Frank Hecker

The poets I’ve featured in this series are mainly those whom I encountered in the 1980s and 1990s, so even those who were alive at the time I first read them are now either dead or in their dotage. (Elizabeth Alexander is the youngest, at 61.)

I thought that for this next-to-last Sunday night poetry post I should feature some younger poets, ones new to me. After doing some Internet queries and sampling a selection of poems, I found the two poets I’m featuring tonight, Natalie Diaz and Danez Smith.

The poems of Diaz I found most compelling were those written about her brother; I chose this one, “My Brother My Wound,” for its surrealistic imagery (CW: violence):

I’ve previously said my piece that people who want to write, at least in an original manner, should just try doing that instead of trying to get a large language model to do it for them. My opinion applies even more strongly to the idea of prompting LLMs to write poetry, whether that be parlor tricks like “write my VC pitch in rhyming couplets” or more serious attempts to create poems that one might confuse with poetical works of past or present.

Like a dog trained to walk on its hind legs, these attempts may entertain for a moment but offer no lasting pleasure. And as with making a dog play the biped, they’re an insult to the true nature of the entity: LLMs are best at producing an average of human creativity and commentary directed along on a dimension or dimensions of choice. However, the best poetry is singular, not collective, the product of particular moments of a particular life.

But recommending poetry and poets — that’s a different task, one where a smoothed-out average of critical consensus is exactly what one might want. That’s the task I set out for ChatGPT in its GPT-4 mode.