I spent some time this summer doing an extended study on translation using AI. My basic question was whether it could help me document translation episodes that have functioned as knowledge generation. I know the phenomenon exists, as I have examples from my own work. I shared these with the AI and asked it to help me find more.
It was more complicated than that, of course, as I wanted certain things and didn’t want others, so I had to provide guidelines. I also wanted time to continue thinking about the project as it proceeded and maybe make adjustments. The solution, which I just submitted as a methodology article to a digital humanities journal, is what I decided to call “punctuated semi-autonomy.”
The method is the main thing in the article, and I don’t want to write too much about it here, as it’s now under review. As I suggested in an earlier post, delay can be productive, and that is at the heart of the paper, which is called “Research at Human Speed.” It was not especially fun to write, as I’m not really too much into methodology. I like to wonder, ask questions, then go explore as I look for answers: that’s pretty much my method.
But it did, also, give me more than 100 episodes from all over the world at different historical periods. This was over a period of about two weeks. That’s the material for exploration. That will be the fun part.
There are also a few big questions that keep coming around, and they made themselves visible in the course of the research as well. One is simply how we work with our tools, our new gadgets. AI is that, but the existential threat voiced of late raises it to a different level, where Hannah Arendt’s ruminations on the new gadget of her own day (the atomic bomb and its progeny) becomes newly relevant. So Arendt’s in there as a frame. There’s also Richard Sennett’s response to Arendt, especially his invocation of Pandora at the beginning of The Craftsman (see my “When Fluency Outpaces Understanding“), which imagines a different interactive framework for humans and their tools.
Then there is the question of thought itself. Some AI proponents seem to conceive of thought as, at bottom, the product of lots of calculations happening very quickly—a variation on the computational theory of mind and, it seems to me, a reductionist corollary of the condition Max Weber described in his 1917 lecture “Science as a Vocation”: the belief that “one can, in principle, master all things by calculation.”
Weber was of course diagnosing rather than celebrating this condition. And there are plenty of older humanistic responses to it. Dostoevsky’s Underground Man, confronted with an imagined person believing in the inevitably positive results of a world governed by calculation, responds, “Oh, the babe! Oh, the simpleton!” I used to teach Sir Gawain and the Green Knight, whose hero fails, in my reading, at precisely the moment he begins to calculate. Other objections come through ideas of positionality, embodiment, tacit knowledge, and practical judgment. What interests me here is less adjudicating among the many accounts of thought than the irony of using a tool that some imagine might replace thinking as a supplement to my own.
My friend Esther suggests I’ve tried to engage with AI on its own terms, which I guess is a possible way of looking at it. I was sort of thinking the opposite: getting it to engage with me on my own terms. And my friend David reminds me that irony is a mode of expression that allows one to entertain a position without totally committing to it. To which I say, well, yes: that’s an accurate description of my use of AI in humanities research at the present time.
###
Russell Scott Valentino 