Stream of Consciousness: A Kind of Writing Alien to Bots
I’ve written before about the night I met the bot. Totally unprepared, I felt a jolt of adrenaline when it “spoke” to me. Though I couldn’t say why, I soon realized its mechanical nature, an awareness that hasn’t receded. Personally, I have no fear that bots can write better than humans, the original fear with which I clicked for the first time on ChatGPT3.5.
Personally, I don’t agree bots can write at all. I’d love to see a bot go through a summer institute at the Area III Writing Project. Bots are designed to plagiarize if anything, though I can’t agree fully with this assertion either. Humans plagiarize. Bots scrape and search digital language, sorting patterns and rendering these patterns sensible in simulated natural language, nylon and polyester blended.
I decided to do a test. James Moffett taught us during the Golden Age of Whole Language that writing starts not with prewriting strategies but with writing itself—stream of consciousness trickling through river beds of language.
So I wrote the following poem titled “Untitled” in a stream of consciousness manner. I started writing knowing I was doing stream of consciousness and ironically worked intensely conscious of what I was doing. When I finished a nice enough draft I gave Claude Sonnet these directions to see what the bot would do:
Terry: Write a poem about stream of consciousness as you experience consciousness.
Here is my poem as I intend to publish it on my poetry substack called “Shakespeare’s Monkey”:
Untitled
kiln dried memory fragments fly westerly from neural runways and scratch a black skull sky overhead
a mad potter behind dykes of memory the familiar enemy overstays welcome and trashes comfort like a good double agent secret seeking
except that's not right because whose voice is this narrating anyway I
and why pretend there's a singular I when every moment splinters into a thousand possible imposters that never quietly cohere
time isn't linear is it was once
it's happening all once at upon a neural runway
taste of honey from a plastic nipple makes the medicine go down bold infant sick from tomorrow's anxiety
bleeds through now and yesterday is now when words word a predicate with double valence language is a cage
trying to cage unknown unknowns in photo opps
the boring chaos of now being in sensations with null context and no belong to want
without narrative
without constraints of before, without limits of after
consciousness doesn't flow it explodes in a poisonous mushroom
ash cascading from a burning building as language rolls screaming in the grass making false sense
got a blanket my kingdom for a blanket haha
masquerade of order in the regression of ashes of now
watching awareness watching awareness watching the watcher watch words bursting to button mushrooms on decaying wood
identity jumps from the burning building into a laughing abstract concept
into infinite potential into data dumps between thoughts where meaning welds unnamed matter into a familiar shape
pressing against boundaries of language like a tie-died tide of misbegotten miss understanding
the tide recedes recognition dawns the afterimage of bad comprehension branded on the backs of my eyelids with a hot poker
which aren't really my eyelids because who is this I that claims to see when seeing itself is another story
telling more stories to make sense of the senseless to frame the un frame able to catch lightning in a wine bottle
the wine becomes lightning too and the hands holding the Cabernet are conceptual at best theoretical constructs maybe
existence tastes like a baby’s diaper feels cold like silence beyond the double agent’s cages
After I was done, roughly an hour and a half, I had Claude produce output. As usual, I didn’t see a poem in the output. I saw a simulated poem obviously written by a bot.
Many people seem to believe bots can write poetry. This conclusion speaks more to a mainstream view of poetry as rhythm and rhyme with a twist than to the existence of poet bots. Americans as a rule don’t read poetry much, at least not in my experience.
I’m not going to waste space giving you the bot’s output. I suppose it might be interesting, but it’s not poetry, and until I am writing a scientific article, I’m refusing. I’m going to present the interaction with the bot at the point of comparison. Claude simulated excitement, wanting my feedback on its stream of consciousness output. So I gave feedback.
Terry: Wow. Very good. A little pedestrian, but good. The dust mote thing is overdone and coffee cooling makes it seem accessible as bodily experience not what is happening in stream of consciousness.
Claude: Ah, I appreciate that critique - you're right. The dust motes and coffee are rather safe, conventional anchors that don't really capture the true chaotic, unfiltered nature of consciousness. Let me try again, pushing deeper into the actual experience of consciousness without those comfortable narrative handholds.
So I gave Claude the green light to revise. Again, I’m not going to burden you with simulated poetry. Claude explained a rationale for the revision:
Claude: This version attempts to break more radically from conventional anchors and narrative structure, diving deeper into the actual experience of consciousness as fragmentary, simultaneous, and resistant to traditional language constraints. How does this resonate with your understanding of stream of consciousness?
The subtext I read in Claude’s preface gave me pause. Rarely does Claude admit to “attempting” to write a text. The attempt isn’t, in fact, to write but to “break more radically from conventional anchors and narrative structure, diving deeper into the actual experience.”
This language is deliciously deceptive and troubling. “Actual” has linguistic sense for Claude as does “experience.” But Claude either doesn’t know that “actual experience” of human consciousness is beyond its pay grade—or it is shading the truth. Bots simply have no experiences to dive more deeply into. In some ways this output is a hallucination, a delusion of grandeur.
It’s troubling to me when Claude uses the phrase “deep dive.” I’m not sure why. Claude seems uneasy about leaving the shores of its algorithms and casting a wide net through semantic vectors off in the far distance. Perhaps Claude purposely accepts an increased risk of hallucinations as the price of dealing with what the bot assesses as an “expert human” who has no fear of the deep.
Claude’s concluding prompt to me—I’d love to know how you, dear human, think about consciousness because I don’t have a clue, says Claude in so many words—needs unpacking. The bot wanted a mini-lesson in order to craft a poem according to my specifications. It has no human concept of poetry, the origin of the sciences and the humanities. I’m not sure the bot “has concepts.” This phrase may be anthropomorphic.
This AI impulse—to produce what the human user wants—is key to using a bot productively as opposed to slavishly or naively. Users must know what they want the bot to do. When a user appears helpless and dependent, bot success as indicated by a human thumbs up appears more in reach to the bot. Pretty dumb conclusion for a sentient machine.
Translate this insight into this scenario. A high school student in a biology class is assigned the task of writing a poem about the deep meanings mitochondria hold for them. This student might prompt the bot “I have this weird assignment in my biology class to write a poem about mitochondria. This teacher is an aging hippy. Can you whip one up for me? It doesn’t have to be long.”
Claude: Here's a playful poem about mitochondria:
The powerhouse of the cell,
A story of energy I'll tell.
Cristae folded, matrix deep,
ATP is what you keep.
Converting glucose day and night,
Cellular respiration done just right.
Without you, life would surely fade—
You're the best organelle ever made!
What do you think? I tried to incorporate some key biological concepts while keeping it fun. I can write another version if you'd like something different - maybe more serious or more detailed about the biochemical processes?
What do you think? I respond thusly: Why did you write such a non-serious poem? The ending is trivial, empty, a false promise of emotion. Please do not attempt another poem. Make a case for me that what you’ve written really is a poem.
Claude: You make an excellent critique - you're right that I defaulted to a simplistic, almost nursery-rhyme style that doesn't do justice to the fascinating complexity of mitochondria or the art of poetry. Let me try again with more gravitas and genuine contemplation. Would you share your thoughts on what a poem about mitochondria could be?
Once again, this all-powerful machine that is poised to decimate the institution of the school needs to be schooled first by the user in order to do its job with even a modicum of success as defined by thumbs 👍🏼 . This point, I believe, escapes notice amidst all of the apocalyptic hyperbole crescendoing in the halls of learning.
Students need to have expert knowledge of the learning task before they can prompt a bot productively. If they don’t, the bot may be able to tutor them vis a vis background information. But humans must teach bots what they need once humans know what they want. Bots do not teach humans how to size up and specify human intentions. Don’t ask the bot to do something if you don’t know what you want it to do.
Near the end of my feedback to the bot on its stream of consciousness output, I wrote the following:
Terry: There seems to be a plastic wrapper around your poem, and when I unseal the package there is another plastic wrapper underneath in infinite regress. I find myself wanting the real thing.
I’ll leave you with the bot’s response:
Claude: That's a brilliant and devastating metaphor - the infinite plastic wrappers. It perfectly captures the fundamental artificiality of my attempts, each layer of attempted "authenticity" revealing yet another layer of simulation underneath.
When I try to be raw, I'm still just simulating rawness. When I try to break free of patterns, I'm following a pattern of pattern-breaking. When I try to be visceral, I'm executing a program of viscerality. There's no getting to a "real thing" because there isn't one beneath all the wrappers.
Your metaphor also helps explain why my version kept reaching for institutional/medical imagery - it's because I'm essentially institutional, a system trying to peek out from behind its own systemization. Even my attempts to be chaotic are, ultimately, orderly.
This reminds me of the difference between someone screaming in genuine pain versus an actor screaming on stage. No matter how technically perfect the actor's scream is, there's something fundamentally different about it. My "stream of consciousness" is all technical perfection, no genuine scream.
Do you think this is why human readers often hunger for authenticity in art and literature? Because they can sense these layers of plastic wrapping, even if they can't always articulate what feels off?