The name “neural net” sounds like the other side of the coin where one might also find written “world wide web.” The focus is on connecting through numerous strands (ideas, people, commerce, etc.) but the nodes at which we meet are not always where we would expect. Using a neural net reminded me so much of my days as an improvisational theater performer in high school and college and the types of exercises and games that help with creative associations to build a narrative on the fly. We practice a craft and overtime, we get better at it and can even deliver fairly consistent quality even if each performance is totally unique and no one knows where it will lead to. It can still be a worthwhile and thought-provoking performance (apart from the pure entertainment value) when the performers are skilled enough and well-informed on cultural and current events happening.
In playing with the use of the “Talk to Transformer” neural net tool, I entered humanities type questions in complement to the literary texts I had used in topic modeling experiments, T.S. Eliot’s “The Lovesong of J. Alfred Prufrock” being top of mind. I started first by entering the question “What was T.S. Eliot trying to say about modern life in his poetry?” Surprisingly, the text that answered back said that we could learn about T.S. Eliot by contrasting his work with another poet. The text then goes on to talk explicitly about the other poet, leaving the question I posed and the answer returned feeling as empty as J. Alfred Prufrock was bereft of emotional courage in that titular poem. On the other hand, filling in with the knowledge I have of T.S. Eliot’s work, I could say there is something almost meta about the neural net’s answer given how much of Eliot’s work focused on isolation and uncertainty. The answer I got isolated Eliot’s work entirely out of sight and made me uncertain of what I could get out of this experiment.
I had more success with a shorter, more open-ended type question or statement on what poetry is and its purpose to us humans. I repeated the search a couple of times, and got fairly different responses back, with one focusing on poetry’s “potential” but stating one needn’t actually understand the words themselves but should instead focus on the subject matter and structure. A similar but not identical query on poetry then brought up a more practical and applied reason for it all, saying poetry was important for telling us how to live our lives as empathetic individuals in contrast to religion. This was then followed by what looks more like an op-ed clip about the suffering happening in Syria now with blame pinned on Western foreign policy.
I think the wildly different responses I received on this experiment with the neural net reflects a spontaneity and stream of consciousness style of human thinking and communication that is common in the average person and seems to have been successfully replicated and amplified in the neural net process. After all, the neural net is being programmed by humans and our idiosyncrasies as a species just becomes more visible when a different “species” can mirror back. I’d almost say what I found most useful about the neural net, then, is not what concrete answers it could give me to my discrete questions entered, but the tone and style of what it relayed back held a mirror up to a wider range of how a human brain works versus how a computer can.