More analysis here: www.youtube.com/live/a6gj7E4...

Which is very, very far from the claim that "AI-generated poetry is indistinguishable from human-written poetry and is rated more favorably"

I know it's just another AI bro paper, but it sucks that almost all the papers I read are like this now.

I think a proper comparison could have been made if they compared the AI poetry to authors known for having good meter and rhyme like Poe or Suess even.

But as it stands now, all the paper is saying is that "people like meter and rhyme"

No duh people are going to like a poem with meter and rhyme to Geoffrey Chaucer, a poet from the 1400s when middle english was all the rage.

I mean, what even is this list?

Here is the poet list: Geoffrey Chaucer (1340s-1400), William Shakespeare (1564-1616), Samuel Butler (1613-1680), Lord Byron (1788-1824), Walt Whitman (1819-1892), Emily Dickinson (1830-1886), T.S. Eliot (1888-1965), Allen Ginsberg (1926-1997), Sylvia Plath (1932-1963), and Dorothea Lasky (1978- ).

Without a direct comparison between real and AI poets, it's hard to claim the two are "indistinguishable."

Ok. Fine. Let's just ignore the "indistinguishable" part and more on to the "rated more favorably."

Then there's this plot.

Notice how the blue and orange bars are offset? Were they originally planning on providing a direct comparison between AI Whitman and human Whitman, but then removed that data for some reason?