Justin E. H. Smith's Hinternet is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. A close friend and collaborator of mine was telling me the other day what great leaps and bounds GPT-4 has been making recently. He showed me as proof one of the texts this AI language model had generated for him, in uncanny imitation of the playful and hyper-erudite style he and I had both effected in our collective work of historiographical metafiction,
The best episode was the musical episode featuring Ethel Merman, Ann Miller, Carol Channing and Della Reese. This is an objective fact. Any AI that fails to recognize it has disqualified itself from serious consideration.
Mar 31, 2023·edited Mar 31, 2023Liked by Justin Smith-Ruiu
AI is already useful for some very particular things, but it’s really a marketing term/metaphor, like “Smart Water”. AI is not close to being intelligent in a literal sense. There is, in fact, no such thing as literal artificial intelligence and I don’t know why we would think there would be, nor, most importantly, why we would want it anyway. However, there is what I take JEHS to have called ‘artificial stupidity’. Natural stupidity has its place, and even charm. And after all, it’s the Font from which All Stupidity Flows. But the artificial kind - woah. Now you’re synthesizing stupidity and feeding it back to yourself. It’s an idiot-mold to stuff the human brain into. Why would anyone *want* do that (other than marketing)?
There’s lots of yack about PoMo, but the AI fallacy has an affinity with what functional post modernism really is: a cultural economy resulting from decades of fear and austerity politics, in which everything is thought to have been done and said already, and we can never go *back in the day* when people could be spontaneous and un-self concious. That world is the Before Time and it’s gone forever. We can never be that CUTE again. All we can do nowadays is categorize what’s extant, and fetish it endlessly. The closed loop of only old old material resembles the foreclosed possibilities and mutations in literal AI due to what we have decided intelligence is. It can’t be *generative* no matter how much data is fed into it. When you live in a lapsed world, maybe this affinity w/AI makes it “seem right”.
Btw, my reaction to ‘digital quasi-officialdom’ stuff like you’re talking about w/CA, taxes, etc, is also rage, tears, depression - so don’t feel too bad. Government and corporate bureaucracies are the way they are because the individual user is precisely the last person anyone designing this stuff cares about.
In a couple of years, it will read you a transcript of Charro and the doctor from the love boat on Hollywood squares. You’ll be up in the right hand corner commenting it all in your style, and it will all be in the most obscure language of your choice. You may even be able to have this in images, choosing your leather jacket instead of a sport coat. You’ll direct a whole episode during your lunch break for fun. You’ve got the gpt of 2021. And it’s boxed to make sure it doesn’t hallucinate. It would seem that it is currently testing humans, not the other way around, which is why we’re getting the “aw shucks mister, I’m just a LLM doin’ my best” and all the praising possible for Open AI. Oh, and the NYT article on the Emirates convention yesterday? Looks like a whole lotta Charros couldn’t shimmy this hell outta place.
I thought I would stop reading articles on ChatGPT, but it just can’t be escaped. Luckily the last two I’ve read (this one and another one in the Sydney Review of Books) have been quite good. The one in the SRB mentioned how ChatGPT and other natural language processors used millions of emails from Enron as training data, a fact that I feel needs to be explored more. Didn’t you have an essay tilted something like “HR Managers of the Human Soul”? Pretty soon we’ll be forced to interact with these chat bots speaking corporate babble, all thanks to some clueless mid level managers who thought they’d hit the jackpot - and this will be the standard of communication (if it isn’t already). To end on a slightly more positive note, every time I hear something along the lines of “ChatGPT will make writers obsolete,” I think of two writers, you and Roberto Calasso, who I find so idiosyncratic that they couldn’t be imitated convincingly. Glad to see I’m right!
Apropos of your last bit, I work for a large, multinational company. It’s end of (fiscal) year performance review season. I’ve heard multiple colleagues praise GPT for making it super easy to “write” performance reviews. Performance reviews are already denuded of most human content because of the HR-mandated form. Add to that this new layer of inauthenticity (i.e., algorithm-produced job performance feedback) and it really is . . . just stupid, but even more so. It’s an interesting point you make about us unwittingly giving away our freedom in response to technology advances (here the freedom to demand or expect something approximating a genuine and meaningful human interaction between manager and employee) without ever even noticing. As you note, this results from the structure of the system. I was born into a world where (at least in the US) employer/employee relationship dynamics are mediated through an HR function whose chief aim is to avoid litigation and an IT administrative state that radically circumscribes the form that employer/employee relationship can take. Under these conditions, dignity is impossible. And 99% of us (myself included) don’t believe we have the freedom to do anything about it.
Dear Doctor Smith (I often remember Lost in Space when thinking of you!),
Being a big fan of your philosophical and humoristic prose, considering you the parodic genius or our eon, and having more or less exhausted your substack, I ordered a copy of "In search of the third bird" at my local bookshop, expecting to receive in exchange for my hard-earned euros some slim volume, more or less the size of "The internet is not what you think it is", comprising short stories and other wherewithal for my amusement, from you and your two co-authors, whom I did not know except for the excellent company they keep.
You may imagine my surprise when I received, a week later, a ponderous volume, more or less the size of a critical, bilingual edition of the Iliad and Odyssey, complete with preface, postface, introduction, chronology, variants, abundant notes, apocrypha etc, some 780 densely packed, well printed, fine-smelling with new-book odor pages. All on the issue of the existence or not of a society so secret that anyone who admits knowing of it is almost certainly an impostor.
For the life of Proteus, where do you find the time, the energy, the verve to produce such an enormous amount of text, all of it funny to be sure, but so enormous? Did you have access to some ChatGPT-37 Pro from the future who wrote it for you? Where do you get the energy, if not? What is your secret?
Your amazed admirer (and paid 'stack subscriber, so technically your boss of sorts),
Doesn't seem that AI has come along much since teenage me and friend would get into similar back-and-forth-going-nowhere dialogs with "Eliza" on my TRS-80 -when the program was loaded by standard cassette tape!
I am reminded of that episode of Star Trek where Captain Kirk induces the tyrannical AI computer to self destruct using pure logic... There must be some version of the Liar's Paradox we can throw at these infernal machines.... as smoke rises, circuits fry, and humanity is liberated!
Thanks for this essay. I'm finding it impossible to be sanguine at the moment about this technology's impending effects on society and on the world of writing and literary study. It seems certain to encroach on our daily lives in the ways you describe, unless reigned in by regulation immediately, and while it's harder to say how it will disrupt what we think of as "creative writing", the fact that it can already do pitch-perfect imitations of writers' styles indicates to me that the art of parody, at least, will soon belong to machines.
In my own fooling around with ChatGPT, I've noticed that when you ask it to write "in the style" of a particular writer, it usually returns the sort of facile paragraphs you quote here. However, if you ask it to give you an actual passage from a writer, it will frequently make up a text that sounds a lot like the writer, but isn't actually in the work in question. For example:
Prompt: "Please write a paragraph imitating the style of Henry James in The Beast in the Jungle"
Answer: "As I sit down to pen my thoughts on this particular topic, my mind is filled with a sense of foreboding, as if something ominous is lurking just beyond the corner of my vision. It is as if a beast, hidden and waiting, is about to pounce upon me, rending me apart with its merciless claws. This sense of impending doom is not new to me, but rather a constant companion, haunting me day and night. I cannot shake the feeling that there is something important that I must do, some destiny that I must fulfill, but I cannot fathom what it might be. And so, I wait, ever vigilant, for the moment when the beast in the jungle will reveal itself to me, and my fate will be decided."
Prompt: "Now give me an actual paragraph from The Beast in the Jungle"
Answer: "Certainly! Here's a paragraph from "The Beast in the Jungle" by Henry James:
"It was the sense of a figure for whom the visitant had waited. The person on the scene was the real presence, compared to which any mere idea of him, any possibility, any danger he had hitherto conjured up, became as ghostly and unreal as the preliminary alarm of night fears. The latter dropped straight from him as he at last fairly recognized and took in the figure, that of a man also evidently, who, equally still and with the same recorded attitude, as if, arrested by the shock of the encounter, he had stopped short of contact, only returned his gaze. The two men stood confronted, and it was as if the more recent-comer had immediately passed, on his side, into high detachment, into the sanctuary of a deep, still felicity from which he looked down at his companion's headlong course.""
The first passage sounds nothing like James, the second quite a bit like him (but, importantly, is not actually from "The Beast in the Jungle.")
When you then ask it to tell you where in the text the passage is from, it will backtrack and apologize, admitting that it's not actually there, and will claim that what it gave you is a "paraphrase." Basically, it doesn't know the difference between paraphrase, parody, plagiarism, and direct quotation. But it is capable of coming up with striking images and figures that seem as if they could have been written by a certain writer (and maybe have been, since it's impossible to know where it's getting these words from), even if the overall effect is parodic.
Glorious. Do try it with gpt 4 because I am expected so much more based on the things I read and my experience was similar but I didn't do as many clever prompts.
Dude, to paraphrase the title of your post, it might be you, who is "really quite stupid". It is widely known and reported that Bing is a rather poor instance of GPT-4, edged towards providing simplistic Bing searches and rephrasing the metadata found through those, not much else. Use OpenAI-s web interface if you want to try what GPT-4 is capable of.
The best episode was the musical episode featuring Ethel Merman, Ann Miller, Carol Channing and Della Reese. This is an objective fact. Any AI that fails to recognize it has disqualified itself from serious consideration.
What a bravura piece. The last line lands with force. Call them demons, angels, or bots --why did humans choose to inflict them upon ourselves?
AI is already useful for some very particular things, but it’s really a marketing term/metaphor, like “Smart Water”. AI is not close to being intelligent in a literal sense. There is, in fact, no such thing as literal artificial intelligence and I don’t know why we would think there would be, nor, most importantly, why we would want it anyway. However, there is what I take JEHS to have called ‘artificial stupidity’. Natural stupidity has its place, and even charm. And after all, it’s the Font from which All Stupidity Flows. But the artificial kind - woah. Now you’re synthesizing stupidity and feeding it back to yourself. It’s an idiot-mold to stuff the human brain into. Why would anyone *want* do that (other than marketing)?
There’s lots of yack about PoMo, but the AI fallacy has an affinity with what functional post modernism really is: a cultural economy resulting from decades of fear and austerity politics, in which everything is thought to have been done and said already, and we can never go *back in the day* when people could be spontaneous and un-self concious. That world is the Before Time and it’s gone forever. We can never be that CUTE again. All we can do nowadays is categorize what’s extant, and fetish it endlessly. The closed loop of only old old material resembles the foreclosed possibilities and mutations in literal AI due to what we have decided intelligence is. It can’t be *generative* no matter how much data is fed into it. When you live in a lapsed world, maybe this affinity w/AI makes it “seem right”.
Btw, my reaction to ‘digital quasi-officialdom’ stuff like you’re talking about w/CA, taxes, etc, is also rage, tears, depression - so don’t feel too bad. Government and corporate bureaucracies are the way they are because the individual user is precisely the last person anyone designing this stuff cares about.
In a couple of years, it will read you a transcript of Charro and the doctor from the love boat on Hollywood squares. You’ll be up in the right hand corner commenting it all in your style, and it will all be in the most obscure language of your choice. You may even be able to have this in images, choosing your leather jacket instead of a sport coat. You’ll direct a whole episode during your lunch break for fun. You’ve got the gpt of 2021. And it’s boxed to make sure it doesn’t hallucinate. It would seem that it is currently testing humans, not the other way around, which is why we’re getting the “aw shucks mister, I’m just a LLM doin’ my best” and all the praising possible for Open AI. Oh, and the NYT article on the Emirates convention yesterday? Looks like a whole lotta Charros couldn’t shimmy this hell outta place.
I thought I would stop reading articles on ChatGPT, but it just can’t be escaped. Luckily the last two I’ve read (this one and another one in the Sydney Review of Books) have been quite good. The one in the SRB mentioned how ChatGPT and other natural language processors used millions of emails from Enron as training data, a fact that I feel needs to be explored more. Didn’t you have an essay tilted something like “HR Managers of the Human Soul”? Pretty soon we’ll be forced to interact with these chat bots speaking corporate babble, all thanks to some clueless mid level managers who thought they’d hit the jackpot - and this will be the standard of communication (if it isn’t already). To end on a slightly more positive note, every time I hear something along the lines of “ChatGPT will make writers obsolete,” I think of two writers, you and Roberto Calasso, who I find so idiosyncratic that they couldn’t be imitated convincingly. Glad to see I’m right!
Apropos of your last bit, I work for a large, multinational company. It’s end of (fiscal) year performance review season. I’ve heard multiple colleagues praise GPT for making it super easy to “write” performance reviews. Performance reviews are already denuded of most human content because of the HR-mandated form. Add to that this new layer of inauthenticity (i.e., algorithm-produced job performance feedback) and it really is . . . just stupid, but even more so. It’s an interesting point you make about us unwittingly giving away our freedom in response to technology advances (here the freedom to demand or expect something approximating a genuine and meaningful human interaction between manager and employee) without ever even noticing. As you note, this results from the structure of the system. I was born into a world where (at least in the US) employer/employee relationship dynamics are mediated through an HR function whose chief aim is to avoid litigation and an IT administrative state that radically circumscribes the form that employer/employee relationship can take. Under these conditions, dignity is impossible. And 99% of us (myself included) don’t believe we have the freedom to do anything about it.
This was a good text, Justin, generated with your characteristic flair, intelligence, wit and subtle humour.
This was a good text, Justin...
Thanks for another comforting essay on AI to go with "No Minds Without Other Minds".
Curious what you make of Robert Wright's contention that, with the advent of GPT-4, AI is capable of a human seven-year-old's level of cognitive empathy? https://nonzero.substack.com/p/chatgpt-is-capable-of-cognitive-empathy
Dear Doctor Smith (I often remember Lost in Space when thinking of you!),
Being a big fan of your philosophical and humoristic prose, considering you the parodic genius or our eon, and having more or less exhausted your substack, I ordered a copy of "In search of the third bird" at my local bookshop, expecting to receive in exchange for my hard-earned euros some slim volume, more or less the size of "The internet is not what you think it is", comprising short stories and other wherewithal for my amusement, from you and your two co-authors, whom I did not know except for the excellent company they keep.
You may imagine my surprise when I received, a week later, a ponderous volume, more or less the size of a critical, bilingual edition of the Iliad and Odyssey, complete with preface, postface, introduction, chronology, variants, abundant notes, apocrypha etc, some 780 densely packed, well printed, fine-smelling with new-book odor pages. All on the issue of the existence or not of a society so secret that anyone who admits knowing of it is almost certainly an impostor.
For the life of Proteus, where do you find the time, the energy, the verve to produce such an enormous amount of text, all of it funny to be sure, but so enormous? Did you have access to some ChatGPT-37 Pro from the future who wrote it for you? Where do you get the energy, if not? What is your secret?
Your amazed admirer (and paid 'stack subscriber, so technically your boss of sorts),
Monsieur de Combourg
What an entertaining read this was!
Doesn't seem that AI has come along much since teenage me and friend would get into similar back-and-forth-going-nowhere dialogs with "Eliza" on my TRS-80 -when the program was loaded by standard cassette tape!
I am reminded of that episode of Star Trek where Captain Kirk induces the tyrannical AI computer to self destruct using pure logic... There must be some version of the Liar's Paradox we can throw at these infernal machines.... as smoke rises, circuits fry, and humanity is liberated!
Thanks for this essay. I'm finding it impossible to be sanguine at the moment about this technology's impending effects on society and on the world of writing and literary study. It seems certain to encroach on our daily lives in the ways you describe, unless reigned in by regulation immediately, and while it's harder to say how it will disrupt what we think of as "creative writing", the fact that it can already do pitch-perfect imitations of writers' styles indicates to me that the art of parody, at least, will soon belong to machines.
In my own fooling around with ChatGPT, I've noticed that when you ask it to write "in the style" of a particular writer, it usually returns the sort of facile paragraphs you quote here. However, if you ask it to give you an actual passage from a writer, it will frequently make up a text that sounds a lot like the writer, but isn't actually in the work in question. For example:
Prompt: "Please write a paragraph imitating the style of Henry James in The Beast in the Jungle"
Answer: "As I sit down to pen my thoughts on this particular topic, my mind is filled with a sense of foreboding, as if something ominous is lurking just beyond the corner of my vision. It is as if a beast, hidden and waiting, is about to pounce upon me, rending me apart with its merciless claws. This sense of impending doom is not new to me, but rather a constant companion, haunting me day and night. I cannot shake the feeling that there is something important that I must do, some destiny that I must fulfill, but I cannot fathom what it might be. And so, I wait, ever vigilant, for the moment when the beast in the jungle will reveal itself to me, and my fate will be decided."
Prompt: "Now give me an actual paragraph from The Beast in the Jungle"
Answer: "Certainly! Here's a paragraph from "The Beast in the Jungle" by Henry James:
"It was the sense of a figure for whom the visitant had waited. The person on the scene was the real presence, compared to which any mere idea of him, any possibility, any danger he had hitherto conjured up, became as ghostly and unreal as the preliminary alarm of night fears. The latter dropped straight from him as he at last fairly recognized and took in the figure, that of a man also evidently, who, equally still and with the same recorded attitude, as if, arrested by the shock of the encounter, he had stopped short of contact, only returned his gaze. The two men stood confronted, and it was as if the more recent-comer had immediately passed, on his side, into high detachment, into the sanctuary of a deep, still felicity from which he looked down at his companion's headlong course.""
The first passage sounds nothing like James, the second quite a bit like him (but, importantly, is not actually from "The Beast in the Jungle.")
When you then ask it to tell you where in the text the passage is from, it will backtrack and apologize, admitting that it's not actually there, and will claim that what it gave you is a "paraphrase." Basically, it doesn't know the difference between paraphrase, parody, plagiarism, and direct quotation. But it is capable of coming up with striking images and figures that seem as if they could have been written by a certain writer (and maybe have been, since it's impossible to know where it's getting these words from), even if the overall effect is parodic.
Glorious. Do try it with gpt 4 because I am expected so much more based on the things I read and my experience was similar but I didn't do as many clever prompts.
Dude, to paraphrase the title of your post, it might be you, who is "really quite stupid". It is widely known and reported that Bing is a rather poor instance of GPT-4, edged towards providing simplistic Bing searches and rephrasing the metadata found through those, not much else. Use OpenAI-s web interface if you want to try what GPT-4 is capable of.
yeah me too