Project

'@artbhot' - Towards a Creative Partner on Social Media

@artbhot is a Twitter bot that leverages stable diffusion technology to translate user tweets into images and animations.  Access to text-to-image networks has generally been restricted to Google Colab notebooks and interactions with the command line , until recently when implementations such as Midjourney and DALLE.2 were made available with access to more user friendly features.  

Embedding this generative deep learning capability amongst social media however provides a unique opportunity to gain insight into how people interact with image and video synthesis algorithms in a social environment. At a time of unprecedented growth and usage of these generative models, and given the pervasiveness of the generated media already online, it is salient to gauge the role this technology could play moving forward. 

The trajectory of this project is set around using an instance of this technology on Twitter to better understand what a meaningful and useful interaction might look like between a user and the generative model - within the scope of treating the system as a ‘creative partner’.   

Current implementations of standard diffusion are set up in such a way that user interactions are akin to an image retrieval service, similar to a Google image search for images that don't exist yet. It seems that a computational agent that has potential as a ‘creative partner’ will need to do more than simply provide an image generation and retrieval service.     

@artbhot currently autonomously engineers prompts from user tweet text and can also use trending topics on Twitter to autonomously tweet media generated from these subjects. 

This project is motivated by an interest in working with generative deep learning to aid users in exploring their own creativity in productive ways  with the philosophy that AI can help guide us to arrive at better and more creative decisions, and support individuals in creative projects and work flows, rather than solving these problems for us.