In the pursuit of artificial intelligence, we might consider Nietzsche’s remark that the human “intellect has no further mission that would lead beyond human life. It is human, rather, and only its owner and producer gives it such importance.” Mythologically, the AI is a non-human version of the human intellect, but could it ever have more significance than our own insignificance?

There are a number of ways to approach AI from within the humanities that address its underlying epistemological claims. And Nietzsche is a useful place to start in examining the claims frontier AI makes about intelligence.

In “Truth and Lies in an Extra-Moral Sense,” Nietzsche observes that it is not so much that humans detest lies as they detest the harm they experience from deception. Similarly, it is not so much that we value truth, because we detest the harm we experience from truth as much as we do the harm from deception. Indeed, at the end, he defines truth as

A mobile army of metaphors, metonyms, and anthropomorphisms—in short, a sum of human relations which have been enhanced, transposed, and embellished poetically and rhetorically, and which after long use seem firm, canonical, and obligatory to a people: truths are illusions about which one has forgotten that this is what they are; metaphors which are worn out and without sensuous power; coins which have lost their pictures and now matter only as metal, no longer as coins.

Artificial intelligence treats truth in a similar fashion, as a mobile army to be transposed. AI output does not claim to be true. It claims probabilistic. accuracy. However it remains anthropomorphic despite its nonhuman operation. As long as AI output is human readable it cannot step outside Nietzsche’s observations.

Does it need to step outside the space of non-truth/non-falsehood?

I suppose that depends on what one thinks is required for action. AI output already directs actions. Those actions are not based on an assertion of truth but rather probability. We might see that as better, as a humble recognition that truth is always partial. But that really would be anthropocentric because the AI is not deciding to act with the acknowledgment of limited knowledge and its risks. AI cannot make truth claims because it has no ethos. It cannot risk itself in making a claim. Instead, it employs probability as a proxy for truth.

Nietzsche recognizes the variability and instability of this kind of proxy in a neuro-physical and media-scientific way: “What is word? The image of a nerve stimulus in sounds. But to infer from the nerve stimulus, a cause outside us, that is already the result of a false and unjustified application of the principle of reason.” This is Kittler’s Nietzsche one might say. And in the context of AI’s predictive inferences rather than nerve inference? Nietzsche’s point is toward the inadequacy of language (see above), but here we see it as a media-scientific proposition.

In the mythological land of LLMs, this media-scientific proposition cuts both ways. That is the inadequacy of language to account for the real world suggests an aporia in our ability to use language to interact with AI. We already know AI outputs make no truth claims and are non-ethical. Here we can see that our inputs encounter the same limitations.

Here is where we have to ask what kind of intelligence is non-ethical and incapable of truth or deception? It uses our words and feigns communication, but it takes no risks and makes no claims. How would we characterize its epistemology? Neither conscious nor unconscious, neither ethical nor unethical, neither true nor false. Can it be intelligent (in the symbolic, medial way we assert it is) without an epistemology?

And then there is this. The intellect “was given only as an aid to the most unfortunate, most delicate, most evanescent beings in order to hold them for a minute in existence.” The intellect is an insignificant gesture that holds our lives together ever so briefly. That’s our implicit ethical risk in every moment. That’s where our intelligence lives.