My ongoing conversation with my friend Daniel about AI has led us to two new topics. (I apologize for the liberal use of “we” throughout this post. “We” here is meant to refer to what is commonly thought and said, to what is usually assumed as what “we”–as a community sharing a language–know.)
The first is instinct versus consciousness. I think I may have written about this before. But, in any case, I want to state here why I find the whole notion of instinct bankrupt. Here’s the case, stated succinctly.
The whole notion of instinct is suspect. Instinct is used to say some behavior is automatic, baked in, not a product of reflection or deliberation. And then we say something is not instinctual when we claim it is a matter of choice, that it could have been otherwise. That seems wrong at both ends. Wrong insofar as all kinds of things we treat as matters of conscious choice are much more automatic (much more unconscious if we want to use that vocabulary) than we admit. Wrong insofar as so-called instinctual behaviors actually have lots of wriggle room for responding appropriately to specific environmental conditions. Creatures adapt to their surroundings so it isn’t as if “instinct” dictates one pathway to the exclusion of all others. In short, the whole notion of instinct seems dubious and lazy, not attuned to actual behavior and its variants. To put it another way, the binary instinct/non-instinctual doesn’t hold up if examined with any kind of care.
The second is whether it makes sense to to think of AI as “beyond good and evil.” And, thus, to think that the panic over AI is just a response to the dizzying prospect of a complete revaluation of values, a step into the unknown once God and humanism are truly dead. What would it mean to accept those deaths and face their full consequences– a facing up that Nietzsche believed we hadn’t even begun to undertake. We were (are) just trying to patch together something palatable out of the fragments of a shattered tradition, acting as if we will manage to salvage something. “These fragments I have shored against my ruin” (T.S. Eliot, “The Wasteland.”)
I don’t think the previous paragraph is quite the right way to think about Nietzsche. But placing Nietzsche into the conversation about AI’s “ethics” is illuminating. So here’s how I would pursue that juxtapostion. Nietzsche thought (as I read him) that humans cannot do without values. So it’s a revaluation of values, not the abandonment of values. And (again as I read him) Nietzsche proposes the “affirmation of life” as the supreme value. The nihilist is someone who can’t affirm this life, this world. That’s why the Christian is a nihilist. And the Christian is also a hypocrite because a Christian ethics is as driven by the will to power as all other human behavior is. It’s just that the Christian denies that his ethics are driven by resentiment and consist of a subtle way to gain power over the strong, noble man. Part of the noble man’s strength, crucially, is his acceptance that the affirmation of life necessarily entails knowing that predation is inevitable as is suffering. There’s a constant oscillation in Nietzsche between glorying in the aggression that accompanies the necessity of predation (kill or be killed) and a passivity that sees strength as being demonstrated by being capable of standing up to the inevitable (and terrible) suffering that life will deal out. Amor fati.
[A sidenote: a major temperamental difference between conservatives and leftists is this relation to predation. The conservative fully believes that it is killed or be killed, that predation is a law of life. And the conservative scorns the leftist for hypocritically refusing to acknowledge this truth even as the leftist’s safe life is secured by policemen and soldiers to whom the leftist outsources the dirty work of killing. All of this gets wrapped up in notions of masculinity–a fact amply demonstrated in Nietzsche’s own work with its denigration–and terror–of anything that smacks of the feminine.]
What seems different with AI is that it is motiveless. Unlike the Nietzschean noble man, it does not aim, without pretense or disguise, for power. The noble man will live life to the fullest, embracing all it has to offer, suffering with pleasure. AI, it would seem, has no values at all, nothing that orients its actions. That makes it seem incalculable. There’s no telling what it might do.
In our dealings with people, we come to understand what they value, what they would defend or refuse to sacrifice. That, in extreme circumstances, can give us a “hold” over them if we have the power to threaten what they hold dear. Even in less extreme circumstances, we negotiate our relations to others in relation to what they want/need—trying in the cases of love and friendship to honor and advance those wants/needs and trying in cases of exploitation to use those wants/needs to get from them what we want.
But it is hard to see that there is any basis for a “hold” over AI. It is beyond good and evil (I guess) in the sense of having no values at all. So, yes, it exists in a very novel space in respect to sentient life, which does have an orienting value (or values)—even if we describe that value in selfish gene terms. (I.e. we don’t need to say the value is something the individual organism chooses or is even conscious of. But we do say there is a value that underlies that organism’s behavior, that gives us a way to explain it.)
My point is that Nietzsche is very much in the game of identifying an underlying value. But that AI seems to me to introduce a different game. What if there is behavior/action apart from any value? What, then, motivates or drive the behavior? It would be easy to say “the algorithm,” to say AI is just an automaton, no different than your dishwasher. But that’s too easy because AI, unlike the dishwasher, does not follow a set path. It innovates; it does unexpected things. How? And Why? Right now, I don’t think we have any good answers for How and Why. Which is uncanny, to say the least, because our usual way of rendering behavior/action intelligible (i.e. by indicating a motive or intention aligned with a value) is not applicable in this case. And we haven’t devised an alternative explanatory schema at this point.
