Don’t get any on you

This is Jessica. In “A Glass-Bottomed Cadillac”, David Hickey describes the advice Hank Williams Sr. received from his father and passes on to his son: “Don’t get any on you, pipsqueak.” By which he meant, don’t let the moral fallout of the road permeate your sense of self. Will Oldham describes the same struggle to retain yourself while the world presses in:

I am still what I meant to be And I’m losing my mind But our burdens must lessen Though our enemies thrive

These days it seems there’s plenty of compromising mess to get on a person who isn’t being careful. By using certain services or purchasing from certain companies, you may be implicitly empowering forces you don’t agree with. Take X, fka Twitter, for example. Until a few days ago they were still enabling anyone willing to pay a small fee to generate child sexual abuse material. To Elon Musk, stopping users from subjugating whoever they please, regardless of their age, was unnecessary censorship. And yet, most of the people I know there went about their business on that platform without so much as a peep. By which I mean breathlessly posting about all the other, apparently more intellectually gratifying look-at-what-the-AI-did-now stuff. Watching AI research conversations play out on social media lately is both exhilarating and exhausting, with the volume of news coming out, and the sense of needing to not miss out on the next source of buzz that it fosters. One could even liken being in AI or ML to being a porn or sex addict, with all the urgency and heavy breathing. But that’s an analogy for a different post.

A less experienced version of myself might have felt personally offended by the thought of friends continuing to support a situation that actively disenfranchised people like me. But one eventually learns that going through life taking such things others do personally is no recipe for peace of mind. I also know from experience that the internal calculus does not always feel so easy when it comes to deciding when to vote with one’s feet. In the case of X’s knowing enablement of child porn, the right thing to do may have seemed obvious (like, don’t send your money every month to an entity that is enabling CSAM at scale!) But I get that severing connections with things you perceive as important to your goals can be difficult, and I think many researchers see their social media influence as part of their identity and evidence of success. Still, the whole situation makes me feel hesitant about going back to that platform.

From a practical perspective, part of the problem with getting too offended by people for not standing up to systems in which they are embedded, even if the principle of it does really does seem obvious, is that it assumes a level of intentionality that they may not have. To be clear, I do not mean that the choice is not ultimately available to them, or that it’s wrong to hold people accountable for their actions. Rather, I have been thinking about how hard it can be to possess oneself, and how many people do not fully possess themselves. By which I mean, they lack either the imagination or the intrinsic motivation to live by their own values. Or as Gram Parsons once put it,

Some of my friends don’t know who they belong to Some can’t get a single thing to work inside

It’s possible to shut many things out in the name of “making it.” In my experience what gets shut out, and what you let get to you, is often not so much a choice as an attribute of your level of self-knowledge and self-ownership about yourself at that point. Are you capable of imagining a version of yourself that remains true to your values even if it causes friction with other perceived needs? Are you capable of imagining a version of yourself that is sure enough of what you’re doing that the friction disappears? Could you love yourself as much or feel as secure in your position without your Twitter account?

It’s easy to accumulate impurities in the quest to be something greater than your current state. As Rafe Meager writes in their recent essay, “In professional life, one is obligated to traffic in a certain amount of bullshit. We all know that. But it has to be finely calibrated, and that is hard.”

Calibrating our behavior and beliefs is hard in part because we can become conditioned to acting against our sense of self as a normal part of personal and professional growth. Actively moving toward the things that challenge you is not necessarily a flaw, and certain things must be endured to succeed in a new game. And often the things that challenge us are the things that don’t come naturally, that are not, at first glance, clearly aligned with our values.

When I was younger, one of the last things I wanted to be was a computer scientist. This is not to say I’m unhappy, or I don’t enjoy what I do now, because I very much do. I just can’t even imagine having to explain to my high school self how that happened, as what we do in computer science (and to some extent in other disciplines I frequent, like statistics) has never felt like the things that I’m most intrinsically motivated to be good at – I always did very well in math and science classes, but I had no passion for it. But somewhere along the way I got disillusioned with the pursuits that seemed more aligned with my values, like art, and so I started gravitating to the things that seemed most different from what bothered me. After enough years of testing to see how far you can take something, it becomes hard to go back, and it does start to feel like what you’re meant to do. Meanwhile, the Destroyer song loops in my mind,

Don’t become Don’t become The thing you hated The thing you hated The thing you hated

It’s a dangerous game, getting very good at letting go of things that once seemed important to who you are, in favor of urges to be something else. On the one hand, there’s a real power to be gained in separating yourself from the things you think you need. Sometimes after you take the first step, it’s like something clicks and there’s a high in suddenly realization that something you were stuck on bothers you no more. Or maybe it still does, but somehow now you can find peace in settling back to watch the tides of your cut-off aspirations and desires continue to pulse back toward what was severed. A way of earning indifference. I believe it’s possible to literally remake yourself this way.

And yet part of me, in looking back, feels like maybe I let myself down. Maybe I would’ve been a better writer or philosopher, both of which I always felt more personal proclivity for. Or maybe I would just feel less disillusioned now if I’d kept up certain hobbies that do matter to me—like writing poetry—more consistently, rather than dropping it for so many years because I couldn’t imagine what it would mean to be a person who did both. It was a failure of imagination: I couldn’t conceive of devoting myself to becoming the new thing while also retaining other aspects of myself.

What’s interesting is that this kind of letting yourself down, by letting yourself drift too far from your original purpose or what you feel like you’re best at, is not necessarily so different from lacking the imagination to do the right thing in the current political moment. There is a loss in both cases, a quiet slipping away of who you really are while you think you’re out there proving it. While you think you’re the one who has your priorities straight, while you’re striving to play the game, or maybe you’re even killing it, comparing how many followers you have or how many papers you wrote this year to those around you. The ignorability of it all is terrifying. Kierkegard got it right when he wrote that “The greatest hazard of all, losing one’s self, can occur very quietly in the world, as if it were nothing at all. No other loss can occur so quietly; any other loss—an arm, a leg, five dollars, a wife, etc.—is sure to be noticed.”

How do you tell whether letting go is growth or self-betrayal—especially when what we “need” narrows what we can even see? On the one hand, by moving more towards statistics and formalization over time, my thinking has expanded to encompass new forms of rigor. But it’s a kind of rigor along narrow lines. In another sense I lost imagination, in that it’s now harder for me to take seriously things that I can’t fit into my formal frameworks.

Did I lose ownership of myself? When I think about what it means to fully possess oneself, I think of Aristotle’s preoccupation with explanations of the inner principles that determine an entity’s states of change and rest. He distinguished that which exists by nature (physei, φύσει) and that which exists from other causes (di’ allas aitias, δι’ ἄλλας αἰτίας). It’s the first kind that’s self-possessed: it “contains in itself its own archê (ἀρχή),” the principle and origin of its entry into presence; the second “does not have its principle in itself,” but finds it in the productive activity of human beings.

This is why failures of imagination feel so tragic. To lack imagination is not only to fail to picture alternative versions of the self; it is to lose contact with the inner source that could have animated them. You become legible, optimizable, and perhaps successful, but successful in the way an object is successful when it performs its intended function. You can be moving quickly, and still be at rest with respect to yourself.

Researchers are beginning to ask how, and if, generative AI systems can attain something like intrinsic motivation. How do you get it to devote itself to an open-ended goal, like creative expression, that can’t be boiled down into a simple reward model? In a recent paper, Charness and Grieco find that AI outputs outperform human outputs (as determined by other humans) for tasks that are more clearly specified in terms of how to solve (“closed”), like writing short stories using specific required words. But AI outputs consistently underperform human outputs for open-ended tasks, like inventing things, where the participant is required to find, invent, or discover the problems. They propose a model for the utility function the agent faces as depending on three factors: the output they’d get from simply following instructions, deviation from the instruction-following output due to randomness (e.g., model temperature), and the utility of exercising imagination. Models can only follow instructions more or less closely, and obtain more diverse outputs through higher temperature, but humans alone respond to the pleasure of bringing imaginative ideas to life, a proxy for intrinsic motivation.

Is intrinsic motivation partly a matter of temporal depth—the capacity to care about consequences that don’t pay off immediately, or even in any obvious reward currency? What kind of “vision” allows an agent to see farther ahead, and be moved by what it sees? Yeats, in A Vision, has a line that keeps returning to me: “The Spirit … may know the most violent love and hatred possible, for it can see the remote consequences of the most trivial acts of the living, provided those consequences are part of its future life.”

The need for this kind of vision feels politically relevant now, as programs are dismantled and regulations rolled back or rewritten in ways that will reverberate for years. As a professor, most present in my mind are the moves that affect science in this country—visa regulation, institutional acquiescence when under political attack, the seeding of doubt in the goals of science or value of education.

But Yeats’ quote also feels very personally relevant. What would it take for us to be able to see this way in our personal choices? When we ignore the alliances we support through our various choices, we narrow our vision on purpose. It’s a survival mechanism, but it comes at a cost.

Ultimately I don’t think anyone owes me an explanation of their internal calculus. But neither do I owe them the assumption that they are in control of their drives. There was a time I might have tried to convince myself that the right interpretation was to give everyone the maximum benefit of the doubt. But as Andrew emphasizes, steelmanning is its own hang-up.

Meanwhile, there are days now where I feel like I’m just waking up, from a period of my life where I became laser focused on goals and outcomes and forgot about everything else. It’s quite painful sometimes, existing with this new awareness of what was always there, an inner “archê” that I’d let go quiet. More bluntly, it’s a cliche mid-life crisis with everything but the sports car. But the waking up is also very beautiful: realizing, suddenly, how real things are, how alive, and how little use there have for whatever games you’ve been playing. You don’t get infinite chances to notice.