Chantelle Hami

1008115781
Sara Ameri Mahabadi
ENG287H1
March 11, 2024
Assignment #2

Choice: Generative AI Experiment

Throughout the past few years, artificial intelligence (AI) has made its way through
various literature. Exceptional poems are typically filled with emotion and lived experiences
which may seem difficult for an AI generator to create due to the lack of human nature. As a
result, I’ve decided to choose the generative AI experiment in order to discover if AI can
generate a well rounded poem with humanistic values. Regarding the prompt I submitted, I chose
romance as the genre because I feel it takes romantic experience in order to deliver a love poem
that is not cliché. By means of the prompts, I wanted to see if the AI responses would deliver
what I expected. Concerning the parameters, I decided to choose a 50 year gap between the two
prompts because I wanted to see the type of vocabulary that AI would generate between the two.
As for the other parameter, I decided to differ the reading difficulty because I feel the vocabulary
from fifty years ago is more difficult than the present day and I want to see if the AI generator
will pick that up; even with an easy read. When crafting my prompts I used specific words in
order to get the results I need. For example, “would be written in the present day/50 years ago”
was used in my prompt because I want the AI generator to generate a poem that can be written in
that era rather than written about that era. When it comes to results I expected the first poem to
describe an old style of love where the vocabulary is outdated. I expect the sentences to be more
straightforward and easy to grasp as well as using common phrases from that time period. As for
the second generated AI poem I anticipated to find a complex vocabulary and sophisticated
language. I expect to see metaphors, imagery, and a deeper interpretation of romance. As time
goes on it is expected to develop literacy skills and create poems that engage its audience.

In respect to the AI generated results, both poems did meet the expectations that I hoped
they would to an extent. Although the poems were on the cliché side, seeing as AI does not have
humanistic values, it was expected. Looking at the first stanza in both poems it is apparent that
the vocabulary and reading ability is a lot easier to follow in the second poem compared to the
first. For example, in the first verse both poems talk about how love was initiated in the twilight,
but the first poem takes an easier approach when stating “our love did bloom” rather than “love
unfurls”. The change of vocabulary from “bloom” to “unfurls” definitely increases the reading
difficulty. Unfortunately, the poems do not articulate the difference in historical period indicating
that AI did not grasp that information in the prompt. Both poems feel as though they can be
written in todays age, but with the exception of different reading levels. What surprised me is the
poem that is suppose to have a more difficult readability fluctuated between easy and difficult.
For example, regarding the first poem, the third verse reads “Hand in hand, we strolled the lane,”
but compared to the second poem the readability still stays the same with the verse “Hand in
hand, we wander through the night,”. I was surprised to see that the same phrases would be used
especially since most parameters were opposite in both prompts. My revised prompt changed the
outcome of the poem because it challenged the readers capability of following along. Most
verses required a deep dive to peel back the layers embedded within the words to comprehend
the beauty and depth. Concerning the parameters, I informed the AI generator to keep the length
at 15 lines, but I noticed when prompted to change the reading ability and the historical period,
the verses decreased. With a simpler vocabulary comes straightforward and blunt verses resulting
in less words being used whereas with a more difficult readability comes more complex lines and
more words being included.

In response to the results, I believe that good literature does not follow specific rules.
There are all types of great literature that comes in different forms and styles. Adding limitations
to literature, as I did in the experiment, keeps literature quite reserved and takes away from the
creative freedom that writers hold. Although the poem may be a cliché, the results coming from
an AI generator changes how we think about literature. It is known that literature comes from the
heart, human experience, and complex emotions, but if an AI generator can create a form of
literature without those values then how can we interpret what is true literature? If someone were
to read the AI generated poems and think they’re appropriate literature then does that effect our
interpretation of literature? In addition, how can one differentiate between a poem written by an
artist versus one written by an AI generator? In my opinion I believe that these poems count as a
form of literature, but it is up to the reader to determine whether its great literature or not.
Personally, I believe that both these pieces are not great literature due to style and perspective.
There is nothing fresh or original that comes from the poem as well as the rhyming style sets me
off. I feel as though the poem is focused on rhyming instead of picking intricate words that work
well for the story. Once again, it is up to interpretation if the poems are good literature or not, but
how can we call something good literature if the writers not got a heart?

Appendix

First AI Generated Poem:

Second AI Generated Poem:

“Create a poem with only 15 lines with the genre of romance that would be written 50 years ago
that doesn’t have a reading difficulty” prompt. ChatGPT 3.5, 13 Feb. version, OpenAI, 9
Mar. 2024, chat.openai.com/chat.

“Revise the poem, keep the genre and length, but change the reading difficulty to hard, and write
it as if it was written in the present” prompt. ChatGPT 3.5, 13 Feb. version, OpenAI, 9
Mar. 2024, chat.openai.com/chat.