The Future is Generative
Or How a Historian Learned to Stop Worrying and Love ChatGPT
Something’s in the air at universities these days. Generative AI has been around for a while, but it didn’t become mainstream until ChatGPT, Dall-E 2, Midjourney, and the new Bing burst into our collective consciousness all at once. Many colleagues tell me they first became aware of the technology after hearing about its potential for plagiarism and academic dishonesty. But I think we are witnessing something far more radical and consequential take shape.
An Unlikely Convert
I think of myself as an unlikely convert to AI. Although I use a lot of technology in my work as a historian, I would not call myself a digital humanist. In my experience, that term is usually reserved for scholars who build text-clouds to visualize word frequencies, print 3d models of artifacts, or use augmented reality—amongst many other things. I know this can be exciting, but it has never felt relevant to what I do.
I see myself as a traditional historian: I assign research essays and book reviews in my classes and I prefer to write books and articles based on detailed archival research. Nevertheless, I take thousands of digital images at archives, use OCR software, and even have an automatic microfilm scanner at home. But for me technology exists on the “input” side of the research equation; I have just never been enamored with digital outputs. I like a good detective story which is what drives me as a scholar. Word clouds and data visualizations offer neither the mysteries I crave, nor the tools to solve them.
In this way, I may be typical of my generation: I am comfortable with technology, so much so that I don’t think much about it. I was born in 1981 and computers have always just “been there” as part of the research and writing process. I only remember typewriters as dusty, odd toys in my parent’s basement. I don’t think I have ever written anything of consequence by hand, except exams.
I was educated entirely during the first digital revolution which brought lots of changes to how we do things as historians, but still allowed the traditional to at least co-exist with the digital. For most of us, computers and then internet just allowed us to do the things historians had always done faster and more efficiently. None of this required special knowledge or skills, that was the domain of the “real” digital humanists.
An End to the Beginning of the Digital Revolution
Generative AI is about to end this détente between the traditional and the modern as we all become digital creators. AI is disruptive precisely because it significantly reduces the learning curve and time necessary to generate new content, be that artwork, music, or text. In essence, it allows us to outsource much of the work involved at becoming good at something like painting, composition, or writing.
The counterargument is that no-one is going to want to listen to an AI generated symphony or read an AI novel because those things are inherently valuable only because they are expressions of our humanity. That is entirely true, but it misses the point. Most composers don’t write symphonies nor are most writers best selling authors. The market for those things is unlikely to change very much. What will change is the market for the mediocre.
Yet if it changes how we create, it will also change how we frame problems, research and analyze the past. AI is very good at finding patterns. That is, in fact, how things like ChatGPT write, by predicting what is likely to come next from a corpus of many billions of pages of ingested text. Finding patterns is, in effect, what historians do too. That will not change, but AI will soon be involved at every stage in that process from locating archival sources to sifting through the records and writing up the results. Although many of us may not be comfortable with the idea, it will be hard to put it back in the box.
Uncertain Futures
So what will this future hold for the humanities and history in particular? That is the million dollar question and anyone who seems to have the answer now is probably wrong. What is clear, though, is that as AI tools are integrated into Google, Social Media, and Microsoft Office over the next few months, we are going to have to feel our way forward. We will obviously have to wrestle with questions of authorship, authenticity, and attribution—that we already know. Above all else, though, we need to contend with the fact that very soon, our graduates will enter a work-world where they are expected to make efficient use of both AI and their own human talents. Figuring out how to get them there is our problem.
And this is, I think, where the future of the humanities and the teaching of history in particular lies. Efficient use of AI requires efficient operators with a good grasp of language, a good general knowledge about the world, and the critical thinking skills necessary to get the best results from the machine and fact check “hallucinations.” In this sense, the solution probably involves putting a little bit of old wine in new bottles. Here we can take the basic skills we have always practiced and taught and dress them up for a new era with the latest AI jargon. Deans and Vice Presidents will no doubt be ecstatic. But my sense is that we also have to make some very real changes, whether we want to or not.
We need to be prepared to rethink how we teach research, critical thinking, and writing. We need to acknowledge that AI can be a useful tool in this process and pioneer effective ways to maximize its potential. It likely means developing new assignments and new approaches to evaluation. For example, generative AI might open up new possibilities for a truly flipped classroom because advanced chatbots like ChatGPT can be effective tutors.
It also means incorporating AI into our own work as well. This need not be as frightening as it sounds as it will happen organically as archives use it to generate metadata for their collections and we get used to “Binging” something rather than “Googling” it.
Getting Started with Generative History
We all have to start somewhere. I have been experimenting with using AI in my own historical research for a few months now and I have found it useful—or just plain cool—in myriad ways. Over the next few months I will be piloting an AI assignment in a third year history course I’m teaching on the fur trade. I want to try out a few ideas with my students to see what works and what doesn’t. This is a dry run, in effect, for next fall when I teach a much larger course organized around using AI tools to research and write. I have no idea how it will go yet, but that is part of the fun. Check back and I will let you know.