There is no question that artificial intelligence is entering - some might say invading - more and more aspects of daily life and "spreading like a plague," in the words of Laurent Dubreuil, Distinguished Professor of Arts and Sciences (College of Arts and Sciences).
For the humanities, that may not be such a terrible thing.
In his new book, "Humanities in the Time of AI," published April 1 by University of Minnesota Press, Dubreuil argues that the arrival of AI may actually present an opportunity to "re-create scholarship."
Dubreuil spoke with the Chronicle about the book.
Question: As a professor of French, francophone and comparative literature who is also interested in modernity and the impact of AI, you have an interesting vantage, straddling the past and the future. Where are you most at home, looking back or looking ahead?
Answer: I am looking even further back, since my core training was in ancient Greek and I continue to have a very strong relation to classical literature and philosophy. In my research, I usually try to adopt a transhistorical view, often making the case for the virtue of anachronism. Technologically speaking, generative AI is a relative novelty, but it is embedded in very old theoretical debates about meaning, thought, language or creation. In this respect, artificial intelligence is as much a thing of the past as it is indicative of the future.
Q: Is there any place in the humanities where it makes sense to use AI? How might it effectively be used?
A: Like in other areas of knowledge, there are, within the humanities, some important technical elements. Take paleography, for instance, which consists of reading older systems of handwriting; or descriptive prosody, where one scans lines of regulated poetry. In such areas, and many others, self-learning automated systems and/or algorithms relying on expert knowledge could prove very useful adjuvants or fine replacements. However, one should never forget that, ultimately, the technical is theoretical: The humanities are the disciplines whose "raison d'être" is to make sense (of life, of the world, of ourselves, of sense itself). We may outsource our techniques and therefore free ourselves from tedious labor, but only if we understand that such means never were an end.
The other big use of AI in the humanities, this time strictly centered on our current generative systems, is to provide a paradoxical quality test. By design, our large language models (LLMs) recycle, recombine and regurgitate what is the most probable (and trite). This is exactly what we should call "generation," which stands in sharp contrast with actual "creation." To say it simply, if our scholarship looks like a response engineered by GPT, then, probably, we are below a certain quality threshold.
Q: At the same time, what does the incorporation of AI mean for the humanities? How can the humanities respond?
A: Doing a passable translation, providing an "objective" summary, applying a preconceived theory to a text or an image, mastering impersonal style, labelling "good" and "bad" behavior - none of this ever mattered much. Some of these tasks might have a virtue when we train ourselves and they could remain as a part of the learning experience, but they will simply need to go from humanistic research, all the more so when LLMs effortlessly accomplish a roughly similar performance. The humanistic ambition is, or should be, much loftier: It is time to reaffirm it.
How we do this will vary. I am launching with Cornell University Press a new book series titled "Mechanema" that will open up an editorial space for such reflections. With my Humanities Lab, we have been undertaking transdisciplinary and artistic projects to illustrate and interpret differences between a plurality of regimes of thought and more one-dimensional modes of computation. New, transcultural assemblages to refurbish our theoretical vocabulary should also be in order. There are many other paths, but acting as if AI did not exist or even was external to the humanities is unsound.
Q: Do you view the humanities as something fixed, or is it a living thing, always evolving, in flux? Maybe another way of asking this question is: Might the emergence of AI help us reconceptualize the humanities?
A: We cannot doubt that many decisionmakers will use the technical advances of AI as a pretext to try to downsize free inquiry, fundamental research, transformative expression and intellectual speculation. At the same time, thanks to the distorted image of ourselves we find in the AI-mirror, it becomes even clearer than before that our task is not anchored in consensus and mediocrity but dedicated to signification, creation and exception across time, space, language, culture. This dedication is not novel; the forms of its realization are what evolves.
Q: Throughout the book you include prompt questions and AI responses. What did you learn from that exercise? Did any of the AI responses surprise you?
A: A couple of years ago, with my colleague Morten Christiansen and other collaborators, I co-designed a scientific experiment where we'd compare original poems with completions by GPTInstruct and by undergraduate students. When I wrote the book, I had already completed that inquiry, which taught me much about the efficiency of generative AI, its difference of implementation vis-à-vis the multiplicity of the human experience, its tendency to recite and the limits of statistical prediction. Then, the prompts and responses in "Humanities in the Time of AI" were not designed to systematically explore a problem but were rather intended as an ironic appearance of AI in a text that might sometimes look critical of it.
This being said, some AI responses for this essay convinced me that LLMs seem much better at some supposedly "human" tasks than what is usually acknowledged, especially in using the language of emotions: There might be no affective ghost in the machine, but the verbal imitation of the rhetoric of touchy feelings is pretty powerful.
