My current conversations with ChatGPT are about Ash Wednesday , which starts like this.
Because I do not hope to turn again
Because I do not hope
Because I do not hope to turn
Desiring this man’s gift and that man’s scope
I no longer strive to strive towards such things
(Why should the agèd eagle stretch its wings?)
Why should I mourn
The vanished power of the usual reign?
If you ask ChatGPT for a poem in the style of Ash Wednesday, about a lighthouse for example, it does this.
In the lonely night, the beacon calls
Guiding lost souls to safer shores
A steadfast sentinel, it stands tall
A symbol of hope that forever endures
Its light shines bright through mist and rain
Piercing through the dark, dispelling fear
A beacon of light that will remain
Through every storm, every passing year
Etc etc. It always writes any kind of verse or pastiche in naff quatrains with a positive vibe. It’s very difficult to extract any other kind of verse from it in my experience, which seriously limits its usefulness for this application.
Just once, it came up with this.
Because I cannot see beyond the shore
Because I cannot see
Because I cannot see the light
Guiding sailors through the tempest’s roar
I stand here in the dark of night
(Why should the wind and waves not fight?)
Why should I weep
For those who venture out into the deep?
…etc., and while it has that model at the forefront, it can produce variations about a desert for example.
Because I cannot find my way
Because I cannot find
Because I cannot find the path
To lead me to the end of day
I wander here in the desert’s wrath
(Why should the sandstorm cross my path?)
Why should I regret
The course that I cannot forget?
But as soon as you start to veer off course and ask it to reference a different poem, it’s back to trivial happy quatrains again.
It’s great fun, though also frustrating. I don’t know if it’s possible to find a reliable way to create better verse with it, or if I’m doomed by the probabilities.