ChatGPT and the case of dactylic hexameter
Another chat with ChatGPT.
This time, I asked it for a one-stanza parody of “The Raven.” It replied:
Once upon a midnight dreary,
As I pondered, weak and weary,
There came a knock upon my door,
From a creature, small and poor,
Who begged for just a little more
Of the bread and cheese I had.
Which reminded me that although GPT can do a surprisingly good job of rhyming, it is not good at poetic meter. After some inconclusive discussion of meter, I decided to ask it specifically for an example of a particular meter.
Me: Can you create an example of a line of dactylic hexameter?
GPT: Sure! Dactylic hexameter is a metrical pattern used in classical poetry, in which each line consists of six metrical feet, with each foot containing one long syllable followed by two short syllables (da-DUM da-DUM da-DUM). Here is an example of a line of dactylic hexameter:
"The brave and mighty warrior stood, his sword ablaze with fire"
In this line, each of the six feet is a dactyl (long-short-short), and the line follows the dactylic hexameter pattern. The long syllables are indicated by capital letters, and the short syllables are indicated by lowercase letters.
In case it’s not obvious, there are a few things wrong with that answer:
- Although “one long syllable followed by two short syllables” is a reasonable description of a dactyl (it would be better to say stressed and unstressed syllables, but long and short are close enough, I’d say), the immediately following description “da-DUM da-DUM da-DUM” is not dactylic at all. In that notation, a dactyl would go DUM-da-da. For example, the word excellent is a dactyl.
- The example line that it provides is neither dactylic nor hexameter. I would call it iambic heptameter—it’s more or less seven iambs. The word warrior would normally be a dactyl, but in a line that otherwise consists of iambs, I would expect it to be pronounced more or less as two syllables: WAR-yurr.
- The statement about capital letters and lowercase letters is just false—the line it provided doesn’t use letter case that way.
So I asked if it knew what a double dactyl was. It gave an incorrect but not completely offbase description, then volunteered an example which is almost entirely unlike a double dactyl:
Henry the Eighth,
He was a king
Who did a lot of things.
He married six times,
But only two were good,
Anne Boleyn and Jane Seymour,
The rest, he wished he could.
The first line is standard, and the second starts out well enough, but the rest of the poem largely gives up on meter, and certainly doesn’t follow the standard double-dactyl meter.
I continue to be impressed that GPT can generate poetry at all. But it really isn’t very good at meter.