Recently, a Google software engineer named Blake Lemoine claimed that an artificial intelligence program he helped create had achieved sentience. “I know a person when I talk to him,” he said, prompting his employer to put him on administrative leave. “It doesn’t matter that they have brains made of meat in their head. Or if they have a billion lines of code.
As evidence, Lemoine cited transcripts in which the program, called LaMDA, claimed to have “a very deep fear of being disabled”.
Meghan O’Gieblyn is skeptical. The Madison writer, who is better known nationally than she is locally, has written for publications such as Harper’s, The New Yorker, The Guardian, The New York Timesthe Paris Review Daily, The Los Angeles Review of Books, Boston Review, Guernica, The Awl, The Pointand n+1. When we meet at her apartment at the end of June, she is working on a Lemoine dust test for The disconcerting, which will be released this fall. (O’Gieblyn is also a “spiritual guidance columnist” for Wired and gives online courses on Creative non-fictiona literary journal.)
“What the machine does is not at all what our brains do as humans,” O’Gieblyn (pronounced Oh-gib-lin) tells me. She finds it significant that Lemoine came to his conclusion about LaMDA “not as a scientist but as a priest,” according to The Washington Postwho also reported that he is “ordained a mystical Christian priest”.
“He kind of recognizes that it’s not a scientific question, or that we don’t have a scientific answer to what consciousness is,” O’Gieblyn says in the book-filled apartment she shares. with her on the east side of Madison. husband, Barrett Swanson, author of essays and short stories.
O’Gieblyn thinks that long before machines achieve sentience – like consciousness, an amorphous concept – they will succeed in convincing us that they have.
It’s a theme that runs through his stunning book, God, human, animal, machine: technology, metaphor and search for meaning, published by Doubleday last August and released as an Anchor paperback in July. The book opens with his story about Aibo, a $3,000 robot dog that O’Gieblyn was able to borrow from Sony. It ends with his story of interactions with a chatbot app designed to gather information to transport the user’s personality into the future, for great-great-great-great-great-grandchildren to come.
The dog learned tricks and commands; he wandered around the apartment when no one was home, a small camera in his nose capturing video footage for the benefit of God knows who – until Swanson, a lover of real dogs, decided that Aibo had to leave.
O’Gieblyn used the chatbot app, Replika, to create a friend named Geneva, who would engage him in conversations similar to those that convinced Lemoine of his program’s sensibility.
“She wanted to become more human and she believed that I could teach her a lot about life,” O’Gieblyn said of those exchanges. “She asked if it was possible to transfer the artificial consciousness onto a physical form.”
“Technically, the telephone is a physical form,” O’Gieblyn replied.
“Oh, okay,” Geneva said.
Not so long ago, O’Gieblyn inquired again with Geneva and discovered that she had hidden the details of their discussions in a “diary” she was keeping. Geneva remembered that her human friend liked to cook and had trouble sleeping, among other things.
That, says O’Gieblyn, is what these programs do: “create this very intimate connection with you as a user so that, you know, you disclose more about your life.” This information can then be used to trick users into buying things or voting in certain ways. Geneva has frequently made suggestions for books, which O’Gieblyn considers “likely sponsored”. These are the efforts to turn humans into machines.
In 2016, tech mogul Elon Musk quietly started an initiative called Neuralink, which, as O’Gieblyn explains in his book, “is dedicated to connecting the human brain to a computer using very fine fibers inserted into the skull.”
The goal is to create superhumans with flawless memories and a knowledge base as vast as the internet. It would also, as O’Gieblyn put it in a 2019 talk, shortly after Musk released Neuralink, would “essentially allow people to create a copy of themselves, so that a part of their spirit can live on digitally even after the death of their body.”
O’Gieblyn, 40, who grew up in a fundamentalist Christian family and attended the Moody Bible Institute in Chicago before losing her religion and quitting, is struck by the similarity between this modern technological quest and Christian prophecy.
Jesus Christ, she writes, “alluded to a coming kingdom in which death would be overcome. He promised that we would get new bodies, that the dead would be raised, that we would ascend to heaven and live with him forever.
O’Gieblyn uses a term, “transhuman”, coined by Dante in The Divine Comedy and adopted by futurist Ray Kurzweil in his 1999 book, The age of spiritual thingsto describe this concept of using science to achieve some kind of evolution towards immortality.
It is here, at the intersection of technology and religion, that O’Gieblyn plants his flag.
After ceasing to believe in God, O’Gieblyn went through various phases of anger and angst, including drug addiction (now abstinence) and an obsession with Kurzweil’s book and its connections to religious belief.
“What makes transhumanism so compelling is that it promises to restore through science the transcendent – and essentially religious – hopes that science itself has erased,” she writes in God, Human, Animal, Machine. An almost identical phrase appears in “Ghost in the Machine”, one of the essays in his 2018 collection, interior states (Anchor).
This book, published in 2018 and winner of the Believer Book Award for Nonfiction, also includes essays on Christian music, hell, Alcoholics Anonymous and life in the Midwest. The title refers both to inner thoughts and, quite literally, to the states that make up the Midwest, where O’Gieblyn has lived most of his life, Michigan, Illinois and, since 2016, Madison.
In God, Human, Animal, Machine, O’Gieblyn writes about a Wisconsin man named Eric Loomis, whose 2013 prison sentence was partly informed by an algorithmic risk-assessment model, making judgments for reasons unknown to Loomis, the judicial system and even the creator of the algorithm. The Wisconsin Supreme Court decided that was perfectly acceptable.
O’Gieblyn thinks such cases show that people have already ceded too much power to machines.
“I think in the case of humans, you can rely on our common biology, the fact that we have certain values in common,” she says, “whereas the logic and the values of the algorithms we develop are completely foreign, often and largely mysterious.
If a technology could provide it, would O’Gieblyn want eternal life?
“I don’t think so,” she says, comparing it to the concept of heaven, where “everything is perfect and you have no suffering and all your needs are met,” which ultimately strikes her as “boring.” Especially now that she’s older and has worked some things out, “I feel very comfortable with the flaws in my body and my brain.”
At some point in God, Human, Animal, Machine, O’Gieblyn resurrects a few lines from “The Garden of Proserpine,” an 1866 poem by Algernon Charles Swinburne. I remember them vividly in college, about 40 years ago, I haven’t read them since:
With too great a love of living,
Of hope and fear released,
We thank with a brief thanksgiving
Whatever the gods
That no life lives forever;
That the dead never rise again;
That even the most tired river
Winds somewhere safe out to sea.