Night-Transport Truck is the product of a series of machine learning experiments and tests, which attempts to address the topic of generative narrative. A collaborative effort, we explored a proposed human-machine spectrum for contemporary narrative: human-told, human-interpreted; human-told, machine-interpreted; machine-told, human-interpreted; machine-told, machine-interpreted…

The following text contains selected passages from words generated by a recurrent neural network trained on a custom data set of compiled contemplative, first-person narrative literature.

The process to achieve such a text is outlined here:
1. Train the model on the customized text file (data).
2. Provide the model with a brief a prime text, which it will use to generate new texts.
3. Repeat the previous step, but replace the prime text with the last few words of the newly generated text.
4. Compile, curate, and edit.


Branching from this, we explored the visual latent space interpolation of images containing the generated text. In this experiment, prime text was consistent, but each key frame represents a different stage of learning in the algorithm's sampling process.

Made in collaboration with Reginald Lin and Annie Yu. For more information, refer here.

UPDATED 11.16.19, 11:21 PST