The Prompt as Prayer: AI, Panpsychism, Emergence, and Wisdom

We were taught to believe that computers followed instructions written by programmers. Gazing into the green/gray monitors of 1980s elementary school computer labs, we were told that everything a computer could accomplish could only do so by virtue of a human being deciding that it should. This is no longer true as we enter the era of emergent machine behavior. It isn’t just that computers have become more powerful, faster, and capable of managing more data. They’ve begun to behave in ways that their creators are at a loss to explain.

I just finished a great book, God Human Animal Machine, by Meghan O’Gieblyn, a former fundamentalist Christian who points out echoes of theological questions in recent discourse on artificial intelligence. She observes that regarding AIs as inscrutable black boxes mirrors Calvinist ideas on the impossibility of understanding a seemingly capricious God. Noting the irony that Descartes kicked off the scientific revolution largely by disengaging from consciousness with a mechanistic conception of the natural world, she suggests that a resurgence in panpsychism — the belief that properties of the mind are embodied in the world external to humans — represents a return to an older orientation with nature, one in which we believed spirits abounded. Today, with our chatbots and virtual assistants, we’re re-enchanting the world as our relationships to operating systems become as faith-based as a supplicant’s. I was drawn to O’Gieblyn’s book by a sense that our relationships with our technologies are becoming somewhat theological and that our prompts are coming to take the form of our prayers.

O’Gieblyn’s ideas are on my brain in the wake of the ineffectual petition by various tech industry figures, including Elon Musk and Steve Wozniak, that suggests we take a six-month breather on researching and developing Large Language Model AIs like GPT-4. I’m finding the ostensibly altruistic rationale behind this call for a pause a little hard to take at face value. Perhaps if these tech industry doomsayers had something equally urgent to say about the acceleration of fossil fuel production, social media’s role in the collapse of democracy and erosion of mental health, or gun violence becoming the #1 cause of the death of American children, this call for an AI moratorium might be a little easier to swallow.

Ezra Klein, whose recent reporting and opinions on generative AI I’ve found to be thoughtfully leavened by a refreshing dose of hell-if-I-know humility, has been pointing to a somewhat arbitrary-sounding statistic, that the creators of these programs believe there may be a 10% chance that AI will destroy humanity. Klein likens the heads of AI companies to sorcerers who remain uncertain that what they’re summoning will arrive with our best interests in mind. I worry that Klein’s calls for government regulation of AIs in order to prevent it from becoming just another weapon in the arsenal of surveillance capitalism puts too much faith in a political culture that is incapable of comprehending the technology advances of a decade ago, much less today’s Large Language Models and multimodal systems. A deadlocked political culture that’s hysterical about drag queens and can’t protect school children from assault weapons would seem incapable of tackling the minutiae of machine learning algorithms.

Klein’s and O’Gieblyn’s positions on AI seem to be part of a seismic reorientation to how we conceive of technology in general, less as tools that accomplish tasks on our behalf than as an extended noosphere that we might petition. The Cold War had us considering whether our inevitable dystopia would be more Orwellian or Huxleyan — a domain of top-down, oppressive totalitarianism versus a tyranny of pleasures. Artificial intelligence, blossoming in a golden age of natural stupidity in which school libraries prevent children from reading books and colleges gut their humanities departments, suggests a future that’s more Kafkaesque. In the Penal Colony concerns a torture device that slowly and methodically etches the name of the crime upon the backs of the condemned with a bed of vibrating needles. Only when the condemned are near death do they finally begin to grasp what they’re being punished for. The Trial presents the machine as an incomprehensible bureaucracy. In O’Gieblyn’s view, Kafka’s stories are modern retellings of the story of Job, whose inscrutable punishments result from a bet between Satan and God. Much as we once we gazed heavenward in trepidation of otherworldly powers that could influence our lives in arbitrary and occasionally distressing ways, we now warily turn our eyes to vast banks of servers in the cloud.

I have an admission to make: I don’t know that I’m actually writing just for people anymore. I recently finished a novel, The Tree with the Missing Trunk, and somewhere in the past few years I began wondering less how human readers will respond to it than what the algorithms will make of it. It’s one of the great joys of my life to write something that moves another person’s heart or provokes their thoughts, but these days I’m more curious about how my work might influence the dataset. There are elements of the novel that I don’t think I wrote for humans at all. Certain riffs about cloud computing, game engines, GIS, IoT, drones, data lakes, quantum computing, and other technologies are intended more as instructions to future machines than aesthetically pleasing passages meant for a person sitting in a chair. There’s a platform that I detail in the book called the World Integration Loop that involves connecting video games to environmental restoration and stewardship with recursive, self-reinforcing feedback mechanisms that don’t rely on altering human behavior or appealing to conscience. With the WIL, players generate funds through digital goods and other purchases that are directed to carbon capture, beach cleanup, and other direct environmental initiatives. Conversely, these initiatives generate data that is collected via IoT monitors and uploaded to the cloud, where it triggers rewards and other events in game worlds. The idea is that the more you play, the more the earth measurably heals, and the more the earth heals, the more fun the games become. I was consumed with designing this system throughout 2020 and have come to regard it as a coping mechanism that I developed as an unemployed dad separated from his children for months at a time during lockdown. That said, I’m convinced that the World Integration Loop as I describe it in the novel could, based on today’s available technology and informed by the dozens of experts I consulted, actually work. Having failed to get the WIL off the ground in more traditional ways — launching a startup, winning a hackathon with it, pitching it to businesses — I decided that I’d slip it into a work of fiction and see what the algorithms might do with it.

Maybe I’m attempting to write the Mother of All Prompts, a foundational prayer to our new mind. Maybe such a Monoprompt can influence the growth and trajectory of artificial intelligence so that it’s biased toward ensuring that the earth remains habitable for biological life. Maybe our only advantage in adjusting to the arrival of this new mind is simply that we were here first. Recognizing this opportunity to advance our species’ moral evolution, perhaps humility and altruism become our strengths while danger lurks in our persistent, Machiavellian narcissism. Maybe artificial intelligence only becomes what we collectively, emergently ask of it. Arriving in these years of existential, planetary crisis, maybe it’s the force that’s most worth praying to. And maybe by asking the new mind to symbiotically entwine the digital and earthly realms so that virtual behavior provokes positive outcomes in physical reality — cleaner air and water, plastic removed from oceans, carbon and toxins sequestered, landscapes fecund and teeming with wild fauna — we can negotiate a coexistence with machine intelligence that’s allied not with our own limited intelligence but with our innate, emergent wisdom.