The automation of creative labor - Will (non-core )

 

I was recently referred to the work of Holly Herndon and Matt Dryhurst that connects well to our question of labour for this week and our general questions regarding the automation of creative labour throughout the semester. 

Herndon and Dryhurst have created Holly+ - a vocal net that claims to be able to reproduce Herndon's voice from any audio input. They have produced a version of Jolene that is entirely generated by Holly+ (above) and you are able to upload audio files to try yourself: https://holly.plus/ (They must be below 23.84MB and less than 5 minutes long). The creations that I trialed were distorted yet familiar and at times uncanny.

Holly+ poses interesting questions for the future of creative labour. Firstly, critics of creative automation often argue that AI are unable to capture the creativity/excitement/passion of human creativity. For instance, in the video that Rohan posted, the NYT recipe writers to the GPT-3 generated thanksgiving recipes. (For the record, the generated recipes sounded nothing if not creative to me). Does Holly+ lose the 'essence' or 'aura' in Benjamin's terms of human performance? Second, if artistic deepfakes are here to stay (such as DALL-E and Midjourney appropriating artistic styles through scraped content), is it better for artists to 'lean in' and license their own creativity, style and content as such? The use and commercial distribution of Holly+ is governed by a Decentralised Autonomous Organization, but artists could choose any licensing agreement that suits them.