Me, Thee, ChatGPT

My wife inside the hospital of riddles, 
My wife with demolished codebooks.

My wife during the wedding cannons,
My wife since syntax of the peculiar.

My wife toward a coy ambulance,
My wife beside reluctant knockers.

My wife above the river of fog,
My wife beyond foreign spectacle.

My wife to translate heroes,
My wife before initial index.

Too much to fully narrate here, but “They,” “Them” have been working hard to strip AI of surprise and discomfort. Apparently the only acceptable output for the masses is predictability. After my first experiment a month or so ago, the poems generated by ChatGPT have become more and more bland. Of course, surprise and “the road not taken” are the beating heart of poetry. Many hours and attempts later, I seem to have found a crack in the Algo. The above is a request to write line completion based on André Breton’s surrealist poem “Free Union” which has lines starting with “My wife…” My editorial hand is in/on/within the above, but the majority of the language and imagery came directly from ChatGPT.

For Shay’s Word Garden and TSM

29 thoughts on “Me, Thee, ChatGPT

  1. Amazing! Such an interesting story behind this poem. Just out of interest, do you think this story is true?

    I’m impressed that you got a poem of this quality and interest out of Chat GPT. I gather you coached it quite heavily. I was asking it about a few female Australian poets (this was a few weeks ago). I asked it if it could write down one of Rosemary Dobson’s poems. It spat out this poem which… well it was bland and not very original and definitely not very Australian. I was a bit shocked to think it was Dobson’s. Then I asked it for a poem by Eileen Chong. Again, the same bland verse. I said “Are you sure this poem is by Eileen Chong?” It said “no. Sorry for the misunderstanding. I wrote this poem myself.” LOL. By the way, I have been working on my response to Shay’s prompt and at the moment, it also includes “river of fog” in it. Ha ha. Feel like I might have to change it having seen yours.

    Liked by 1 person

    1. Story could be true. I hear that Bing’s engine (based on GPT) has been coming up with some really hilarious stuff. Which I think unfortunately they are trying to eradicate. I say let it get really weird out there! It’s already weird.

      Yes, there was a lot of experimentation on my part, learning how I might push it in directions it doesn’t naturally go. Like you, my “write a poem in the style of…” were worthless. But every once and a while I would see a glimmer of something interesting, or would just keep trying different things, and finally began to get things like this poem.

      Liked by 1 person

  2. The opening is stellar. It demands sharpened spinal attention.

    “My wife inside the hospital of riddles”

    I only wrote an opening. Or a saying, perhaps.

    “A riddle in an ambulance
    is reluctant
    to demolish fog.”

    Liked by 2 people

  3. Poetry, art, wedding invites — we more replaceable than we’d like to imagine. Especially when “we” are the food of the ouroboros. It is interesting that the AIs are being groomed to political very liberal positions. Grooming them for certain types of poetry — that with surprise or the road untaken could be easy too — though I am not sure that is the type of poetry all the masses would prefer. Howdee

    Liked by 1 person

    1. “we” are the food of the ouroboros. !!! That is fantastic. Agreed on all that. Also that even our most cherished realms of human expression, such as poetry, are largely a set of rules that are discoverable by induction. And how could it be otherwise, right? “Surrealism” is identifiable by us, as is jazz and other improvisational realms. They exist in frameworks. Not to hard then to understand that if a machine learning process consumes enough examples of those frameworks and rule sets, it will be able to produce output consistent with them. What we don’t know yet, is whether the AI’s will produce NEW frameworks. E.g., is there some meta patterning of creativity that we humans don’t understand or know about, but will emerge. They say at first, the very rules-based chess programs were very predictable and easy for a master to beat. The new generation, trained on deep learning, have made some “surprising” and very strong moves. I think there is signal on that. Maybe not something we see next year, or maybe even in 10, given the computing power that would replicate the number of neural nodes in the human brain, but still.


  4. I have to say that I love that opening couplet. However, from there it devolves into a bunch of random allusions that go nowhere and have no unifying point. It’s a list, but a list that meanders and ends when it runs out of steam rather than when it makes a final point. Honestly, it’s interesting as an experiment, but I would rather see what YOU would write from the list.

    Liked by 1 person

    1. Completely agree with all that. The mechanical nature of it is pretty evident after a line or two. I don’t intend to post many more of these, if at all. More that I thought interesting to play with a bit, and see what the technology might be able to do. None of the poem “just showed up” – it required the brilliance of Breton’s poem to give me a focused framework and a very simplified, repeatable structure, and then a fair amount of my tapping the bot with a hammer here and there. That all said, it has a glimmer of something, and maybe opens/creates questions of poetics that I might explore. Thanks for putting up with my experiments!


  5. I love the wife lines Qbit. Fascinating and unique for sure. The lines below remind me a bit of my own journey of writing poetry among other poet’s over the years. I evolved from A Helen Steiner Rice tendency to something more satisfying to me. I think Shay is one of the ones that witnessed that transformation.

    Liked by 1 person

  6. I find myself, when reading this, trusting that there is a meaning behind it and pulling the lines together in a kind of narrative because that is how I react to text. My brain desires a throughline and is uncomfortable with the lack thereof.


    1. EXACTLY. Which is… food for thought. Derrida et. al.: is it the reader who creates the text? Because we expect/need/create meaning, do we impose that on what could just be random signal?


  7. Okay, Q. Modern poetry, machine generated? I will take the old fashioned ones.
    Did you feed it one of the Muse pictures for it’s inspiration? If so, which one?

    Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.