Signs of the Time: Making AI Legible

Abstract

Humanity's fascination with artificial life is long-lived, appearing in ancient mythology (e.g.Galatea, Talos) and more modern fiction alike (e.g.The Creature in Shelley's Frankenstein, HAL9000 in Arthur C. Clarke's 2001).Arguably, however, it was Alan Turing's seminal research question "Can machines think?" (Turing, 1950) that gave rise to the field and the technologies that we now call AI.In the 70 years since Turing posed the question, AI has experienced cycles of inflated expectation and troughs of disillusionment.While the ethical and technical complexities of Artificial General Intelligence (AGI) remain as ontologically challenging as ever, the AI field has developed an array of powerful computing techniques including Neural Networks, Expert Systems and Machine Learning.Facilitated by a growing abundance of data, cheap computing power, and advanced data science, these techniquesin particular Machine Learning-have become come widespread.Whilst these AIs excel at pattern recognition and prediction tasks, we have not created any 'thinking' machines, however, there are plenty of reasons why we should put time and effort into thinking about them.