How can technology help us in preserving our past? How to reimagine it? What can a collective subconscious of the past tell us about our future?
To answer these questions, I trained an image generation neural network on a large corpus of transcription drawings of Sumerian cuneiform tablets and used it to generate entirely new cuneiform writings. I iterated on the output of the cuneiform-AI and created a variety of new hyper-digital/physical/augmented reality versions of this ancient form of memory-encoding. In the latest iteration I brought the AI cuneiform together with some of the original tablets at the Met-Museum - in AR.
Cuneiform is regarded as one of the earliest systems of human writing and encoding of information. It was common during the Sumerian period in the fourth millennium BC. A blunt reed was used to stamp words and letters into small pieces of clay. The content of the tablets ranged from financial data to epic adventures. The neural network learned cuneiform by abstracting pixel-data of the items in the tablet-corpus. After that it was able to generate new tablets on its own. During this process the AI maps out the collective unconscious of the Sumerian civilization 4000 years ago, abstracts it mathematically and uses this to continue the production of entirely new cuneiform writings. If the texts make sense at all still needs to be verified by cuneiform experts. The generative approach in combination with augmented reality represents a playful form of AI-driven memory: a hyperspace archeology that breathes new life into historic artifacts and puts them into a contemporary context.
year: 2017, 2019
work: laser etching, processing, 3D rendering, AR, machine learning (DCGAN)
AR Quick Look is used for the image above - if you are on iOS, click on the cube symbol in the top-right corner of the image and have your own cuneiform cube in AR.