Speaking as a non-academic, I think I can say that those of you on the front lines of the Humanities are the true heroes of our era! You have the opportunity to reshape education and make the Humanities the centerpiece of our culture. I admire your approach of experimenting and sharing feedback on how different strategies might work.
I'm sure this increases the teaching burden with zero additional compensation, but if it's any consolation, I have found the Humanities the most important and long-lasting part of my education. My only regret is that I never reached out to thank all those responsible.
Thanks, Paul. Most of us in the humanities, at least in my department, are fighting the good fight. I don't see us returning to a central position in the culture any time soon, but we will go down swinging!
Recent ChatGPT "mishap." My wife and I dozed too much during the first episode of Unforgotten, Season 5. We started episode two and realized we had no idea who was who. So I asked GPT for a summary of the first episode. It confused us more because it was just wrong about who and where the characters were. It was annoying, then funny how wrong it was. Similar to its map attempt in your post.
I do think it's good for research. But mostly to go a level deeper. I needed to know cutting-edge nail polish colors in 2012 (a novel set then) and for that it was great. I could then go to google and click through to ads for the color I chose (for the name––Malice Red)
And I still think that humanities classrooms should have "dumb" devices that just allow a student to type out an essay. I pity the teacher who'd have to decipher my handwriting. Or I pity myself for having to write in block letters so slowly to make myself legible that I wouldn't be able to complete my essay.
Last year I quit smoking what Geoffrey Hinton is smoking, and I don't miss it, so I won't say "I want some of what Geoffrey Hinton is smoking"; but that's what I think when I read his confident pronouncements about the inevitability of constructing systems we cannot describe, or blindly iterating to them, or catalyzing their emergence, or however he thinks about it.
My hot takes on LLMs are getting cooler by the second; everyone is realizing “AGI” isn’t coming. Gary Marcus is a strong writer on this subject, if you want more encouragement! Apple’s most recent (very bearish) research paper on LLMs is good too, will link to both below!
I’m about to co-teach an introductory seminar on classical antiquity, and I think this is brilliant. If my colleague is agreeable, we’ll try out some of your ideas. For what it’s worth, I started my teaching career before desktop computers.
I think it's an almost impossible task to expect your students to engage themselves rather than engage AI, particularly in these days of large class sizes and remote study. There maybe some potential still in the Oxbridge 'personal tutorial' approach where there is close discussion with a tutor in a (very)small group of students. There is little scope for device use for conversation/discussion in such a small and focussed group. A requirement for essays for assessment would need to be in person, hand written or via a simple word processor with no internet access. To be honest, I rather like the ideas that trying to avoid AI produces - in person reading and discussion, small groups, invigilated, in-person internet-free examinations, using a library etc etc. Like stepping back into a more engaged educational context. One that I personally missed out on and still long for. There is hope if this can take hold, but it's expensive and intense. It also requires those teaching to be deeply and widely engaged in their subject - and I'm not sure that's always the case now.
Fair points, but most of my courses with English majors and graduate students are small and discussion-driven, and most of the student buy in enthusiastically. And even in the larger general-curriculum courses with non-majors, there is always a cohort of students who come in good faith and are interested in the work. I teach to these students and accept the fact that I can't reach all of them, though I try.
I’m sorry, but this is simply delusional: “We can emphasize these skills by designing assignments that require students to connect course material to their own lives, to engage in nuanced debate, or to produce something genuinely original. The emphasis should be on the process of thinking and creating, not just the final product.”
There’s no requiring anything. Every step, even the most personalized, can be fed back to the LLM.
Absolutely. As I said to a room full of English majors yesterday, our only option is to trust each other. Perhaps I'm naive, but I am convinced that most of them are participating in good faith. I accept that this is likely not the case in general-curriculum courses.
It's also likely not the case in English courses taught by faculty the students don't like much (I assume they all adore you).
Perhaps a bigger concern is how much AI threatens pedagogical property. I've been thinking about this as a book coach, how I offer some free resources as lead magnets that could be turned into AI prompts. Not as good as working with me, but probably more serviceable than I'd like to think.
So all that syllabus sharing and assignment sharing could encourage more brazen plagiarism of colleagues. Here's this assignment John did, tweak it for my class in the American Novel, etc. And the more savvy students, probably including some majors, might feed old drafts with your comments on them into the system either for tips on improving new work or as a way to "train" the LLM in your standards. Or maybe those resources could be leveraged to produce a "close reading" of a text that aligns with your rubric.
You haven't mentioned Hua Hsu's essay in the New Yorker, but I think he shows pretty plainly that even above average students are doing this without a lot of guilt.
AI shows evidence of having been designed to appeal to the cognitive biases, such as confirmation bias. The socratic approach has the potential of unlocking a key to using AI, which is that the questions have more importance than the answers. Answers turn on facts, which are always changing. The questions are more durable.
Just for fun, I consulted my friendly neighborhood bot for suggestions on topics for a English Lit graduate seminar on literary criticism using Ulysses as the vehicle. Here's what I elicited. I am no authority on either literary criticism nor post-modernism, but it seems a passable simulacrum.
1. Authority, Narrative, and Voice
* To what extent does Molly’s unpunctuated monologue destabilize traditional narrative authority?
* Does the lack of punctuation democratize meaning-making, or does it obscure voice in ways that reinforce Joyce’s (male) authorship over Molly’s “voice”?
2. Embodiment and Language
* How does Joyce’s rendering of Molly’s embodied sexuality challenge or reinforce the tendency of modernist texts to intellectualize experience?
* Does the linguistic stream of bodily sensations subvert male rationalist discourse, or does it reinscribe it by turning embodiment into literary experiment?
3. The Politics of Representation
* Is Molly Bloom’s soliloquy a feminist gesture, granting a woman subjectivity at the climax of the novel, or does it remain a male-constructed fantasy of female consciousness?
* How might we reconcile the radical form with the problematic provenance of its creation (Joyce ventriloquizing a woman’s inner life)?
4. Intertextuality and Deconstruction
* How does “Penelope” undermine or echo earlier chapters of *Ulysses*? For example, in what ways does it resist Bloom’s own attempts to define or narrate Molly throughout the novel?
* What happens to meaning if we read Molly’s “Yes” less as affirmation and more as différance—a sign that endlessly defers closure?
5. Subjectivity and Postmodern Multiplicity
* Does Molly’s stream of consciousness exemplify postmodern subjectivity—fragmented, decentred, contradictory—or is it still caught in modernist faith in interiority?
* How might her soliloquy resist or collapse binary oppositions (male/female, speech/writing, desire/domesticity, fidelity/infidelity)?
6. Ethics and Readerly Complicity
* How are readers implicated in the voyeurism of accessing Molly’s intimate thoughts?
* Does the novel invite us to critique that voyeurism, or simply indulge in it under the guise of high art?
Interesting. Some of those are pretty interesting questions, though some are predictable. My problem with students using AI for this kind of brainstorming is that human brainstorming (for me, at least) is a vital part of the process of discovery and of the writing process in general.
I was just discussing Geoffrey Hinton's claims with my mother. My mother is almost entirely analog in her approach to technology - no matter how many times we have tried to teach, she cannot navigate 'smart' technology - so her source for hearing about Hinton was through a radio broadcast. Her interpretation of Hinton's words was that he was saying AI would take over the world and get rid of humans, to which I replied, "Is it going to grow arms and legs?"
Which brings me to the question of why? Why do we persist in poking the bear? Why do we keep pushing the envelope? We should all know by this time, the enormous environmental impact of AI, in the water to cool the data centres, in the electricity needed to power them. So, as we face significant environmental challenges related to climate change, where water will be a precious resource, and electricity production could increase impact on climate, why?
As a millennial born in the mid-1980s into a low-income family, I remember a time when not everyone had a computing device, and the world wide web wasn't a thing. We did OK. We still communicated, via phone and radio, over long distances. We still had mass communication through television and print. We still could listen to music, on record, tape, or radio. We could even use computers, even the internet, without AI. It is possible to do with less and/or limited technology. We could just say no, but the more we chose to integrate AI into those previously analog and basic digital systems, the harder it will be to extricate the system intact if we need to down the road. Like most problems in the world this is man-made.
I grew up before the internet and spent countless long summer afternoon engrossed in books, while listening to vinyl records and was perfectly happy doing so. I lacked nothing. If I wanted to learn about something, I went to the library.
Your question as to why we are where we are, unfortunately, I think the answer lies in the desire of many to build up huge amounts of wealth. True, there may be some idealists among them, but they would be nowhere without massive capital investment. They have made fortunes out of highjacking teenage (and adult) brains.
Speaking as a non-academic, I think I can say that those of you on the front lines of the Humanities are the true heroes of our era! You have the opportunity to reshape education and make the Humanities the centerpiece of our culture. I admire your approach of experimenting and sharing feedback on how different strategies might work.
I'm sure this increases the teaching burden with zero additional compensation, but if it's any consolation, I have found the Humanities the most important and long-lasting part of my education. My only regret is that I never reached out to thank all those responsible.
Thanks, Paul. Most of us in the humanities, at least in my department, are fighting the good fight. I don't see us returning to a central position in the culture any time soon, but we will go down swinging!
Gemini made me part of the nefarious "we!"
Recent ChatGPT "mishap." My wife and I dozed too much during the first episode of Unforgotten, Season 5. We started episode two and realized we had no idea who was who. So I asked GPT for a summary of the first episode. It confused us more because it was just wrong about who and where the characters were. It was annoying, then funny how wrong it was. Similar to its map attempt in your post.
I do think it's good for research. But mostly to go a level deeper. I needed to know cutting-edge nail polish colors in 2012 (a novel set then) and for that it was great. I could then go to google and click through to ads for the color I chose (for the name––Malice Red)
And I still think that humanities classrooms should have "dumb" devices that just allow a student to type out an essay. I pity the teacher who'd have to decipher my handwriting. Or I pity myself for having to write in block letters so slowly to make myself legible that I wouldn't be able to complete my essay.
Last year I quit smoking what Geoffrey Hinton is smoking, and I don't miss it, so I won't say "I want some of what Geoffrey Hinton is smoking"; but that's what I think when I read his confident pronouncements about the inevitability of constructing systems we cannot describe, or blindly iterating to them, or catalyzing their emergence, or however he thinks about it.
Thanks, Mills. I needed to hear that.
Gary Marcus: https://open.substack.com/pub/garymarcus
Apple’s sense of LLM limits: https://ml-site.cdn-apple.com/papers/the-illusion-of-thinking.pdf
My hot takes on LLMs are getting cooler by the second; everyone is realizing “AGI” isn’t coming. Gary Marcus is a strong writer on this subject, if you want more encouragement! Apple’s most recent (very bearish) research paper on LLMs is good too, will link to both below!
That's great! Will read both forthwith!
I’m about to co-teach an introductory seminar on classical antiquity, and I think this is brilliant. If my colleague is agreeable, we’ll try out some of your ideas. For what it’s worth, I started my teaching career before desktop computers.
I think it's an almost impossible task to expect your students to engage themselves rather than engage AI, particularly in these days of large class sizes and remote study. There maybe some potential still in the Oxbridge 'personal tutorial' approach where there is close discussion with a tutor in a (very)small group of students. There is little scope for device use for conversation/discussion in such a small and focussed group. A requirement for essays for assessment would need to be in person, hand written or via a simple word processor with no internet access. To be honest, I rather like the ideas that trying to avoid AI produces - in person reading and discussion, small groups, invigilated, in-person internet-free examinations, using a library etc etc. Like stepping back into a more engaged educational context. One that I personally missed out on and still long for. There is hope if this can take hold, but it's expensive and intense. It also requires those teaching to be deeply and widely engaged in their subject - and I'm not sure that's always the case now.
Fair points, but most of my courses with English majors and graduate students are small and discussion-driven, and most of the student buy in enthusiastically. And even in the larger general-curriculum courses with non-majors, there is always a cohort of students who come in good faith and are interested in the work. I teach to these students and accept the fact that I can't reach all of them, though I try.
Also, of course, I speak from a UK perspective, which is likely different…
I’m sorry, but this is simply delusional: “We can emphasize these skills by designing assignments that require students to connect course material to their own lives, to engage in nuanced debate, or to produce something genuinely original. The emphasis should be on the process of thinking and creating, not just the final product.”
There’s no requiring anything. Every step, even the most personalized, can be fed back to the LLM.
Absolutely. As I said to a room full of English majors yesterday, our only option is to trust each other. Perhaps I'm naive, but I am convinced that most of them are participating in good faith. I accept that this is likely not the case in general-curriculum courses.
It's also likely not the case in English courses taught by faculty the students don't like much (I assume they all adore you).
Perhaps a bigger concern is how much AI threatens pedagogical property. I've been thinking about this as a book coach, how I offer some free resources as lead magnets that could be turned into AI prompts. Not as good as working with me, but probably more serviceable than I'd like to think.
So all that syllabus sharing and assignment sharing could encourage more brazen plagiarism of colleagues. Here's this assignment John did, tweak it for my class in the American Novel, etc. And the more savvy students, probably including some majors, might feed old drafts with your comments on them into the system either for tips on improving new work or as a way to "train" the LLM in your standards. Or maybe those resources could be leveraged to produce a "close reading" of a text that aligns with your rubric.
You haven't mentioned Hua Hsu's essay in the New Yorker, but I think he shows pretty plainly that even above average students are doing this without a lot of guilt.
AI shows evidence of having been designed to appeal to the cognitive biases, such as confirmation bias. The socratic approach has the potential of unlocking a key to using AI, which is that the questions have more importance than the answers. Answers turn on facts, which are always changing. The questions are more durable.
Just for fun, I consulted my friendly neighborhood bot for suggestions on topics for a English Lit graduate seminar on literary criticism using Ulysses as the vehicle. Here's what I elicited. I am no authority on either literary criticism nor post-modernism, but it seems a passable simulacrum.
1. Authority, Narrative, and Voice
* To what extent does Molly’s unpunctuated monologue destabilize traditional narrative authority?
* Does the lack of punctuation democratize meaning-making, or does it obscure voice in ways that reinforce Joyce’s (male) authorship over Molly’s “voice”?
2. Embodiment and Language
* How does Joyce’s rendering of Molly’s embodied sexuality challenge or reinforce the tendency of modernist texts to intellectualize experience?
* Does the linguistic stream of bodily sensations subvert male rationalist discourse, or does it reinscribe it by turning embodiment into literary experiment?
3. The Politics of Representation
* Is Molly Bloom’s soliloquy a feminist gesture, granting a woman subjectivity at the climax of the novel, or does it remain a male-constructed fantasy of female consciousness?
* How might we reconcile the radical form with the problematic provenance of its creation (Joyce ventriloquizing a woman’s inner life)?
4. Intertextuality and Deconstruction
* How does “Penelope” undermine or echo earlier chapters of *Ulysses*? For example, in what ways does it resist Bloom’s own attempts to define or narrate Molly throughout the novel?
* What happens to meaning if we read Molly’s “Yes” less as affirmation and more as différance—a sign that endlessly defers closure?
5. Subjectivity and Postmodern Multiplicity
* Does Molly’s stream of consciousness exemplify postmodern subjectivity—fragmented, decentred, contradictory—or is it still caught in modernist faith in interiority?
* How might her soliloquy resist or collapse binary oppositions (male/female, speech/writing, desire/domesticity, fidelity/infidelity)?
6. Ethics and Readerly Complicity
* How are readers implicated in the voyeurism of accessing Molly’s intimate thoughts?
* Does the novel invite us to critique that voyeurism, or simply indulge in it under the guise of high art?
Interesting. Some of those are pretty interesting questions, though some are predictable. My problem with students using AI for this kind of brainstorming is that human brainstorming (for me, at least) is a vital part of the process of discovery and of the writing process in general.
I was just discussing Geoffrey Hinton's claims with my mother. My mother is almost entirely analog in her approach to technology - no matter how many times we have tried to teach, she cannot navigate 'smart' technology - so her source for hearing about Hinton was through a radio broadcast. Her interpretation of Hinton's words was that he was saying AI would take over the world and get rid of humans, to which I replied, "Is it going to grow arms and legs?"
Which brings me to the question of why? Why do we persist in poking the bear? Why do we keep pushing the envelope? We should all know by this time, the enormous environmental impact of AI, in the water to cool the data centres, in the electricity needed to power them. So, as we face significant environmental challenges related to climate change, where water will be a precious resource, and electricity production could increase impact on climate, why?
As a millennial born in the mid-1980s into a low-income family, I remember a time when not everyone had a computing device, and the world wide web wasn't a thing. We did OK. We still communicated, via phone and radio, over long distances. We still had mass communication through television and print. We still could listen to music, on record, tape, or radio. We could even use computers, even the internet, without AI. It is possible to do with less and/or limited technology. We could just say no, but the more we chose to integrate AI into those previously analog and basic digital systems, the harder it will be to extricate the system intact if we need to down the road. Like most problems in the world this is man-made.
I grew up before the internet and spent countless long summer afternoon engrossed in books, while listening to vinyl records and was perfectly happy doing so. I lacked nothing. If I wanted to learn about something, I went to the library.
Your question as to why we are where we are, unfortunately, I think the answer lies in the desire of many to build up huge amounts of wealth. True, there may be some idealists among them, but they would be nowhere without massive capital investment. They have made fortunes out of highjacking teenage (and adult) brains.