Here’s another conversation (clash?) between Thomas Brady of Scarriet and Christian Bök on AI and poetry.
First you can listen to the conversation between Anthony and Christian, the link below which Brady found on FB.
Listen to Episode 5 of “Fate of the Arts” (in which Anthony Etherin and I hang out, chatting about THE XENOTEXT and the impact of AI on the future of poetry). —CB
Enjoyed the discussion but the duality Bök articulates is moot. No choice is necessary between “emotions recollected in tranquility” and “negative capability,” between careful Wordsworth and ecstatic Keats. Bök went so far as to assert that it’s “impossible” to write poetry from the self’s “truthful experience” and in the grip of “present feelings,” simultaneously. I found this a very puzzling assertion, since writing presupposes the freedom of the poet to manage a poem’s composition in any manner which strikes the poet’s fancy—the indirect quality of any poem must be assumed to exist, simply on account of the nature of language, on the nature of poetry and on the poet’s freedom already alluded to. That indirect quality necessitates a hidden intent—is the Wordsworthian “telling of his story” sincere? That question can only be answered by biography—which is not the discipline of the poem. AI has no Wordsworthian biography, which makes Bök’s heralded division irrelevant to AI poetry, but we can’t demand less of AI poetry. This isn’t fair, especially if, going forward, we are supposed to take AI seriously (and Bök is nothing if not enthusiastic on the potential of AI poetry). Bök, it seems, wants to have his cake and eat it, too. According to Bök, AI has infinite potential as a poet. But, also according to Bök, human poets are limited by aesthetic (Wordsworth v Keats) constraints. No. The very opposite is true. The human poets will always surpass AI poetry precisely because constraints (variety, not just obstacles) exist for human poets (Wordsworth v Keats is not random—it has real sources)—which are—thru creativity and variety itself—overcome.
Christian Bök responds:
You’ve misconstrued the distinction: the dialectic in Romanticism arises between two types of “expressive” poetry (the Wordsworthian models) and two types of “non-expressive” poetry (the Keatsian models). In the Wordsworthian models of the “egotistical sublime,” poetry is either “emotion recollected in tranquility” (i.e. “cognitive”: self-conscious and self-assertive) or “the spontaneous outburst of feeling (i.e. “rhapsodic”: not self-conscious, but self-assertive). In the Keatsian models of “negative capability,” poetry is “impersonal” (suppressing the self on behalf of other forces that might speak): it is either “mannerist” (not self-assertive, but self-conscious) or “automatic” (neither self-assertive, nor self-conscious). These four ways of “willing” (being self-conscious) and “telling” (being self-assertive) correspond to four different kinds of games (mimesis, ilinx, agon, and alea), testing different aptitudes with different rules for success — and judging the outcome of one game by the rules of another game constitutes the reason for most philistinic, parochial disputes among poets. For now, AIs play the “automatic” game (which we judge for the quality of its oracular uncanniness, arising from an ergodic pretense), while most poets believe that the only game in town is the “cognitive” game (which we judge for the quality of its original sincereness, arising from a mimetic pretense). The poets of the “egotistical sublime” often regard the poets of “negative capability” with much suspicion — and hence, such poets dislike AI (for the same reasons that they dislike avant-garde poetry written by drawing words from a hat or by rolling dice from a cup). The “poets” who play one game, while discounting the validity of the other games, might lack some capaciousness of imagination, which we associate with “creativity” itself.
Thomas Brady answers:
Thanks for clarifying, Christian! I don’t think it changes my critique, however. A poem’s “expression” or “cognitive” presentation of a “self”— labels or descriptions such as these—fall short of what any poem in the context of being a poem is doing. Within this context words such as “expression” and “self” are merely words. A person could be shouting at us—and we might naturally term this as “expressive.” But the individual could be simply raising their voice on a whim, or in fun. With this simple example, I can topple the scholars’ theories of “expression” in poetry. It isn’t that I don’t accept these categories which indeed are “real;” I do, and respect them. We get them from Wordsworth the man (as well as the poet) and from comparing letters to Fanny Brawne with poems on Fanny by Keats. However, in postulating on AI, all our cogitations lie in the future (AI is still “embryonic” in its development, as you said). Keats and Wordsworth belong to the past. The future human poets (AI poets, too) can, and will, use irony (the most vital tool of poetry, in my opinion) to conflate categories of “expression” and “non-expression,” “self” and “non-self.” I believe the “games” of poetry which you entertain are far less real and solid than you maintain. Again, they have a source, but we have no choice but to trace these sources back to Wordsworth with a headache and Keats with a cough. That is, history, which is real, but not real for poetry, per se. You are thinking like AI already in a present-towards-the-future and therefore you have already lost yourself, as I see it, as a poet and a critic. AI parrots the scholars, even if they are wrong or limited—human data is all AI ultimately has to go on. The “superior” AI ability to “run” with something, if it starts with a false premise, will only increase error algorithmically. We will always have Keats as he was, and you and I discussing Keats over here, and AI over there. AI will never be able to cross the boundary to where Keats was and to where you and I are. I think this is very important to keep in mind. For our sanity, if nothing else. These categories are far more legitimate, I believe, than the “games” of poetic division which you are positing—in the act, I believe, of turning yourself into AI. Let’s not do that.
Christian (now in the fight of his life) explains himself:
History certainly “feels” real for the poets who write in response to the poets of the past. Are not poets the worst because they “parrot” scholars (right or wrong)? Are we humans no less constrained than AI by the data that we have “to go on.” Do not poets often “run” with something by starting from a “false” premise (just like any machine), turning some “conceit” into a total trope? Do not poets enjoy “increasing error” through the use of a rule to see where it might take their language? Is there no room for “play” amid all this serious concern. I do not “posit” the four games described (as something theoretical): they already exist before us, built into the very structure of writing itself (at least insofar as we admit that writing involves some relationship to permutations of intentionality and expressiveness). I might even suggest that AI is the perfect example of “negative capability” — insofar as it appears to be impersonal, with no ego, acting as an “agent” through which larger forces, outside itself, might be birthed into existence.
I admit that I distrust the poets who tell me that some boundary is “never” going to be crossed in the world of poetry (because I, myself, have crossed a boundary or two, pulling off several “impossible” feats that my peers in their cruelty have dismissed as “never” going to happen) — and I witness the advancements of machines, which are never going to play chess better than humans, never going to solve any unsolved, mathetic conjectures on their own, never going to navigate a roomful of obstacles with human grace, never going to invent a new drug, a new gene, a new tool, or a never going to paint a unique canvas in the exact style of Rembrandt (beautiful enough, down to the brushstroke, to fool the experts), etc. — and yet AI has already done all of these things (and much more, even the Rembrandt), while we, in turn, keep readjusting the benchmarks for what must constitute a “pass” into our fellowship of creativity (shifting each time to a newer “never” for reassurance).
The critics seem to dismiss each of these accomplishments of AI in the same way that abusive parents might rebuke a child for not being so “mature” in its development as an adult — (whereas I remain curious about the “growth” of such a child, taking an interest in its primitive, but colourful, drawings in crayon, tacked to the fridge). The poets of the “egotistical sublime” dismiss “negative capability,” I think, in part because unself-conscious, unself-assertive art implies that creativity might not be “special,” inhering in the “self” of a given, human being — but instead might be an inherent property of the universe itself, available to everything within it, even machines, giving expression to itself in whatever way seems fit for the job. Cheers!
Thomas Brady cruelly closes the door on AI:
Absolutely human poets are flawed—they “parrot,” they begin from “false premises” and all the rest. The major religions might even say the fact that humans are flawed is the whole point, but your “AI religion” (if you’ll allow me that phrase for the sake of argument) wants the whole point to be that AI has no flaws and that there is nothing (at least in the world of poetry, if not consciousness itself) which is inaccessible to AI. You defend AI as if it were a child of yours. As a radical critic of AI’s claims, I understand my claims might upset you—if not AI.
AI is quantity—it has no quality. We know that quantity can look like quality—this is a pretty good definition of the “sublime,” a key term for the Romantic poets. Negative Capability implies a view of the Sublime in which a poet’s admiration of quantity (like the height of a mountain swooned over by Shelley) does not interfere with speculation on how much quality really does belong to whatever particular example of the Sublime is under review. I don’t think “ego” has anything to do with Negative Capability. There are unhealthy egos and healthy egos—it is unfortunate for AI that it has none. Keats coined Egotistical Sublime (to attack Wordsworth—the Lake Poet was also mocked by Poe and Byron) and Keats also invented Negative Capability. Most poets of the Egotistical Sublime would never admit being such. AI is not capable of Negative Capability. It can have no doubts about anything.
AI is the result of a fast computer. I agree it can “do wonderful things” for us. But it’s no Child of Poetry. Blame my opinion here on my “egotistical sublime,” if you will, but per this discussion, (and sure, AI belongs to “the universe”) it matters not.
*********
Scarriet 