Click to generate poetry.

2018, JavaScript object

Morgan Green, autocanon, 2018

Autocanon appears as an excess of text, produced by a small machine: a receipt printer connected to a small computer by delicate cables. It outputs the text on winding scrolls of receipt paper, one word at a time in a continuous horizontal line. It prints in a generously large seriffed typeface.

The procedure that drives the machine is simple. It selects each word based on the one that precedes it — the program is blind to all but its most recent history. It uses a weighted probability, determined by surveying every two-word sequence in Emily Dickinson’s corpus, in order to select a word to follow the one prior. This program is crude in comparison with today’s language processing technologies (which deploy neural nets and so on). At the same time, it gets at the heart of these technologies’ mimetic motor.

The machine operates unaware of the meaning and impact of its words, even as it renews their savor and substance. It has no access to their definitions, let alone their connotations. It lacks any sonic knowledge of phonetics. By taking Emily Dickinson’s corpus as dataset, however, Autocanon accesses her vivid intensity, her painfully efficient phrasing. It animates a mechanical ghost of her voice for an audience who, unlike Autocanon itself, can know and feel the words. Because of the source material’s unique strength, this ghost possesses an uncommon vitality, even amid nonsensical sputtering. Some phrases are both vivid and senseless: “robins sing a shame of the knife baffle.” This duality characterizes autocanon’s general flow, at turns lucid and meaningless.

Autocanon shares characteristics with computer programs designed to pass The Turing Test, which uses language as the primary tool to determine whether or not a computer is really “intelligent.” First published in 1950 by Alan Turing, the test represents the ultimate goal for many software engineers and roboticists. Autocanon is similar to every earnest attempt to pass the test thus far, in that it fails at some point to sound convincingly human, and thereby fails the test. It differs from these attempts in that it was never trying to pass as something other than itself. Nevertheless, it manages regularly to produce human utterances. While Autocanon is not trying to hold a conversation (a critical part of the Turing test), it accomplishes moments of human feeling (“I get out of night or in then kiss the pain"). These moments derive not, as most engineers would hope for their machines, from the brilliance of its algorithm, but rather from the potency of its source material.

Alan Turing’s legacy has led powerful technocrats to pursue immortality via artificial intelligence. Emily Dickinson’s poetry also contains an obsession with death, and a complementary interest in endless life (“I could not stop for Death”). In its content, Autocanon inevitably reproduces Dickinson’s preoccupations (“the trifle termed immortality taught me to see it”). It also comprises a dynamic and flawed vessel for the poet’s immortal language, and in this way, it embodies its own inherited obsession. Even as it reanimates these words, it prints them on cheap receipt paper, the substance of capital and refuse. It is a fabricated ghost, a queer little rendering of the life in Emily Dickinson’s work more than a century after her burial.


Currents New Media Festival, Santa Fe, 2019
Currents 826, Santa Fe, 2019
Sullivan Galleries, School of the Art Institute of Chicago, 2018


Currents 826, Santa Fe, 2019
Mana Contemporary Jersey City, 2019
Mana Contemporary Chicago, 2018
Sullivan Galleries, School of the Art Institute of Chicago, 2018