Katherine C.

Sunday, December 4, 2022

Music 256A / CS476A, Stanford University

Word count: 404

Final Reading Response

This week, we read Ge’s essay “Humans in the Loop: The Design of Interactive AI Systems” and watched Allison Parrish’s talk “Experimental Creative Writing With Vectorized Words”. I enjoyed both pieces. The single idea/theme that I choose to respond to is Parrish’s pioneering work in NLP on creating writing by representing English words in vector spaces based on their phonetics.

This concept of converting words into vectors with dimensions corresponding to sounds, rather than meanings, is new and very fascinating to me. Previously, when thinking about the problem of determining the similarity between two texts, I was bound by traditional NLP methods such as tf-idf and word2vec that represent words mathematically based on their frequencies, distributions, and context, which all reveal something about their meaning. Parrish gives an overview of these methods in the first half of her talk, so initially, I thought the video would just be about the application of these well-known methods in a novel realm (creative writing and poetry), as in the Frankenstein Genesis example.

However, I was then surprised by her experiments with phonetic similarity, specifically, when she said, “Poetry is not just about word choice and meaning. It’s also about sound and the voice. Speaking poetry out loud is very important to me.” This reminds me of several concepts from the textbook Artful Design. An obvious one is sound design – interesting sounds create motion over time, for example, when adding kiki and bouba sounds creates artful, fun-to-read poems. Another is technology in search of the sublime – when we focus on how words feel when we speak them with our mouths, and how words feel when we hear them with our ears, we are creating a deeply human aesthetic experience that is beautiful and has the ability to transcend the medium (poetry).

I am curious to see where Parrish goes next with her poetry experimentation and creative coding. So far, the examples shown all fall into the category of texts originally written by humans (e.g., book by Mary Shelley, poem by Robert Frost) and manipulated by the computer either semantically or phonetically with vectorized words. I wonder what sort of results would appear when the writing is fully generated by the computer. I am also interested in exploring how similar methods of computational creativity can be applied to other forms of art, such as digital images and music. Overall, this was an extremely provocative video to watch!