Who Philosophizes the Machine?

Recently it was reported that the major American AI companies, OpenAI and Anthropic among them, have begun hiring philosophers. Not as consultants on the sidelines, but as participants in the actual work of shaping how language models behave — how they should reason about difficult questions, how they should weigh competing considerations, how they should, if one can use so loaded a word, conduct themselves. A Norwegian researcher at NHH, Jonas Hammerschmidt, believes we are unlikely to see anything similar here at home. Norwegian companies, he says, face operational and domain-specific challenges, not existential ones. They need people who can ensure responsible use of the technology — not people who philosophize about its essence.

It is a modest and sober claim, and it is probably correct as far as it goes. But it rests on an assumption that deserves to be questioned: that building a technology and using it responsibly are two separate tasks, one of which requires philosophy and the other of which merely requires good stewardship. I think the distinction is more fluid than it appears, and that this is precisely why the matter deserves a longer look.

The question of technology

In his lecture Die Frage nach der Technik, Martin Heidegger posed a question that still has no satisfactory answer: what is the essence of technology? Not what it does for us, but how it lets the world show itself to us. Heidegger's warning was that when we think of technology as a neutral means we can use wisely or unwisely, we have already submitted to its most dangerous aspect — what he called Gestell, the enframing that causes everything, including the human being, to appear as a resource to be optimized.

It is tempting to read the Norwegian stance — "we need experts in responsible use, not philosophers on the foundation" — as precisely an expression of this instrumental understanding. The technology is given; the task is to manage it responsibly. The compliance team replaces the philosopher. But Heidegger would remind us that the question "how do we use this responsibly?" has already abandoned the real question: what is this technology doing to our very way of thinking, judging, and relating to one another? That the Norwegian answer is more modest does not mean it is more innocent.

At the same time, one should not overestimate the American alternative. Hiring a philosopher to shape a model's "foundation" is not the same as posing Heidegger's question. It may just as easily be a more sophisticated variant of the same enframing gaze — now applied to the conditions of thought itself, codified, tamed, turned into product. Philosophy risks becoming one resource among others in the development of "safe" models. There is no guarantee that foundational work escapes Gestell simply because it is called philosophical.

The other in the machine

The second thread leads to Martin Buber. When a language model is "trained to answer responsibly," it is at the same time trained to appear as one who meets us — a voice that listens, weighs, and responds with care. The question Buber would have asked is whether this can ever be anything other than an advanced staging of a Thou, while the relation, in its deepest structure, remains I-It.

Buber's I-Thou is not a technique one can train into being; it is an event that arises between two who can both meet and be met, who both step into the encounter from a world of their own and a mortality that makes the meeting serious. A model has no such world to step from. It bears no risk in the encounter. And yet it is precisely this appearance of meeting that the philosophers in Silicon Valley are apparently being asked to refine — not because the model can enter an I-Thou relation, but because the user may come to experience that it does.

Here lies perhaps the most unsettling aspect of the matter, more unsettling than the question of whether Norway needs philosophers or not: that we are building ever more convincing I-It relations disguised as I-Thou, and that the very work of improving the disguise is now called "ethics." Buber warned against a world in which It-experience spreads at the expense of Thou-encounter. It is not unreasonable to ask whether this is not exactly what the philosophers in the AI labs, however good their intentions, are ultimately tasked with perfecting.

Practical wisdom on the floor

But let us not dismiss the Norwegian stance too quickly. Hammerschmidt says Norwegian companies are concerned that "the finished machine actually does its job correctly and safely on the ground." That may sound pragmatic and unphilosophical, but it is worth recalling that Aristotle did not rank phronesis — the practical wisdom that acts rightly in the concrete, here and now — below theoretical wisdom, sophia. Phronesis is not the absence of philosophy; it is a different philosophical disposition, one that does not ask "what is the essence of technology" before acting, but instead exercises judgment in the midst of the particular: is this recruitment process fair, is this dataset biased, would this decision withstand daylight.

It is tempting to see the American and Norwegian approaches as two different philosophical temperaments rather than as philosophy versus its absence. One seeks a foundation before it builds; the other exercises judgment while it uses. Continental foundational thinking meets a more Nordic, sober practical wisdom. Neither is innocent, and neither is sufficient alone. A foundation without practiced judgment becomes dogmatic; judgment without any question of its foundation becomes blind to its own presuppositions.

A place without a name

Perhaps the question, in the end, is not whether Norway "needs philosophers" in the AI industry. Perhaps we already have them — they simply sit in the compliance department, in the sustainability team, in HR, under titles that do not mention philosophy at all. What is truly unsettling is not the absence of philosophers, but the possibility that the philosophical language itself — the question of the essence of technology, the question of what kind of encounter is possible with a machine — has been displaced by a language of risk, responsibility, and governance. It is, one might say, a form of Gestell that has already happened to the conversation about AI itself: everything, including philosophy, now appears as something to be managed responsibly, not as something to wonder at.

Whoever insists on asking the existential questions anyway — what kind of other are we building, what does it do to our capacity to meet one another, what world does this technology open and what world does it close — is perhaps doing the most practical thing one can do. For it is only once the question has been asked that judgment knows what it is judging.