As "artificial intelligence" grows more complex, it's modes of conception and misconception may start to overlap with that of a human's. The video, "Deeply Artificial Trees", represents the result when a system tries to find objects which are not present, while the audio is a generative representation of what a machine "thinks" it is hearing. The video is an experimental investigation into machine learning and perception of AI.
The images are generated by an algorithm called "deep dream" which looks for patterns of models (say a spider) in a image. If the image does not actually have a spider in it, the system then proceeds to draw one in the place where a pattern looks closest to the model, in effect "seeing" something which is not there. The audio is generated by an algorithm similar to one called "wavenet" which learns a sound and is able to then generate that sound from a model.
More info: Instagram
Deeply Artificial Trees (excerpt)
Spider crawly
What?
Deer things
