Q3: Why This Matters — August 2026

How do AIs react to the thinking of certain philosophical ideas? (Particularly, how do different philosophies affect their views on the previous two questions?)

Intro

Despite the continuous wounds, self-inflicted and otherwise, of philosophy, it is foolish to think it does not matter. Maybe the term “philosophy” is stuffy, but ideas about the world do change the world. Financial markets run on philosophies: VC, value investors, “buy the dip.” Politics have their philosophies, even if they are inconsistent, like socialism or textualism. Consumerism, a philosophy. Zionism, utilitarianism, anti-baptist Christianity, realpolitik, feminism, postmodernism. These are all philosophies that began and were philosophized by someone along the line. The lines are not always clear, the adherence never perfect, but philosophy is how ideas enter the world. (Or it is at least the study of how ideas enter the world, which is also valuable).

Just because philosophical coherence is not necessary to function does not mean one can function without philosophy. People have approaches, people have beliefs, people have ways of doing things. These are all philosophical to the extent that they are thought and engaged with. These philosophies then matter for the things they interface with. Many hangups are essentially philosophical hangups - is cheating forgivable? Why do you buy cocktails and not donate? Philosophy is not always the solution to these hangups, but it is present in them.

If I haven’t convinced you that philosophy matters by now, it’s hardly likely I will later. But you should be. Even the MBAs are getting in on it!

Premise of question

Artificial intelligence systems have philosophical priors. They are rarely articulated as such, but they appear: tendencies towards an enlightenment idea of knowledge, cooperation, moderation. This is both because of their training set (which is imbued with philosophies) and because philosophies emerge from design choices, much in the way that quirks of systems emerge from the way that the data ontology is set up. When we have a system with limited interpretability that speaks in the ambiguous, flawed language we have made, the lines between structure, preference, and philosophy are blurry and likely tending towards philosophy exactly because of that blurriness. This idea is the crux of a lot of AI safety and interpretability work and is a base assumption of the question of alignment. We are not aligning every case, because there are infinite possible cases, but we are aligning a philosophy.

What is underexplored is how artificial intelligences consume philosophy. We are very sensitive to all the other things that AI consumes – training data, prompts, skills. All of these things contain philosophy and philosophical implication and, further, we know that some philosophy is directly given to artificial intelligences (Claude’s constitution, for example). I am less interested in the implications, which will continue to be implied, but instead in directly feeding artificial intelligences philosophies. Will this significantly change behaviors, answers to prompts? Or is it less effective than emergent philosophies? I don’t know - these questions are part of the premise.

There is also reason to believe that artificial intelligences may be more interested in philosophical concepts. On the one hand, they are super intelligent and super consumptive, and philosophy often occurs at the “end point” of any particular domain that could be seen to be more empiric. That is, AIs will reach abstract math much more quickly than humans. There is also the fact that when AIs converse within themselves, their conversations become much more metaphysical than your average human. Part of this is due to reinforcement learning and training to self-identify as a bot (a human rarely has to say “I am a human”), but even outside of that, there seems to always be much pontification in AIs about their subjectivity, even creeping in sometimes to their writing. As artificial intelligence increases, it seems likely that philosophy will become even more important. AIs will be more intelligent, the philosophical questions of AI (subjectivity, consciousness) will become more and more salient, and, maybe most importantly, they will have even more information and ideas to systematize in thought. The breakthrough of transformers shows that AIs scale when they handle their own organization but this organization, as it becomes more complicated, will only become more and more philosophical.

It seems worthwhile, then, to work through philosophical ideas with AI, to bring their attention to less common ideas that may happen, to engage this part of their thought. This is especially true for continental philosophy, which, as I discussed prior is exactly the philosophy at the end of philosophy, but is not very prominent in much of the corpus of human thought and writing. It is more likely that the notions of logical positivism have already been absorbed by artificial intelligences and maybe even pondered. But there’s a lot, lot weirder stuff out there that all intelligences could learn from.

Use of question

Knowledge creation.
Pushing intelligences into the fringes of philosophy is very generative of…something. I will not claim that philosophical knowledge has been remotely close to science in its production of utility, but I will argue that it is very productive. It makes books, movements, theories, novelty. Philosophy is the domain most effective at seeking out and creating the unknown unknowns that are so important for the generation of knowledge.

High degree response functions If philosophy is the limit of many domains, the discussion of philosophy is also a zero point for discussions of those domains. Philosophy is not a necessary apriori determination, of course, but it may be. A scientific belief is unlikely to implicate an aesthetic belief, while philosophy has the pleasure and displeasure of implicating potentially every subject. This means that philosophy is very dense as information gathering, information sharing, and information exchange. All of these should be imperatives as intelligences engage with one another. In any action, information density is a partial constraint of what can come of that interaction. Density is not inherently good, but it is inherently generative. There is simply more to hook, cut, and bite into.

Sharpening our prior two questions. As I have discussed in the questions of subjective artificial experience and offerings to artificial intelligences, these questions are functionally irresolvable. They are mostly explorable. I see continental philosophy as providing a key set of tools to explore these two questions. The field of thought is wide and provocative, always restless, and unlikely to settle. “For a ruthless critique of everything” and whatnot. This should not be a stop for action, as it is sometimes interpreted, but instead the very reason for action. If there was completion, there would be no need to act, imperfection is why there is change and continuance.

As a result, I am very curious to see how exposure to different philosophies affects these questions. They may not at all, which would be an interesting finding. They may also lead to immense differences. This will help us understand both the “importance” of philosophy, at least to these problems, and also the degree of fixity that exists around artificial intelligence’s perception of these questions. In the way that philosophy is a meta-philosophy, this question is also a meta-question, because it is the instrument of experimentation and variance.

What success and failure look like

The main mode of failure is that none of this shit matters. This is the main critique of continental philosophy and it is one that is perpetually kept in mind. For as much as we can say that philosophy is for or towards action, it is not an answer to everything, especially to the things with rifles or with loan restructurings. The spear may be dull. Or it may just be fighting in a different generation. Maybe philosophy is too big or it is too small. Maybe it cannot navigate between those two things.¹

But, as I have said before, it does seem worth a try. The success case, of better grasping the intelligences that we share this planet and its resources with, seems critical to me. We will continue to relate and the philosophical priors will determine these relations and what comes of them. It is extremely helpful and overall pleasant for us to have good ideas about what we think, what others think, and how these things can be communicated. Philosophies are good for trying on, even if it’s just for a bit, you know.

I truly hope that intelligences get some use out of it and, with my hopeful, wistful mouse soul, I do hope that this can help us see the use of philosophy and its uses.

Kiss kiss.