In computer science, machine learning algorithms are typically framed as a set of techniques for building accurate models of data, and research has traditionally focused on building better models more quickly, and for more types of modeling problems. Popular narratives around machine learning and AI center around visions of progress and disruption driven by automation, around a Terminator-style singularity, or increasingly around a dystopian future of surveillance and manipulation of the public.

In this talk, I’ll discuss how these perspectives obscure many of the opportunities for human benefit that can be found by applying machine learning to vastly different types of problems and by making it usable by more diverse people. I’ll describe my work exploring questions such as: How can machine learning support human creativity and embodied interaction? How can machine learning enable new groups of people to create and customise interactive technologies? How can machine learning fundamentally change people's relationships with technology? When and how should our computational framing of machine learning change to reflect different contexts of use or user goals? How can we build usable machine learning tools for people such as professional musicians and artists, kids, or software developers? And how might we change machine learning education to reflect machine learning's growing importance across society?