AI Learning, or Just Facilitating?

In touch with a good friend and fellow writer yesterday, he commented on attending an AI academic’s lecture and not finding it that informative – certainly nothing new to his own areas of interest and experience. My friend was ‘playing with ChatGPT and poetry’ rather than engaging with the generating images/researching coding focus of the lecture, and he also commented on how AI ‘won’t write poetry at all well – it’s all rhyming doggerel’, and this is definitely my own experience with it.

I became interested in ChatGPT a year ago, December the 5th to be precise, after reading about it in The Guardian and then had my own engagement with it – quite general, with some disturbing aspects, but including the writing of poetry. You can find links to my posts on this here: https://gravyfromthegazebo.blog/?s=ChatGPT&submit=Search

After yesterday’s reminder of the poetic doggerel, I did return to ChatGPT having left it behind some months ago, and explored its rhyming propensity a little more. Nothing has changed in its algorithmic compulsion to rhyme – which is surprising considering the wealth of information and models at its disposal – but I did manage to coax it into a momentary break from that rhyming response. Still limited, but some kind of learning. If I am assed enough to return, I will ask Al (have called it this from day 1) to write a poem without rhyming, and will not be surprised at the familiar rhyming response. Here are yesterday’s exchanges:

rhyme1rhyme2rhyme3rhyme4

Note – the poems appeared in conventional stanzas, accentuating the rhyming, but these get presented as ‘prose’ paragraphs when copying to save.

Leave a comment