Ted Underwood has recently written about machine learning as cultural technology: A more interesting upside of AI. While I don’t find that the idea of LLMs as cultural technology tells us much about how they actually work, that’s not what interests Underwood. He remarked in a recent comment on Facebook:
But my own attitude has taken a strong turn toward Marx's line, "the point however is to change it." I'm still interested in understanding cultural history. But I'm interested in it more and more as a map of resources we could actively remix to produce new things. If LLMs don't do that exactly the same way as the human brain, I think ... I might be fine with that?
I’ve certainly got nothing against changing the world. In fact, I’ve been approaching that in some of my recent use of ChatGPT and Claude. Let’s take a look.
Sonnet 129
Consider this recent blog post: Toward a Tantric reading of Shakespeare’s “Th' expense of spirit in a waste of shame,” with two mandalas.
First I had ChatGPT create a mandala based on Shakespeare’s Sonnet 129, “Th’ Expense of Spirit.” Then I asked it to gloss the poem’s concluding couplet: “All this the world well knows; yet none knows well / To shun the heaven that leads men to this hell.” Here’s what it concludes:
So yes, that couplet acknowledges the tragic bind not with moralistic condemnation, but with rueful solidarity. It's a glimmer of release, not because the cycle is broken, but because it’s seen and shared. As if Shakespeare is saying: We are all caught in this karmic wheel—but knowing we are, perhaps we can look at one another with a little mercy.
Then I ask it about the hunter simile in lines six and seven, and possibly eight as well. Here’s part of my prompt:
In it we have a hunter laying poisoned bait to catch an animal. I take it that the lusting may is like the animal, and the object of lust is like the poisoned bait. But where in the body of the poem is there a figure parallel to the hunter? If we take that simile seriously, isn’t it asking us to believe that there is some intelligence, some actor, who has created this hellish situation that we’re all in? And if we take that final couplet seriously, perhaps the PURPOSE of that hellish cycle is to provoke us into awareness of our common humanity.
In my own work on the poem, that is central. It may be a secular poem, but it displays a pattern found in Christian theology, and pattern that informs Milton’s Paradise Lost. ChatGPT has no trouble responding to my prompt. Here’s part of its response:
Tantric traditions, especially in some strains of Hinduism and Buddhism, deliberately do not avoid the energies of desire. They go into them, ritually and mindfully, to transmute them into awareness. Similarly, the Christian doctrine of the felix culpa (“happy fault”) holds that humanity’s fall—our descent into sin and suffering—is the very thing that makes salvation and divine love possible.
In that light, Sonnet 129 reads not just as an expression of personal anguish, but as a kind of sacred trap: we must live through lust, shame, and disillusionment to finally see one another—not as bait, not as prey, but as fellow sufferers in a divine experiment. The hell of lust is not just punishment—it is revelation.
It concludes: “Shakespeare becomes almost a mystic, using sonnet form like a mantra to push us into confrontation with the sublime.”
How does this respond to Underwood’s desire to treat LLMs as “as a map of resources we could actively remix to produce new things.” Well, a tantric reading of an ostensibly secular poem from a Christian culture, that seems rather new to me. I suppose Stanley Fish might regard it as a reading generated within a different “interpretive community,” one populated by Hindus and Buddhists. But there’s no interpretive community here at all, just the “cultural technology” of the LLM underlying ChatGPT.
It’s not at all clear to me just what to make of this, and to be honest, I’ve not thought all that much about it. At the moment I’m mostly interested in the fact that we can do this sort of thing at all. That’s new.
Here’s more of that some kind of thing.
Shakespeare, Disney, Coleridge
Here’s a working paper: Claude on the Eightfold Way: Sonnet 129, Fantasia, Kubla Khan. It records a conversation I had with Claude about those three texts. Here’s the abstract, which I’ve edited out of a series of summaries I’d asked ChatGPT to create:
A) The circular structure of Shakespeare’s Sonnet 129 mirrors contemplative technologies found in both Hindu/Buddhist and Christian traditions. The sonnet's recursive exploration of desire creates what can be understood as a secular meditation technology that operates through psychological rather than explicitly spiritual means, transforming destructive cycles into opportunities for deeper understanding.
B) Walt Disney's Fantasia (1940), functions as a cinematic mandala structured around the mystery of sound-sight relationships. Despite appearing as popular entertainment, Fantasia employs sophisticated contemplative architectures, including radial organization around central insights, recursive motifs (particularly hands as symbols of creative consciousness), and a progression that builds contemplative capacity rather than advancing linear narrative.
C) Coleridge's “Kubla Khan” provides the most striking example of unconscious contemplative mastery. Structural analysis reveals the poem's recursive architecture, where the emblem “A sunny pleasure-dome with caves of ice” functions as what modern AI research would recognize as a semantic attractor—a compressed representation in high-dimensional meaning space that contains the informational content of the entire preceding text. This insight, impossible before the advent of large language models, illuminates how consciousness naturally creates contemplative technologies through semantic compression and decompression cycles.
D) There is a consistent meta-pattern across all examined works: (1) presentation of a psychological or existential problem, (2) circular exploration that deepens rather than resolves the problem, (3) semantic compression into forms that contain the entire exploration, and (4) recursive return that transforms the initial problem into contemplative opportunity. Is this a fundamental cognitive architecture that emerges when consciousness operates at optimal integration?
E) Finally, the conversation itself becomes an example of the contemplative consciousness it analyzes, suggesting that such deep structural patterns represent universal features of human cognitive architecture rather than culturally specific techniques.
The general point is the same as it was with the previous example, involving Sonnet 129 alone: We have a Hindu and Buddhist thought applied to secular texts from an entirely different culture.
And there’s something else, something specific to Stanley Fish. He made his name as a theorist by announcing something he called “affective stylistics” in an article published in 1970. The idea is that the meaning of a text is to be found in the process that unfolds as the reader moves through a text, word after sentence, after paragraph, etc. It’s a method he’d arrived at in his book on Milton, Surprised by Sin: The Reader in Paradise Lost (1967). Forget the fact that his criticism is not affective, and forget that, once he’d announced the method as such in the 1970 article, he didn’t much practice the method, whatever it is. What’s interesting about this working paper is that ChatGPT is interested in the process by which the reader, or audience member in the case of Fantasia, moves through the text.
It talks of (in D) “circular exploration,” “semantic compression,” and “recursive return.” And who’d’ve thought of comparing these three texts? Well, I did, not ChatGPT. But it was able to do the work that validated the comparison.
Are we changing the world yet? Probably not, but things are looking interesting.
Homo Ludens, Pleasure, Gross National Happiness
As you probably know, I’m working on a book whose current title is, Play: How to Stay Human in the AI Revolution. In that book I will be arguing that we live in a world where the institutional structure is dominated by the economic metaphor of the efficient marketplace. In that world one’s place in society is tied to one’s place in the occupational system. As AI displaces people from the occupational system they will, in effect, be left to drift in a world that is now bereft of meaning for them, something I explore in a recent article in 3 Quarks Daily, Trapped in Work Mode: The Real Challenge of AI is not Technical, It’s Conceptual, Mythic, and Institutional. We need to replace an ethic of Homo economicus with one of Homo Ludens.
I note that the small kingdom of Bhutan has already done that. In 2008 Bhutan adopted a new constitution, one in which Gross National Happiness (GNH) becomes that nation’s guiding philosophy. To that end Bhutan has instituted an index for measuring GNH.
While happiness and pleasure at by no means the same thing, they are closely allied. But Western culture has never quite known what to do with pleasure. We tolerate it, market it, even celebrate it in the abstract—but we rarely take it seriously. Pleasure is seen as secondary, indulgent, or morally suspect. Hinduism and Buddhism, however, are better suited to the intellectual task of grounding an ethos where pleasure is treated as a first principle. That’s something I’ve explored in conversation with ChatGPT: Redeeming Pleasure through Hinduism and Buddhism.
Now we’re talking about changing the world. Just how that will happen, who knows? But I don’t doubt that the creative and effective use of LLMs, cultural technology, will be part of the process.
No comments:
Post a Comment