In Monterey’s rugged expanse, where Pacific whispers weave ancient tales, Cypress shadows waltz with the rhythmic tides.
Garrapata’s canyons, cradling secrets close, redwoods embrace silence, on nature’s temple floor, the land’s heartbeat throbs.
Monterey pine, stoic sentinel on the shoreline, a witness to stories etched in enduring stone, each breath of wind, a stanza in the coastal hymn.
Asilomar’s dunes, where monarchs briefly pause, fragile wings painted in hues of sunrise, eucalyptus whispers tales of unseen migration.
Carmel’s streets, an echo of artistic musings, lupines unfurl along untamed paths, wildflowers nod gracefully in the sea-kissed breeze.
On Point Lobos, sea lions luxuriate on granite stages, cliffs cradle a theater of crashing waves, pelicans trace seamless circles in the cerulean sky.
Monterey Bay’s depths, a realm of enigmatic mysteries, kelp forests sway in a silent aquatic ballet, sea otters weave through an emerald canvas, and the humpback’s song echoes in liquid serenity.
Wharf’s clamor, Cannery Row’s whispered tales, echoes of industry, resilience of the eternal sea, in quiet coves, a tranquility persists, untouched by time, a poem woven by elemental hands.
Bit frustrating! ChatGPT initially popped out total drivel – here’s a taste:
“In Monterey’s embrace, where waves caress the shore, A haven of beauty, enchantment galore. Nestled ‘neath the sun’s warm, golden kiss, My hometown whispers tales of bliss.”
It stubbornly wanted to do 4 line stanzas that rhymed and were very flowery like it was trained on material from the 1800s. It didn’t follow “You are poet Gary Snyder.” It didn’t write remotely in his style. I finally got a bit of improvement by telling it to write in free verse and not rhyme, but it was limited in its originality. I felt like it would have done better with more prescriptive styles like limericks or haiku. It had a few serendipitous lines that I enjoyed:
“nature’s temple floor, heartbeat of the land.”
“wind’s stanza, hymn to the coastal wilderness.”
Here’s what happened when I told it to make it a haiku:
Monterey’s whispers,
Cypress shadows dance with tides,
Nature’s hymn unfolds.
Overall, the initial impression is “Wow!” But then you kind of look harder and it’s missing something. It’s intriguing to think about using this as a writer because it spits out all kinds of stuff and sometimes there’s a word or phrase that you can riff on. I decided to edit it to better understand what I didn’t like.
https://docs.google.com/document/d/13lwrcHlKtTt_hlu_HfTbo_buI9jW4CTZpMq30F4nOts/edit?usp=sharing
I don’t love my version, but I like it slightly better. I realized that it had a pattern – naming a place, then putting some details. It had too many descriptive adjectives so that’s something concrete I could use to make it less cheesy and trite (ie: no adverbs, limited adjectives, minimal similes). I don’t know if I would let it do the first draft next time. That could be interesting – to create the core structure myself, then use ChatGPT.
Created the image using Dall-e…like the poem: not quite! That’s some kind of freaky mangrove/cypress mash-up!
I met similar resistance when asking it to write like Wallace Stevens. Dr. Suess was a bit better, but not by much. It may have just been that it could use rhyme again . . . which it really, really wanted to do.
I think some of this intentionally built into the models to prevent (additional) problems around copyright. You also end up with some additional options with the paid versions that enable better prompt contextualization from specific source material and they allow the prompt writer to control things better (steerability).