Conversational Computing: Speech Synthesis from Assistive Technology to Artificial Intelligence, 1930–1980

Abstract

This dissertation shows how a strange alliance between art and disability led to the creation of computerized speech. Most scholars have viewed the history of representational technologies as the replacement of the hand by mechanical forms of reproduction; they have similarly defined modernity as a march towards disembodied data. I offer a dramatically different perspective by showing that conversational computing—what Ed Finn has called “the holy grail of artificial intelligence”—grew from bodily knowledge and tools borrowed from the artist’s studio. Put simply, the emergence of user-friendly computing and natural language processing did not depend entirely on information theory or disembodied deliberation; instead, an unlikely combination of hand-painting and disabled bodies gave rise to the notion of human-computer symbiosis. To be clear, it was not only the labor of scientists and artists that facilitated this transition. My work demonstrates how the blind community accelerated the technological translation of graphic text into other sensory modalities. Starting in the 1940s, the use of artificial speech as a vocal prosthetic catalyzed digital computers’ transformation from glorified calculators into the talking chatbots we use today. In this sense, blind scientists and programmers were not the beneficiaries of charity. Instead, they were active participants whose labor was later co-opted for non-disabled consumers’ virtual assistants. As the voices of these AI agents and deep fakes amplify, so too does the value of attentively listening to the audible past of synthetic media. To be sure, the theoretical and historical story of reproductive technologies—from the photograph to the phonograph—has been well-told. Extending this narrative, my dissertation articulates the unheard history of mechanical synthesis and conversational computers.