“The future is now,” Liza wrote.

In 2020, a book written by a machine was published in Slovakia. This event offers opportunities to consider literature outside our everyday patterns. To rethink the concept of authorship and its various forms, creativity, cooperation, and receptive activity, while copyright and the essence of literary community also enter the debate.

Liza Gennart is the name of a neural network that writes contemporary Slovak poetry. It is based on a fine-tuned GPT – 2 language model. Thanks to it, the neural network has learned certain language patterns and tries to replicate them or write in a similar mode. Therefore, all the words in the poems were created by the neural network itself. The GPT – 2 language model was trained on the English texts from Wikipedia and Reddit. Thus, it learned a general English language structure: from word formation and sentence creation to whole paragraphs. If we wanted to make this neural network create poetry in the Slovak language, the input material of a large number of contemporary Slovak texts was needed to tune the GTP – 2 language model. This emphasized the neural network parts that create the patterns resembling these poems. We created a big database (2108 works) from (mostly) poetic collections published by contemporary Slovak publishing houses: Drewo a srd, Aspekt, FACE, Skalná ruža, Dive Buki, Európsky dom poézie and Ars Poetica. We also added all the numbers of cultural and literary journals VLNA, Glosolália and Vertigo. To fill the database, we used also some relevant poetic works from the digitized collection of copyright free works called Zlatý fond denníka SME.