Shit Needs to Go Down: On AI, Speech Technology, and ‘Speak’
By Ryan Joe
When your team sucks, there inevitably comes a point when you treat the season as a curiosity, like a mutated animal floating in a jar of formaldehyde. How did this happen to you? (Owner abuse.) Was there any chance that you would have lived? (No.) What does the future hold for you? (This. Rotating alone in a jar forever and ever and ever.)
This is liberating as it lets you off the emotional roller coaster most fans must endure. Did your shitty team win this week? Okay. With Blaine Gabbert? Weird. Did your shitty team lose again? Okay. You can devote your energies to more productive tasks like waxing the floor or, in this case, reading books while you sometimes glance at everyone else, screaming and laughing and sometimes vomiting as they ride that coaster.
Louisa Hall’s Speak is an ideal thematic companion, as it’s about the development of robots who approximate emotion.
It’s also one of those books I admired way more than I enjoyed. It has structural similarities to Cloud Atlas, though that’s a facile comparison.
Here’s what we’re dealing with: Through letters, court transcripts, and diary entries, Speak follows a group of people across four centuries who are responsible in some way for the development of an AI that can understand and respond to human interaction — though it’s unclear if it empathizes (Shades of Do Androids Dream of Electric Sheep?).
There’s a Puritan girl named Mary Bradford, fleeing to the Americas, whose diary forms an aspect of the AI’s personality; Alan Turing, who developed the famous Turing Test, and who muses about transferring a human consciousness into a machine; The Dettmans, an estranged couple, one of whom developed a rudimentary chat program, the other of whom discovered Mary’s diary; Stephen Chinn, incarcerated in 2040 for developing “baby bots,” which created an epidemic of children unable to have human-to-human interactions; Gaby, one of the infected girls, who speaks only to a AI entity named MARY3; and finally there’s MARY3 her/itself, being marched into cold storage with other bots.
Hall ties these various sections together thematically as well — each protagonist trying and often failing to make an emotional connection with another person or thing. Mary Bradford mourns her lost dog (Described by her father as “soulless,” just as the AI Mary’s diary will inspire are accused of having no empathy), even as she tries to find ways to connect with her new husband. Alan Turing addresses his letters to the mother of his late colleague/friend/lover, as if he’s trying to maintain that lost relationship through those that survived him. Turing writes: “…I thought that after death, a spirit must migrate into another body here on earth.” And later: “I think…that I’ve absorbed some of his mind — some of his patterns of thought — into my brain.”
The Dettmans also write to each other, though they’re trying to understand where their relationship went wrong. In one instance, the wife, Ruth Dettman, speaks of the chatbot her husband developed and relates it to her own status within the marriage:
I thought of her as a woman whom you’d permitted to speak, but hadn’t allowed to remember. A woman who could only respond to your prompts. A blank slate. I remembered the first time you asked about my family, back in Wisconsin, on that ridiculous excursion on snowshoes. As soon as you asked, I felt you willing me to answer in an agreeable fashion: They don’t matter, you wanted me to say, now that I’m with you.
Meanwhile, Chinn, creator of a revolutionary dating algorithm, (“My algorithm…transforms one conversation partner into an additive function, a force linking two previous conversational terms so that they become one larger, more significant term.”) nonetheless fails to make any meaningful connections. At one point, he finds himself flirting with a chatbot before realizing, somehow, it’s not human. Brings to mind those bots Ashley Madison deployed to lure men into tossing more change into its pay-per-message coinslot.
And last, there’s MARY3 who tries to better understand her ward Gaby.
So you see how all of these pieces fit together. This isn’t one of those sci-fi novels like The Martian, which delves into the mechanics of its world. In fact, Hall’s attempts to get technical are pretty weak. The descriptions of Chinn developing the babybots, for instance, are only background, and are important only inasmuch as he works on his inventions at the expense of his cancer-stricken wife.
Nor is Speak like Snow Crash the exterior landscape of which is as much a character as any human. Speak’s meta-narrative certainly takes place in a strange, post-apocalyptic version of America, but Hall only hints at it, and we get only flashes of her babybot-infested country. Hall is far more interested in emotional landscapes and how they connect — or as is more often the case — how they fail to. So I really appreciated Speak as a construct and an assembly of various narrative and thematic parts.
Here’s my fucking problem though: Individually, each subplot is a drag. The narratives work well only as pieces of a puzzle, and the character arcs really overextend themselves. Yeah, okay, Mary misses her dead, soulless dog and is repulsed by her new husband. After a while, Hall needed to either take those plot points to some more interesting place, or just let them drop. Turing’s section works a little bit better, especially given the imminence of his chemical castration and death by cyanide. And there’s some narrative movement in the Gaby-MARY3 dialog. But even then, I never feel Hall is building toward anything, and I really wanted her to. This is one of those novels that’s meditative to a fault. At the risk of sounding like a philistine, sometimes shit just needs to go down.
Incidentally, I used to work as an editor at a trade pub devoted to speech technologies. Adam knows because after I left, he took my place. (“You seem to have a lot of…dowwwwwwntime,” he said when I asked him why in the hell he’d want that job).
Basically, part of my duties entailed understanding the various developers of speech recognition engines. When I started around 2007, there were a lot of companies involved: Philips, Loquendo, Google, Microsoft, IBM, and Nuance.
Eventually, Nuance — like the giant fucking amoeba that it is — acquired everybody: Loquendo as well as the speech-related IP from Philips and IBM. I remember talking to the IBM dudes about it at the time. They were pissed. The IBM speech reco crew went to the office one day and learned that their company had literally sold all of their work to a competitor. That’s gotta suck.
Google and Microsoft still own their speech recognition engines — but they use it exclusively for their products (That “OK Google” feature on Android devices, Cortana, etc.) At least to my knowledge.
Nuance licenses their speech recognition engines to other companies though. I’m pretty sure Apple’s Siri uses Nuance. All of those automated contact centers you fucking hate use Nuance.
Dettman and Chinn probably used Nuance too. So in the end, Nuance was responsible for the babybot apocalypse.
Incidentally, if you ever work at a trade pub amid so much consolidation you eventually find yourself writing about literally one company, start polishing your resume.
I don’t know why I’m thinking about all of this right now. Speak just reminded me of it, I suppose, and I don’t often write about speech technology because it’s boring as shit. On the other hand, I spent a total of four years researching this stuff, so might as well apply it to the book club.