Words and Numbers

I’ve gone back and forth about the acceptability of the use of large language models for any kind of work. My hunch is that to communicate with my colleagues, with my students' parents, with my friends, it is not acceptable to use large language models. I think when they were first released, I unreflectively ported that assumption onto many other uses they could have. However, for various reasons, I’ve been convinced that there are acceptable uses for these tools. Deep down, I think it has to do with the distinction between words and numbers.

Human beings use words; words reside at the heart of our existence as our creaturely inheritance. Man, made with words, uses words as a part of the imago Dei. A Greek counterpart to this Judeo-Christian tradition is Aristotle’s logikos, which is usually translated “rational” in the definition of man as a “rational animal”.¹ We might say, though, if we were being stilted and literal with the translation of λογ- roots, that human beings are “word animals”. By this we would mean that they are the kinds of creatures who use words.

In fairness, we also use numbers for lots of things.

Hence the liberal arts, seven of them, divide into the trivium and the quadrivium: three verbal arts, four mathematical arts. One thing you’ll notice, though, if you try to understand the relationship between the trivium and quadrivium is that the trivium can explain the quadrivium, but the reverse is not true. You can explain numbers with words, you cannot explain words with numbers.²

My own heuristic for large language model use has thus become: I never use it for the composition of words, but I do use it for tasks reducable to numbers. The reason for this is that deep down, the words “composed” by large language models are in fact just numbers in words' clothing. As D. Graham Burnett wrote in the New Yorker in 2025:

“That guess is the result of elaborate training, conducted on what amounts to the entirety of accessible human achievement. We’ve let these systems riffle through just about everything we’ve ever said or done, and they ‘get the hang’ of us. They’ve learned our moves, and now they can make them. The results are stupefying, but it’s not magic. It’s math."³

If students habitually utilize these mathematical systems to mediate their engagement with the humanities, they cannot be said to have either read or written. We can no longer require students do the reading or the writing as such. So what’s left? Only this: give them work they want to do. And help them want to do it. What, again, is education? It is at least a non-coercive rearranging of desire, or as Weil puts it, promoting the acquisition of knowledge in love.