In this column distributed by the Elon University Writers Syndicate, Professor Rosemary Haskell asks whether the language-generation capabilities of generative AI will impact the thought process.
By Rosemary Haskell
Text-generating artificial intelligence devices such as ChatGPT send shivers down many spines, particularly the sensitive vertebrae of writing and literature instructors.
“Write a paper about Hamlet’s Oedipus complex,” I ask ChatGPT and lo, there is an essay… in some shape or form. The optimist in me says, well, a student will need to edit and add to and check sources for that text, and probably extend it. In fact, by the end of the process, he will actually have some understanding of the topic and of the play.
But will the student actually have to read Shakespeare’s drama? Will he or she actually have to struggle to sift the play’s components, to shape a focus and to develop the tiny germs of their own ideas? The answer appears to be “No.” The student, or anyone, won’t have to do any of those things.
How much does the absence of such intellectual activity matter? As often happens, we can ask George Orwell to sharpen our thinking. His 1946 essay “Politics and the English Language” denounces contemporary English because it is stuffed with “convenient” off-the-rack phrases that writers reach for without thought and then “tack together like the sections of a prefabricated henhouse.” No tough brainwork is needed. We may become like the “dummy” speaker who, by “letting ready-made phrases come crowding in” and regurgitating them “has gone some distance towards turning himself into a machine.”
But why did Orwell fear this semi-automatic method of merely producing text instead of composing a work themselves? He saw it as the abdication of thought leading to the concealment of meaning — from others but, more sinisterly, even from writers and speakers themselves. Using other people’s words and phrases — the ones already doing the rounds — eliminates the hard work of composition. Furthermore, Orwell said this thought-lite writing promotes political “orthodoxy” – the acceptance of the status quo.
OpenAI’s ChatGPT accesses a word hoard that’s “already out there.” To echo Orwell, I wonder if this huge warehouse of verbal garments that are off-the-rack rather than bespoke will limit our thought. To use another metaphor, will AI’s script keep us treading water rather than swimming forward to new thoughts, new ideas, which are formed partly through our idiosyncratic and probably less correct ways of expressing them?
Orwell was gearing up in his 1946 essay for the later horrors of his novel “1984,” where totalitarian Big Brother quashes rebellious thought by promoting the official language of Newspeak. That language drastically reduced the store of words and so, the novel argues, reduced the range of thought, especially politically unorthodox thought.
In my darker moments, I cannot help but consider the text-generating functions of Chat GPT as Orwellian agents of conformity by channeling the internet world into my brain. Of course, such linguistic Pavlovianism has been occurring for years: text prediction in my email writes things before I do. I’m becoming more compliant: why not let Bill Gates write my emails for me? His Microsoft brain saves me trouble. I think less.
As a teacher of undergraduate writing and literature, I feel as though I am on the leading edge of this newish text-generating territory. Students, and perhaps all of us, have always found ways to avoid reading assigned texts. Now that late-adolescent “don’t want to do my homework” temptation has been made much easier to succumb to.
Still, I try to put the AI robo-writer in perspective. After all, the internet didn’t destroy reading, writing and thinking but opened up a huge vista of more easily accessible information. To start my research, I no longer have to trek to a library, thumb through drawers full of moldering index cards, heave giant indexes off shelves and laboriously transcribe authors, titles, dates. The great electronic machine-brain can do all of these things for me. “When I was your age, I actually had to get up out of my chair and walk uphill both ways to the library,” was a common refrain. I don’t even bother to say that anymore, so distant are those analogue days.
And yet, if ChatGPT makes not reading “Hamlet” or Toni Morrison’s “Beloved” the default option, if it deprives us and our descendants of the intoxicating and painful pleasures of creating from scratch our own idiosyncratic and imperfect paragraphs, I won’t be able to forgive it.
—
Views expressed in this column are the author’s own and not necessarily those of Elon University.