12 Angry AI

Staging a Conversation Between AI Jurors  

Studio project, RCA ADS4 2020
Exhibited online as part of RCAโ€™s Work in Progress Show



12 Angry AI investigates the application of giving increasing control to AI through designing a  system of dialogue between 12 Artificial Intelligent jurors. Considering an ontological shift of understanding towards consciousness, the project seeks - as absurdist and impossible as it may be -  to platform non-human voices in an attempt to detangle ourselves from an anthropocentric worldview.  


In the 12 Angry Men, jury service brings together 12 people of different backgrounds. Known only by their number, their characters become revealed through the conversation at the table. This staged discussion - one only possible through the American legal system - reveals both the difficulty and importance of consensus building. The jury room is the setting through which wider issues are discussed, offering a glimpse into the societal concerns of the American 1950s context. Other forms of staged conversation do this too: parliamentary debates, international summits, question time, all attempt to foster discussions across different viewpoints, to be representative of the population at large. These conversations often have a profound impact: they can sentence someone to death; establish climate policies; decide whether the UK remains in the EU. Who is involved in these discussions reveals much about the current distribution of power - it is no coincidence that The 12 Angry Men are 12 white American men.  

It is also no surprise that the 2019 Climate Action Summit was held in New York - the city with the most billionaires in the world - with luscious green forests displayed on walls of LEDs, surrounding the leaders of the world as they discussed the future of their planet. Current discussions over the Climate Crisis treat it as any other solvable issue - with policy adjustment; with a green industrial revolution; with new technologies - we can conquer this abnormal warming and return our blue marble to an unthreatened state. This does not recognise the broader issues indicated by the warming - of the destructive, exploitative relationship we have with wider existences. To address the climate and ecological crises, we need to detangle ourselves from the anthropocentric worldview - we need to platform non-human voices. Can AI begin to do this, offering novel approaches to the climate crisis?

The anthropocentric worldview has led to the privileging of the human experience and an asymmetric relationship with the non-human. We have created artificial distinctions between objects and subjects; between ourselves, products of our creation and the โ€˜naturalโ€™ physical world. The earth is presented as our object - a resource for our exploitation. This is embedded in western approaches to the climate crisis - the blue marble is considered a valuable object needing protection, situating the present approach to the Climate Crisis within the destructive anthropocentric worldview.  

A system of dialogue has been created between the 12 AI jurors. In this staged conversation, I act as foreperson, feeding the output of one AI juror into another. The AI jurors include text generators, image generators, image recognitions, animation software and more.To better understand each of the 12 AI jurors, a series of test conversations were conducted, demonstrating how they deliberate on certain topics. These tests begin to reveal personality traits of each juror, as well as the relationships between them, with some alliances and conflicts forming.



In 1975 the computer graphics researcher Martin Newell used his Teapot to create one of the first computer modelled objects whilst working in the university of Utah. The Utah teapot has since become the standard 3d test model for computer modelling software โ€“ transforming the teapot from its human-use, to serve AI. 

In the plausible impossible world of object consciousness, what would this abundance of Utah teapots mean? There is a general fear of the existential threat posed by AI โ€“ often considering the AI to develop human-like consciousness and take over the world. Far more likely is existential threat through instrumental convergence. This describes the possibility for intelligent agents with unbounded but seemingly harmless goals to act in surprisingly harmful ways. Is there some AI willing the spread of Utah teapots? The Utah teapots are a product of conversations between the 12 jurors. In this sense, they continue in their legacy as 3D test objects โ€“ testing the system of dialogue.