“…theoretical research in its academic form is the privileged place for these functions to be confused.” (Jacques Derrida)
I’ve learned a great deal this summer. I’m not sure I’ve resolved any of the core issues around artificial intelligence and academic writing, but I have reached a much better understanding, both of how these technologies work and of how our scholarly discourse approaches them. Intellectually, it has calmed my nerves a bit; I’m no longer too worried that robots will suddenly gain moral standing and legal rights in any revolutionary way. Professionally, however, I’m perhaps more worried than I once was; I have to admit that artificial intelligence is more advanced than I had thought, perhaps even than I had thought possible. I’m still confident that it will never be conscious (or even sentient), but it is increasingly able to seem so. In twenty years, I may very well be out of a job. Fortunately, I’ll also be retiring. I’m too old to learn to code!
Anyway, I wanted to make a few quick closing remarks before leaving this topic for a while and getting back to the ordinary business of writing academically. I hope you’ll grant, as I said at the start of the summer, that language models are nominally in my wheelhouse, but I understand if, as a regularly reader of this blog, you think I’ve been off on a bit of a tear. I promise that the next few months will be devoted to the art of learning, the craft of research, and the altogether human pleasures of writing. In any case, here are three quick paragaphs about robots and writing on the way out, each organized around a sometimes subtle distinction that I’ll no doubt be thinking about far into the future.
Robots/Machines. All robots are machines, and not all robots are humanoid. What is the important difference? I think it’s that a robot seems to “serve” us; it exists in what looks like servitude. It’s a machine that interacts with us on a scale that feels “social” and this naturally evokes our sympathy. A machine that does something for us, i.e., something that we would otherwise have to do ourselves, gives us a different feeling, when we watch it work, than one that we do something with, i.e., one that we use to accomplish some goal. We drive, we vaccuum. Once the car drives itself or the vaccuum moves by itself we begin to identify with its predicament. Part of this has to do with the fact that it is doing its work within the same physical constraints as we do ours and with roughly the same urgency. It’s got the room to clean; there’s a speed limit to observe. That seems to be how we distinguish “robots” from mere machines. They occupy space and time on a human scale.
Rights/Rules. The key here is a sense of freedom. Rights give us what Daniel Dennett called “elbow room”, a space in which free will can operate meaningfully. Rules can, of course, be broken, but that word itself suggests that you either follow them or you don’t work. (I’ve always liked the irony of “working to rule” as a form a labor unrest.) We can have rights and enjoy the freedoms that they suggest without ever invoking them. Thinking about the reaction of the state to people like Julian Assange and Edward Snowden, a dark thought once occurred to me: if you want your privacy you’ll have to keep it like a secret. Rights, like privacy, exist in so far as they are respected, while rules must be enforced, like a secret must be guarded. We could imagine rules that make us do only what we’d want to do anyway; but we’d rather simply have the right to do those things. It just feels right.
Writers/Authors. This one has perhaps been talked, if you’ll pardon a little joke, to death. What is an author? What is writing? What is the future of the book? These questions have been addressed from every angle by Barthes, Foucault, Derrida, Inc. & Co. Etc. “An author performs a function,” said Barthes, making a useful distinction, “a writer, an activity.” For the author language is constitutive, for the writer, it merely supports a practice. For the author, language is an end in itself; for the writer, it is a means. I may not be getting that exactly right, but it’s this sort of distinction we’re talking about. For me, it’s all about authority. The author has rights, moral rights, not just legal ones. The writer gives us information but the author takes responsibility. While I’m loth to admit it, “writer” almost becomes a pejorative. Indeed, the more I think about it, this distinction is probably moot in regards to robot rights. Robots, I would say, can’t be authors; they can’t even write.
It has been suggested a number of times during these discussions on Twitter that I’m a poor scholar. I like to think I have cobbled together a workable philosophy over the past thirty years (since I was an undergraduate in that strange trade), but it is true that I don’t approach machine learning, artificial intelligence, and the rights of robots with the same, let us say, “discipline” as people like David Gunkel and Josh Gellers. (Perhaps this paragraph should be rehearsing the distinction erudite/dilettante.) I hope, however, that I have acquitted myself as a plausible “educated layperson”; i.e., the sort of citizen who might one day have to make an “informed decision” about the rights of our robots or the governance of our machines. Perhaps I have managed only to be, as I have described myself in another context, a legitimate peripheral irritation. So be it. Whatever the future may hold, for now AI is certainly a subject on which we may hone our natural intelligence.
_________
This is the last post in a series of fifteen that started with “Robot Writes” back in June, and proceeded through “A Hundred Thousand Billion Bots,” about Raymond Queneau’s famous book; “The Artifice of Babel,” about Borges’s famous library; “Sentience on Stilts,” about Blake Lemoine’s infamous claim for LaMDA; “The Anxiety of Artifice,” about the existential dread of machines; “An Infamous Device,” about the Tower of Babel; “Automatic Sensemaking” and “Are Language Models Deprived of Electric Sleep,” two experiments with GPT-3; “The Automatic C,” a reflection of the ability of machines to pass exams; “Handwriting,” an attempt to recover my wits; “Do Transformers Desire Electric Rights,” an attempt to answer a challenge from Steven Marlow; “Subject-of-a-Text,” a reflection on animal rights; “I Am the Text. The Text is Me. (Or, There Is Nothing Outside the River),” a close analysis of the “personhood” of the Whanganui River; and, “The Virginia Incident,” which is almost a deconstruction of the “rights” of delivery robots in that commonwealth.