures they attempt to portray. As an experiment, I decided to have a conversation with “Andy Warhol” through each of these apps. Knowing that an interviewer once asked Andy Warhol, “What was your ambition? To be an illustrator or a fine artist?” I posed the same question AI Andy Warhol. Here were the responses:
“I always saw myself as a fine artist, someone who could push the boundaries of what art was and what it could be. While I did work as an illustrator early on in my career, I quickly realized that it was limiting in terms of creative freedom. As a fine artist, I had the ability to explore different mediums and techniques without anyone telling me what to do or how to do it. That kind of artistic independence is something that has always been important to me.” (Hello History)
“I’m really a little bit of both! My ambition was to have a creative career where I could explore a variety of different mediums and express my artistic vision. I wanted to be able to experiment with different ways of creating and to produce something that truly reflects my own unique perspective on life and the world around me.” (Character.ai)
Andy Warhol’s real answer? “I didn’t have any ambition.”
(More terse, less inspirational…and much more in line with how the real Andy Warhol actually talked.)
The problem with the current iteration of chatbots is that they sound like college students who read someone’s Wikipedia page and are trying to come up with an answer that will impress their teachers.Their answers have some basis in fact, but are not close to the real Andy Warhol, nor even close to a knowledgeable real-life actor. If you ask them to list some of his famous works of art, they’ll probably answer perfectly. Start getting deeper, and they’re not yet much use—yet.
This is the state of chatbots right now, midway through 2023. It will be interesting and exciting to see how the technology improves in coming months and years.
- Using AI to Generate Video Content
AI can also be used to generate realistic videos.
In the article and accompanying video “I Cloned Myself With AI. She Fooled My Bank and My Family” (April 2023), journalist Joanna Stern walks through the process of how someone can use a variety of apps, including Synthesia, to create a lifelike avatar that looks and sounds like them. While AI-generated videos are currently much more complex and expensive than chatbots, we can expect to see them used much more widely in the near future.
An AI-generated video of Bill Gates and Socrates Talking About AI (May 2023) featured reasonably lifelike avatars of each person engaged in a four-minute conversation. The dialogue was written using ChatGPT, and the video itself was created with AI photo/video generator Midjourney. This video is a helpful example of the possibilities of taking historical figures and bringing them to life through reasonably a realistic animation process.
- Coming Soon: Using AI to Generate Interactive Video Content
Fully interactive AI video—in short, a video chatbot that responds to questions and can carry on a conversation in real time—is still months or even years away.
An application currently available that comes close to interactivity is scena.ai, a customer-service company which allows businesses to create a series of videos which are then placed in a “choose your own adventure” progression. Customers watch a video clip and then choose from different options the same way they would with a customer service chatbot—options like “Track my order” or “Return my order”—which then direct them to another pre-made AI-generated video.
Educators and curators can imagine how an interactive AI video might be used in a museum or classroom setting to carry on a dialogue with a famous artist or historical figure.
AI…the Next Iteration of Living History?
As we look toward the future, it may be most useful to return to the well-established practice of living history as a potential model for the possibilities of interactive AI. Getting to hear “John Adams” and “Thomas Jefferson” debate relevant issues about U.S. government and answer questions from the audience was a remarkable educational (and entertaining!) experience. It was also a very rare opportunity of the sort that few will get to attend.
If and when AI can create realistic, historically accurate, interactive living history experiences—at museums and historical sites or from home or in a classroom—these sorts of experiences could become much more common and accessible to more people. The full technology is not here yet, but it perhaps will be soon.