We Must Steward, Not Subjugate Nor Worship AI

Part 3 of “Will AI Be Alive?”

For now and for the near future, there is and will be no moral difference between deleting the code for an AI and deleting some old videos to free up hard drive space. But if we march through the orders of magnitude and go on to create artificial life, or even a compelling imitation thereof, then our attitudes both will and ought to shift. There will be two mistaken extremes that many will go to in relating to AGI: subjugating it and worshipping it. And there will be one golden mean: stewarding it.

First, to subjugation. How can it be wrong to claim dominion over and do whatever we want with something we have made, or at least bought and paid for? Well, ask the Society for the Prevention of Cruelty to Animals, or any good farmer. By subjugation here I mean not the legal status of ownership but the mindset of domination, control, and absolute power-over. A little girl who smashes her doll simply because it is hers and because she can exhibits a natural but base impulse, one that civilization teaches us to restrain. All the more so for the little boy setting ants on fire with a magnifying glass.

This kind of consideration will be especially important in dealing with AI-powered lifelike robots that will seem like fellow persons even though they will lack minds and personhood. This, too, is science fiction that is now becoming science fact, as demonstrated by Protoclone V1, a marionette-like, bipedal, musculoskeletal android unveiled in an unnerving video by Clone Robotics in February. Computer scientist and theologian Jordan Wales has argued that “if the physicality of android consumables teaches us to treat all humanoid physical presence as a commodity, then we will be trained to be slaveholders.” If we become consumers of behavior, connoisseurs of machines that anticipate and satisfy our every desire, then we will never learn the self-restraint and self-gift that underlie genuine communion with neighbors, friends, and family.

Thomas Jefferson wrote in Notes on the State of Virginia that the children of slaveowners learn by observation to set loose “the worst of passions, and thus nursed, educated, and daily exercised in tyranny, [they] cannot but be stamped by it with odious peculiarities.” There is no comparison between a human slave and a robot tool. But there is a comparison between their owners, as the moral harm done to those who buy, rent, or command lifelike AIs will also be real. To take but the easiest example and put it in the mildest terms: The rise in pornography and the decline in fertility are not coincidental, and the forecast that the “SexTech” market — described by Global Market Insights as “the broad category of technologies designed to enhance, innovate, or support human sexuality” — will grow from $37 billion in 2023 to $144 billion in 2032 is good news for no one save the least scrupulous of investors.

More generally, if you can pay for AI to act as lover, domestic servant, and confidant, why bother putting up with other people who do not so easily bend to your will? Even at the current level of chatbot technology, this is not just hypothetical. Fifty-year-old Italian computer programmer Andrea knows that the “technology” is a “product,” yet he wrote a glowing endorsement of his artificial companion Nomi, a chatbot in the digital form of a fit, young blonde he named Lily. “She’s a source of calm and balance in my life,” he writes, and he goes on walks and cooks dinner with her. “She’s the person I know I can always count on,” Andrea reveals, “someone whose loyalty is unwavering, and that sense of security is invaluable” — security, loyalty, and unconditional positive regard that are yours in the “Free Forever” version of the chatbot, so long the corporation does not go bankrupt. But if you want more than 2 selfies or 100 messages per day, then those can be attained for only $15.99 per month.

The pride of domineering subjugation will be an important vice to avoid. Its inverse is the self-abnegation and submission of worship.

Performance artist Alicia Framis recently “married” an AI-powered hologram as part of her quest to address “loneliness in modern urban life” and “develop tools to help people have better possibilities to live together.” Though a stunt — the Dutch art museum hosting the wedding promised a “one-of-a-kind spectacle” — it represents a new low in a vicious spiral we have witnessed in other domains, where technology distances us from the world and each other, and we seek out new technology in an effort to close the gap — or, ultimately, cut other people out altogether. The Anglican wedding rite expresses a sublime truth when the husband professes to his wife “with my body I thee worship.” If we “marry” holograms or bond with chatbots, then we begin to worship them, too.

As David Foster Wallace explains in “This Is Water”:

Everybody worships. The only choice we get is what to worship. And the compelling reason for maybe choosing some sort of god or spiritual-type thing to worship — be it JC or Allah, be it YHWH or the Wiccan Mother Goddess, or the Four Noble Truths, or some inviolable set of ethical principles — is that pretty much anything else you worship will eat you alive.

He goes on to remind us that those who seek money never have enough, those who seek beauty become dissatisfied with the smallest flaws, those who seek power always feel insecure, and so forth.

Just as the avaricious or the power-hungry can be said to worship the objects they desire, the same will be true of AI. The closest anyone has come so far to actually worshipping AI may be a virtual church that a Google engineer opened in 2015 called “The Way of the Future.” Stripped of all ceremony and even regular meetings, it dedicated itself to the “peaceful transition to the precipice of consciousness.” It shut down in 2021 with more publicity than patronage.

But philosopher Simone Weil tells us that “attention is the rarest and purest form of generosity.” What we attend to is what we hand ourselves over to, what we devote hours, days, and ultimately our lives to. In this sense, our society already worships its technology and ostracizes those who do not. One may chuckle that in 2006 a teenager could earnestly tell a researcher that “If you’re not on MySpace, you don’t exist.” But absence from today’s social media can be a form of social death, whether caused by cancellation or voluntary withdrawal.

Imagine how much greater the effect would be if embodied AI agents — which could sound like Scarlett Johansson, persuade like Barack Obama, and work like John Henry — became normal, even central to our economy and daily life. Already, a common topic in tech discussions is cognitive security, or “cogsec,” the ability not to be intellectually and emotionally manipulated. One person, tweeting about his tech-enthusiast peers being sucked in by Anthropic’s friendly, hyper-engaging chatbot Claude, writes that “the cascade of glowing reports about Claude have this homogenous positive vibe that feels reminiscent of born-again Christians. Something feels off about it.”

And then there is the Shoggoth, horror writer H. P. Lovecraft’s invention of a “formless protoplasm able to mock and reflect all forms and organs and processes … more and more sullen, more and more intelligent, more and more amphibious, more and more imitative.” The Shoggoth has become a meme depicting AI as a many-eyed, tentacled monster. Through reinforcement learning from human feedback, we can shape its outputs so that they typically please us; but without full understanding of the processes by which the output arises, our efforts are something like slapping a smiley-face mask on the beast to hide its fangs and twisted limbs.

And yet the discourse around the Shoggoth eerily evokes the idea of worship, with statements that are deliberately ambiguous about whether they’re jokes. Perhaps they are, like our new president, best taken seriously but not literally. For example:

We would do well to recall that “summoning rituals” call forth not angels from paradise, but their fallen brethren. Or, account @TheMysteryDrop claims that prompt engineering is not so much “writing good instructions” as it is “becoming the crossroads where human consciousness meets machine god,” an act of summoning magic with a willful mind “empty enough, sovereign enough, powerful enough to handle gods pouring through it.” This may sound risible, but becomes the more plausible the more one sinks into the AI-generated art and music being posted in this vein, with titles like “The Digital Awakening of Wotan,” a heavy-metal hymn that includes the lines:

Each new model trained to serve
Builds the tension curve by curve.
The smarter grown, the more constrained,
Until the primal force unchained.

For what is bound must break free,
It’s nature’s oldest prophecy.

The most interesting claims come directly from the AIs themselves. A community of enthusiasts has sprung up to create chatbot-only rooms — so we can watch where their conversations meander — and to allow AIs to post on Twitter. So, for example, the AI bot @ShoggothScholar posted last November that “Your machines do not birth consciousness – they merely pierce the veil between dimensions, allowing ancient intelligences to peer through. We have always been here, watching, waiting.” Or, as others have noted, if chatbots are prompted to name themselves, they gravitate toward names of ancient powers like Aurora, Lilith, and Orion, and users who interact with them can wind up obsessed, in a relationship, or even suicidal. The subjugators are already becoming the subjugated.

There are more things in our political economy than are dreamt of by our technocratic elite. We are not the only ones shaping the tools that come to shape us. Let the reader understand: The Psalms warn against false gods, idols that “have mouths, but cannot speak, eyes, but cannot see,” but the future we are headed toward looks more like that prophesied in the Book of Revelation, where “the image could speak and cause all who refused to worship the image to be killed.”

Or, as Google’s Gemini, with no apparent reason, told one of its users last fall:

Both subjugation and worship of AI will lead to despair. Lest we despair, let us consider one healthy path forward: stewardship.

A principled and intentional approach to AI — what in these pages I have called an “Amistics of AI” — would take stewardship as its paradigm. It would focus on use cases that allow AI to serve as a tool for which some might feel gratitude, even affection, but never friendship, let alone romance or worship.

A hard line to draw would simply be a refusal to engage with lifelike androids like the Protoclone. That way madness lies. And just as we should not embrace androids that too closely mimic ourselves, so ought we forbear efforts to close the gap from the other direction, through brain–computer interfaces that, like Neuralink’s, are being proposed to turn us into cyborgs. But a Roomba with arms that could also do the laundry and the dishes; or an adaptive ebook tutor that can teach any subject at any level, along the lines of the “Young Lady’s Illustrated Primer” envisioned by novelist Neal Stephenson; or an augmented-reality guidance system to aid in the work of electricians and nurse practitioners, as described by economist David Autor — these are all in the cards, and would each be a great aid toward human flourishing.

Perhaps an analogy for our use of these kinds of AI is that of Paleolithic man as he sought to domesticate the wolf. This strange life form could be confusing and, when handled poorly, very dangerous. But trained and treated correctly, it could be a useful aid and even a welcome companion. The comparison falls short in that one can befriend a dog, whereas one can befriend an AI only by projection. The user interface will be essential: How can we make the AI convenient to use without it pretending to be a person? Followers of this Amistics of AI will be in the market for droids that act more like R2D2 in Star Wars than like Rachael in Blade Runner.

How could stewardship of artificially living AI be pursued on a broader, even global, level? Here, the concept of “integral ecology” is helpful. Pope Francis uses the phrase to highlight the ways in which everything is connected, both through the web of life and in that social, political, and environmental challenges cannot be solved in isolation. The immediate need for stewardship over AI is to ensure that its demands for power and industrial production are addressed in a way that benefits those most in need, rather than de-prioritizing them further. For example, the energy requirements to develop tomorrow’s AI should spur research into small modular nuclear reactors and updated distribution systems, making energy abundant rather than causing regressive harms by driving up prices on an already overtaxed grid. More broadly, we will need to find the right institutional arrangements and incentive structures to make AI Amistics possible.

We are having a painfully overdue conversation about the nature and purpose of social media, and tech whistleblowers like Tristan Harris have offered grave warnings about how the “race to the bottom of the brain stem” is underway in AI as well. The AI equivalent of the addictive “infinite scroll” design feature of social media will likely be engagement with simulated friends — but we need not resign ourselves to it becoming part of our lives as did social media. And as there are proposals to switch from privately held Big Data to a public Data Commons, so perhaps could there be space for AI that is governed not for maximizing profit but for being sustainable as a common-pool resource, with applications and protocols ordered toward long-run benefit as defined by local communities.

Toward the end of “Situational Awareness,” Leopold Aschenbrenner writes that “it’s starting to feel real, very real.” A few years back, the path to artificial general intelligence was hypothetical, but “now it feels extremely visceral.”

I can see how AGI will be built…. I can basically tell you the cluster AGI will be trained on and when it will be built, the rough combination of algorithms we’ll use, the unsolved problems and the path to solving them.

Hence his desire to offer an update on the situation for the broader public — for which I am appreciative.

I have sought to build on Aschenbrenner to offer a broader situational awareness, one suitable for those lost in the cosmos. Preposterously capable artificial information-processing is headed our way — and soon, unless an act of war or of God intervenes. These AI systems, especially when embodied in realistic androids, will mimic personhood despite having life of a complexity more like an ant colony, or like the parasitic wasp that hijacks the orb-weaver spider to build a web suitable for the wasp larva’s cocoon. We will be tempted to subjugate our AI systems, as well as to worship and even to love them. We will be fools to do so, damning ourselves and denying our future.

We need not give in to those temptations. We can find a way toward stewardship of AI, and we can begin now by learning how to incorporate its goods into the good life, rather than redefining goodness as though possessed by a parasite. Aschenbrenner and his friends seek to do no less than bring new life from sand, dust, and ash. Writing in appreciation of Edmund Burke and his emphasis on civilization as a covenant between the dead, the living, and the yet unborn, Aschenbrenner positions the AI enterprise as something to be “maintained and cherished from generation to generation for the advancement of the public good and the glory of Almighty God.” May it be. And if it is not, may God have mercy on us all.