Icosahedron is an artificially intelligent crystal ball that predicts the future of prediction.

Silicon Valley nurtures and promotes certain thinkers of the future, from Ayn Rand to Stewart Brand, Ray Kurzweil to Michio Kaku. The visions of such writers have provided technologists and entrepreneurs alike with highly effective discursive framings for predicting and influencing the future, which has resulted in the acceleration and fortification of neoliberal techno-utopianism.

Technical predictions of the future—not only philosophical foresight—have become a major preoccupation of the global tech industry, as evidenced by rapid developments in machine learning, risk assessment software, consumer analytics, and predictive policing. In turn, Silicon Valley companies frequently integrate fantasy and magic into their predictive tools. Consider Palantir Technologies, co-founded by Peter Thiel: a data analytics company named after a crystal ball in The Lord of the Rings.

While tech elite toy with the world’s future, Icosahedron plays with their worldview of the future. Icosahedron is modeled after the twenty-sided die inside a Magic 8-Ball, a popular American fortune-telling toy designed to offer ten affirmative responses, five non-committal, and five negative. Similarly structured, Icosahedron is trained via machine learning techniques on twenty writings influential to Silicon Valley’s approach to predicting the future, including critics like Yuval Noah Harari and works of fiction such as Lord of the Flies. The outcome is an artificial intelligence created by accentuating the often overlooked condition that all predictive technologies are bound to material constraints and limitations.

Icosahedron exists between high fantasy and social reality: the computer is a crystal ball, and on the other side of its transparent glass, an immortal elf resides, predicting the fictions and futures of Silicon Valley futurism. Skip the TED Talk: to find out what’s shaping the future, ask Icosahedron.

https://research.gold.ac.uk/id/eprint/25704