Portal de Periódicos da CAPES

Can the Computer Speak?

2023; Duke University Press; Volume: 95; Issue: 2 Linguagem: Inglês

10.1215/00029831-10575260

1527-2117

Christopher Grobe,

History of Computing Technologies

Macs do it, phones do it—even educated drones do it. And by it, of course, I mean “respond to our words with a synthesized speech of their own.” How did it come to pass that the Internet of Things was suffused with this ability to recognize, parse, and synthesize human speech? How did people come to think of such recognition as “listening,” such parsing as “understanding,” and such synthesis as “speaking”? How did technology and culture combine to make this development thinkable—even, apparently, desirable?Two recent scholarly books attempt to answer such questions historically. Simone Natale’s Deceitful Media focuses on how we came to treat machines as social agents capable of conversation. Specifically, it shows how the dream of artificial intelligence (AI) blurred the lines between computer science and adjacent social scientific fields. Liz W. Faber’s Computer’s Voice focuses on how we came not just to anthropomorphize speaking machines but to gender them, too. Specifically, it draws on the history of American sci-fi TV and film to show how AI was gendered—and gendered differently—in successive eras of modern computing.These books are quite different from each other, methodologically. Deceitful Media comes out of media studies, and its project is historiographic: it rewrites the history of AI to include a better account of how technologists thought humans perceived intelligence. The Computer’s Voice, on the other hand, comes out of film studies, and its project is driven by feminist psychoanalysis: it identifies gendered tropes in art that features talking machines, and then it unsettles those tropes through close study. What these books share, however, is the idea that people’s fantasies about machines have real-world force in shaping how technology gets built and deployed.▪ ▪ ▪Deceitful Media’s innovation is to tell a joint history of two fields that are often treated as separate concerns: conversational AI and human-computer interaction (HCI). Examining key moments in the history of both fields, Natale argues that the quest for machine intelligence has never been separable from the attempt to get humans to “treat things as social agents” (4). Naturally, Natale’s story of this phenomenon begins with Alan Turing’s imitation game, which “placed humans rather than machines at the very center of the AI question” (20)—and it culminates in a withering account of the Loebner Prize competition for chatbots, which has reduced Turing’s philosophical game to an annual search for ONE NEAT TRICK!!! to fool some judges. Along the way, Natale offers several case studies—some famous, some obscure—of how “the race for AI was being run not only in the realm of technology but also in the realm of imagination” (38). He explores how something as simple as early time-sharing systems, which allowed multiple users to interact separately and simultaneously with one computer, achieved a feel of responsiveness by “employ[ing] the human nerveware as well as the computer hardware” (41, quoting early computer scientist Herbert Simon)—and he uses such examples to demonstrate a more general point: insights gleaned from “the new human sciences” of the early twentieth century were essential to midcentury computer science (12). In short, while hardware and software get all the credit, knowledge of “nerveware” was essential in the pursuit of AI.As Natale continues this history up to the present, he demonstrates a growing concern with “social interfaces”—famous, like Siri, and obscure, like Microsoft Bob—and their ability (or failure) to sustain “the mindless nature of imaginative social relationships between humans and machines” (83). (Mindlessness is a keyword from HCI, and encouraging people to interact mindlessly with machines is a goal of much research in that field.) The tactics for encouraging such mindlessness are examples of what Natale calls “banal deception” (7). These tactics are deceptive, he suggests, in their attempts to exploit the quirks of human perception and cognition, but they are banal in the sense that people “embrace deception so that they can better incorporate AI into their lives” (7). In other words, we are “deceived” not because we are fools but because such suspension of disbelief brings us benefits. By consenting to treat our iPhones, for example, as quasi-human conversation partners, we stand to gain Siri’s smooth integration into our lives.This is a persuasive account of how conversational software has developed since the 1950s, but Natale has greater ambitions for his coinage, banal deception. He would like to persuade us that all modern media are “deceptive” in how they exploit quirks and blind spots of “the human sensorium and intellect” (12). Deception is not a time-limited phase in the history of each technology, lasting only as long as people’s initial ignorance, but is instead “an irremediable characteristic of [modern] media technologies” (12). For example, audio compression algorithms “deceive” our ears, Natale argues, by stripping out sounds we won’t miss very much—just as the frame rates of projected film leave out images we’d scarcely be able to see.The stakes of calling these phenomena deception—banal or otherwise—are unclear in Deceitful Media. Natale might argue on one page that people knowingly “embrace deception” (29), while on the very next page he might assert that people “[fall] into a state of narcosis that makes them unable to understand how media are changing them” (30). This latter, Black Mirror-ish tone about technology is common across academic media studies, but it distracts from Natale’s true argument: “AI is a relational phenomenon, something that emerges also and especially in the interaction between humans and machines” (31). This relational approach, the bread and butter of HCI, was baked into AI from the start, Natale argues. And it persists in the “personality teams” behind today’s voice-user interfaces, like Alexa or Siri. These groups of social scientists (and sometimes artists) play a key role in determining how conversational AIs are constructed.▪ ▪ ▪If Natale provides a disciplinary prehistory of conversational AI, Faber provides its imaginative prehistory. Specifically, they track the emergence and transformation in sci-fi media of what they call the “acousmatic computer” (15). (An acousmêtre, in film theory, is a figure whose voice we hear but whose body we do not see.) For Faber, the cultural history of this trope begins with the shipboard computer on Star Trek (1966) and culminates with the release of Siri on iPhones in 2011.Within this forty-five-year time frame, Faber gathers an eccentric mix of sci-fi films and TV programs, grouped mostly by decade. These groupings are then offered as proof of how talking computers have been variously gendered over the years. So, for instance, chapters 2 and 3 both center on the same decade (the 1970s), but they explore different imaginative tendencies: whereas chapter 2 focuses on the hyperfeminine caricature of computers offered in such films as Dark Star (1974) and TV series such as Quark (1977–78), chapter 3 draws our attention to the hypermasculine, militarized stance of computers in films like Colossus (1970), THX 1138 (1971), Rollerball (1975), and Demon Seed (1977). After isolating this kind of gendered trope, each of Faber’s chapters studies examples of that trope through the lens of feminist psychoanalytic theory.Faber justifies this psychoanalytic approach as a necessary alternative to the cognitive bias in tech-industrial thinking about AI. Such counterballast is sorely needed, but this book will not teach you how to weigh one against the other—or how to revise technological thinking in light of arguments like these. A final chapter does analyze Siri in comparison with Samantha (the intelligent operating system in Spike Jonze’s Her [2014]), but it does so by focusing on the discourse around Siri (e.g., Apple’s choice not to use gendered pronouns in referring to it). In other words, it sidesteps those realms where cognitive approaches are applied to technology. It studies technology by first reducing it to narrative.This leaves us with the need for more books on this subject, for instance, books that make good on the historical narrative implied by Faber’s title. How exactly did we get “From Star Trek to Siri”—and what, if anything, stayed with us throughout this half-century journey from fiction to fact? How, if at all, has conversational AI been shaped or limited by the narratives developed by sci-fi auteurs? In the voices of today’s “smart” machines, can we hear any echoes of screen actors’ performances? Have designers of computer-synthesized voices learned anything from the professionals who capture, create, edit, and mix sci-fi sound? Do the dialogue styles of these films and TV series bear any resemblance to the conversational flows created by voice-user interface designers? In other words, what truth is there (if any) to technologists’ frequent claims that they have taken Star Trek’s talking computer, or HAL-9000, or any other sci-fi creation as their model? To answer questions like these, one needs a double expertise: not only Faber’s commitment to studying art but also Natale’s understanding of technology and the thinking that lies behind it. Together, these two approaches might help us confront AI as the strange hybrid thing that it is: both a technical problem and an artful dream.