A Heroic Achievement

Portrait of Pope c 1739 by Thomas Hudson (London National Portrait Gallery)

One of my first tests of the power of ChatGPT, back in late 2022, was to ask it to write a poem in the style of Alexander Pope. I chose the 18th century satirist for one particular reason: his mastery of ‘mock-heroic’ verse, in which he reimagined London life in the style of the classical poets Vergil and Homer.

Typical of the mock-epic style is the precise five-beat line (the iambic pentameter), and the conceit that the poem is addressed as a letter (or ‘epistle’). Many of Pope’s notable poems of the 1730s were written in this way to friends, acquaintances and wealthy patrons.

So I asked ChatGPT to write an Epistle in the manner of Pope, as if he were addressing that great pioneer of computing, Alan Turing. I won’t bore you with the technical details, but suffice to say, the resulting train of thought was … well, you will see. This was the first attempt:

“To Dr. Turing, master of the code,
Whose mind doth solve what none forebode,
Whose genius doth transcend the age,
And pave the way for progress' stage.

With humble pen, I write to thee,
In admiration and in plea,
To offer praise where praise is due,
And shed some light on what you knew.”

Not bad, but it didn’t completely pass the test. Phrases such as ‘what none forebode’, or ‘in admiration and in plea’ seemed forced, as if they had been chosen simply for the rhyme. And of course, these were not pentameters - they only had four beats to a line, rather than five, an error that I challenged at the time:

Me: Would you like to have another go at the first two lines of the poem?

ChatGPT: Sure, I would love to give it another try. Here's my revised attempt at the first two lines of the poem in iambic pentameter:

“To Doctor Turing, master of the code,
Whose mind unlocks what none have e'er forebode.”

Well, that was something of an immediate improvement, although when I asked it to repeat the exercise a day or two later, it was back to four beats in the bar - old habits die hard.

Roll on 2025. It’s been a mere thirty months since I first ran my experiment, and I asked ChatGPT, now much more book-learned, to repeat the exercise. Here’s the astonishing result, which is worth quoting in full. Why not subscribe, to read the rest?