Working Groups/Narrative Future of AI

From MozillaWiki
Jump to: navigation, search

Narrative Future of AI
Narrative Future of AI was launched by Algowritten in the pilot cohort of Mozilla’s building trustworthy AI working group. This project aims to address problematic cultural biases of machine learning through the creation of a series of media works that challenge and explore bias in new algorithmic technologies, such as GPT-3. The project is led by researchers at The School of Digital Arts (opening in September, 2021), which is the founding member of an emerging consortium of North of England universities focusing on AI in digital storytelling, comprising The Centre for Immersive Technology (University of Leeds), XR Stories (University of York) and the Institute of Art and Technology (LJMU).

Problem
The future of digital storytelling will involve the increasing use of algorithmic tools, both to develop new forms of narrative and to find efficiencies in creative production. However, unsupervised algorithms trained on massive amounts of web-based text come with issues of bias most harmfully pertaining to gender, race, and class. These issues are compounded by problems of transparency and explicability related to large complex algorithms.

Solution
As creative applications for AI emerge and have authorial identities assigned to them, it must be clear where their authorial voice originates and who it deems worthy of inclusion in storytelling. This includes information on training data related to the authors, the stakeholders of the AI authors, as well as genre conventions and storytelling techniques that may continue to be damaging to the representation of marginalised groups. This project seeks to review typical biases that occur from writing with the GPT-3 API. The outcome will be a series of science fiction stories and feedback from working group members on their observations of bias and problematic AI behaviours. This analysis will form the basis of our first set recommendations for creative writing with advanced machine learning tools.

Contributors
Algowritten, which explores bias in algorithmic writing, is maintained and edited by Dr David Jackson and Marsha Courneya who are academics at the School of Digital Arts, Manchester Metropolitan University, UK.

Resources
Project site Short story collection