Robot artist to perform AI generated poetry in response to Dante
Dante’s Divine Comedy has inspired countless artists, from William Blake to Franz Lizst, and from Auguste Rodin to CS Lewis. But an exhibition marking the 700th anniversary of the Italian poet’s death will be showcasing the work of a rather more modern devotee: Ai-Da the robot, which will make history by becoming the first robot to publicly perform poetry written by its AI algorithms.
The ultra-realistic Ai-Da, who was devised in Oxford by Aidan Meller and named after computing pioneer Ada Lovelace, was given the whole of Dante’s epic three-part narrative poem, the Divine Comedy, to read, in JG Nichols’ English translation. She then used her algorithms, drawing on her data bank of words and speech pattern analysis, to produce her own reactive work to Dante’s.
Ai-Da will perform the poems on Friday night at Oxford’s Ashmolean Museum. While Ai-Da is not the first AI to have been taught to write poetry, organisers said Friday would be “the first time an AI robot has written and performed poetry, as a human poet would do”.
“We looked up from our verses like blindfolded captives, / Sent out to seek the light; but it never came,” runs one of her poems. “A needle and thread would be necessary / For the completion of the picture. / To view the poor creatures, who were in misery, / That of a hawk, eyes sewn shut.”
In another, Ai-Da writes: “There are some things that are so difficult – so incalculable. / The words are not intelligible to the human ear; / She can only speculate what they mean.”
Meller, an art specialist, said that the words and sentence structure of the poetry are all AI generated from Ai-Da’s unique AI language model, with “restricted editing”. “People are very suspicious that the robots aren’t doing much, but the reality is language models are very advanced, and in 95% of cases of editing, it’s just that she’s done too much,” he said.
“She can give us 20,000 words in 10 seconds, and if we need to get her to say something short and snappy, we would pick it out from what she’s done. But it is not us writing.”
Meller described it as “deeply unsettling” how language models are developing. “We are going very rapidly to the point where they will be completely indistinguishable from human text, and for all of us who write, this is deeply concerning,” he said.