I recently read Mary Oliver’s wonderful book, Rules for the Dance. Oliver is one of America’s great poets, and a colleague here at the University of Florida (Dr. Brenda Smith) introduced her to me. In chapter 9, a short chapter of just two pages, Oliver (1998) briefly talks about how the sound of a poem conveys meaning too. She notes, for example, that the liquids l and r are “just that, watery sounds; they suggest softness, fluency, motion” (p. 61). And to illustrate her point, she shares an excerpt:
Pillow’d upon my fair love’s ripening breast,
To feel for ever its soft fall and swell …
(Keats, “Bright star, would stay steadfast…”)
Oliver contends that every category of sound carries meaning. Some “words rasp, or whistle, or sigh. Or shout. Or whisper.” In other words, sound is never neutral. The other example in this chapter is instructive as well. Oliver notes that the “letters representing mute sounds are b; c, k (hard), q; d; g (hard), p and t. Each is a quick wack of a sound, emphatic and vibrationless, a sound that refuses to elide with any other” (p. 60). To illustrate this idea, she shares the sentence:
The dog jumps into the car
Did you notice what happened as you read that sentence? Oliver writes that there is “an unfillable instant of nothingness” between the letters g and j. And when such silences are used in a poem, they act as “a snap and a click, … an enforcer of the self-containment, and so the certainty of what has been said.” Once again, she illustrates with an example:
‘Men work together,’ I told him from the heart,
‘Whether they work together or apart.’
(Frost, “The Tuft of Flowers”)
One of the most dramatic illustrations of how sound communicates meaning can be found in the poem, We Real Cool by Gwendolyn Brooks. In his delightful introduction to poetry class on Coursera (Sharpened Visions), Douglas Kearney analyzes the effect a reading of this poem has on a listener. The author’s choice of words and the structure of the poem give it a rebellious sound. That is, until one hears the final line – we die soon – that sounds, as Kearney says, “like a coffin lid being slammed down.” It signals the abrupt end to not just a poem but a life. As is the case with Oliver’s examples, a poem’s aural landscape imparts meaning too, a meta-meaning, if you will, that goes beyond a mere reading of the words.
The importance of these poetic observations for artificial intelligence is instructive. For too long, AI has been hobbled by an anemic understanding and definition of intelligence. The Turing Test fails to capture the nuanced, complex, and multi-dimensional nature of human intelligence. In fact, Alan Turing hoped to sidestep all the difficult questions linked to intelligence by developing a simple binary test. If a human could not tell if they were interacting with a machine – instead of another human – via a simple chat interface, then the machine, at least in Turing’s mind, was “intelligent.”
And with that, the field of artificial intelligence was off and running. For many years, the field stumbled forward like a deaf mute, but text-to-speech and speech-to-text technologies have now changed that dynamic. After converting a sound sequence – think poetry reading – to a spectrogram, AI models can do some interesting things. But as we saw earlier, sound conveys meaning too. That is, a human can infer the presence of a coffin lid just through the sound of a poem itself. This dimension of human intelligence and meaning-making has yet to be fully explored. The time has come for AI to face the poetic music…
My first introduction to the sound and color of words came from a college professor as he read poetry. To me, words are like music notes or dabs of color on an artist’s palette, waiting to be placed in just the right places to produce a word symphony or artistic masterpiece. Choice and placement is the work of highly creative artists!