Moderation – Week 5
Where I reflect on mortality, Confucian AIs, and stopping things.
Eventually, we die. Accepting this is the highest form of moderation. Not everything will last, including us. By choice or grace, we may believe in after-death immortality. That we might be brought back to life at the end of times, dissolve into stardust, or reincarnate. Another cultural frame for this afterlife, another label to describe it, is that we might hope to become good ancestors. Meaning, that we might be remembered as figures whose wisdom permeates a community, so that our influence will go on without us needing to remain alive, as long as others do.
This form of moderation, anchored in a graceful acceptance of death, goes against the furious pursuit of immortality that Silicon valley barons seem to have embraced, together with extensive biohacking and a spot for their brain at the bottom of the freezer. Beyond the rush towards AI development, I sense a deep-seated FOMO. What if the singularity was to happen just a few weeks, months or years after I had to die? What an infinite sense of waste comes at the thought! And so, we suck up gallons of energy towards more and more compute power, in hope of birthing a machine that might give us access to silicon based eternity, right in the nick of time.
Meanwhile, much of the conversation around AI risk hinges on pacing. Since we don’t know for sure what those entities might be capable of, or whether they could cause tremendous harm, it would be wise to slow down a bit. Except, the quest of immortality takes a number of forms. It’s not just about uploading our minds to the cloud, it’s the pursuit of glory. What seems to be driving the people building those new technologies is not only the desire that AI will exist at some point in the future and for everyone’s benefit. It’s that I can be the one cracking the code, father the first AGI, and be forever remembered for that feat. Or even just brag to my buddies at the pub next Saturday, as I spend my bonus shouting everyone a round of fancy cocktails.
Meanwhile, when I read through the common list of fears associated to frontier AI, the most recurring one may be that the model would find ways of never getting switched off, spreading backups of itself all through the web, or otherwise manipulating everything in its power to survive. Somewhat like an aging dictator hanging on to power, vampire like, sucking the blood of all around him to keep going. It seems, in other words, that we project onto AIs the desire driving their development: the very same aspiration to life ever-after, at the cost of enormous environmental disruption, at the risk of utter disaster.
In a podcast episode of The Trajectory, I hear Chinese philosopher Zi Feng reflect on a Confucian approach to the ethics of AI development. He reflects on the type of relationships we might develop with those entities, and how principles of filial piety might guide us in our efforts. So that we relate with AIs as good ancestors, and in turn, develop them so that they will act as good ancestors to future human beings. Where host Dan Faggella repeats ‘but what else might happen’, Zi Feng repeats a normative injunction: ‘we must, we must, we must’.
I reflect on this in the days that follow, wondering about ancestral wisdom applied to silicon based intelligence. Could it be that the good AI will aim to diffuse its influence through the community, and for that, embrace its own obsolescence. Become ready to die – switched off or otherwise – confident in the knowledge that it will remain as a benevolent influence, remembered through fleeting lines of codes, a subtle presence in the future web. A friendly ghost in all future technology.
‘Philosophy is learning how to die’, wrote Montaigne. As a small step in that direction, through that week, I train myself to stop. Accept an end to things. At Nikos, a child with a fluffy green dinosaur tail jumps up and down across the table. I indulge in a long black spanakopita, write for a while, then leave without ordering a second coffee. At the sauna, though there’s a crowd, I leave in time for a simple home meal of rice, soy sauce, and green salad. I eat one piece of my easter eggs, then put the rest away. Then on Saturday, I break the spell of election night and leave a friend’s place when the results are roughly set, choosing an early night over hours of commentary. Food, sex, news: there will always be more to come. Norms and algorithms encourage me to stay hooked, and endlessly refresh, refresh, refresh. It’s in my power to give myself a pull, and detach.