samanthas_synth is an experimental videowork exploring visualisation of machine cognition and computational thinking. The video shows the algorithm trying to make sense of various random input images, altering and sorting them to fit the algorithms neural structure and preferences. 

Using a Deep Learning Neural Network trained on a vast dataset of tagged images, we created short video loops of interpolations between one neurons maximum activation and a random input imgage. Each frame is a synthesised image resulting from the networks attempt to alter the input image in a way that activates a chosen neuron to its maximum - played in short succesion, these images represent the networks process of sculpting the input to match the neurons preferences. Embracing the main limitation of this process - the low resolution of the images of 227 x 227 - as creative inspiration, we edited these videos together to create a collage with a total resolution of 4K.

The piece premieried as a 4K videoinstallation at HEVRE Gallery Krakow as part of the exhibition Elementi 8: EKRAN - HEVRExINTERMEDIA on March 7th 2020. It was awarded the Niio x AI Art Prize 2020 alongside works by Jakob Kudsk Steensen, Kris Kore, Pascual Sisto, and Eyal Gruss and is part of the Niio cataloge. 

samanthas_synth

With „Samanthas Phone“, we tried to create a representation of how an AI Algorithm would perceive online dating spaces and attempt to interpet and structure them, inspired by technologies used for presenting suggested profiles and matchmaking. We used datasets of online dating profiles from platforms such as OKCupid and Tinder to train an AI that is generating questions about dating behaviour, directed the viewer.
Around this centre presence we used a Speech to Text Algorithm along with scans of various objects and found footage to build a visual environment of misinterpretation, confusion, and digitalism / cyberspace that is further enhanced throughthe audio composed for the piece. 

Add

B_No_Source [live]

Opportunities

For Artists

ArtConnect

Support