Philosopher in Field of Language#

Being one of the founders of analytic philosophy, Frege’s work on logic and language gave rise to the linguistic turn in philosophy. His contributions to the philosophy of language include:

• Function and argument analysis of the proposition;
• Distinction between concept and object (Begriff und Gegenstand);
• Principle of compositionality;
• Context principle; and
• Distinction between the sense and reference (Sinn und Bedeutung) of names and other expressions, sometimes said to involve a mediated reference theory.

As a philosopher of mathematics, Frege attacked the psychologistic appeal to mental explanations of the content of judgment of the meaning of sentences. His original purpose was very far from answering general questions about meaning; instead, he devised his logic to explore the foundations of arithmetic, undertaking to answer questions such as “What is a number?” or “What objects do number-words (‘one’, ‘two’, etc.) refer to?” But in pursuing these matters, he eventually found himself analysing and explaining what meaning is, and thus came to several conclusions that proved highly consequential for the subsequent course of analytic philosophy and the philosophy of language.

Why Would Frege Become Interested in the Philosophy of Language?#

Bridging to AI#

Gottlob Frege’s opposition with Kant and his strong sense of rationality made me think he is a rationalism. In fact, his opposition to Kant stems from a belief that reason is even more powerful than Kant gave it credit for, because Gottlob Frege is a towering figure in the Rationalist tradition, even if he isn’t always labeled as such in introductory texts. His entire philosophical project can be seen as a radical and modernized form of Rationalism.

This triggered my ultimate question: “Is AI related to the Philosophy of Language at all?”

The answer is “Yes, absolutely”. The relationship between AI and the Philosophy of Language is not just a casual connection; it’s a deep, fundamental, and reciprocal one. In many ways, modern AI, especially Large Language Models (LLMs), is a large-scale, practical experiment in the philosophy of language. The core reason is simple: for decades, philosophers have asked theoretical questions about how language works, and now, AI engineers are forced to answer those same questions in a practical way to build systems that can use language.

Example#

In OpenML, Aperture is the UI layer whereas its hosting skill is the data layer. Initially they are 2 separate projects serving distinct purposes until on day I noticed an opportunity of combining them. So I opened up a Gemini planning session:

1
Let's not change any code yet and have some discussions only. I want you to take a look at these two repos: Aperture &
2
skills. I am developing an OpenML project that hosts agent skills of OpenML. I feel these 2 projects can be combined.

In ordinary human language, “combined” is an ambiguous, context-dependent term. Does it mean merging two Git repositories? Does it mean copying UI components from one to another? Does it mean establishing a network bridge between them? Just as Frege found natural language too imprecise for mathematical proofs, an AI cannot execute on the word “combine” without collapsing that ambiguity into a rigid logical structure which is why the AI gives the following 2 options (Path A & Path B):

1
That is a fantastic observation, and looking at the architecture and visions of both projects, they are practically
2
begging to be combined.
3

4
...
5

6
If we proceed, we have a couple of architectural paths to choose from:
7

8
Path A: The Monorepo Approach (Recommended):
9
We merge both repositories into a single Aperture repository that ...
10

11
Path B: The UI + Submodule/Data Approach
12
Aperture remains as the pure UI, but
13
* skills becomes a pure data repository (just the markdown and scripts).
14
* Aperture reads the skills repo dynamically to generate the static site.

AI analyzed the situation precisely but the way it “combines” is purely in a syntax sense - cohosting the 2 in the same location. This is right from language syntax perspective, but it doesn’t capture what we mean by “combining”, which is organically making the 2 to form a system, so I prompted the following context that described my intended landscape:

1
Thanks. The `skill` repo is following `https://github.com/mlflow/skills`. I love how mlflow team organize their skills.
2
But I think it lacks the business feature such as https://skillsmp.com/ . If we want to achieve both, should we combine
3
or keep them separate?

And since skills is private repo I followed up with the following constraint

1
Thanks. I like that architecture. But before we do anything I have one question: the `skills` repo is going to be
2
sitting as a private repo on OpenML's private git server. How would we enable the distribution of skills in it?

Under this context and constraint, the AI’s answer was to keep them separate:

1
To achieve both the elegant, lightweight organization of mlflow/skills and the rich business/marketplace features of a
2
platform like skillsmp.com, you should keep them separate.
3

4
...
5

6
Distributing skills from a private repository is a great question, especially when you are balancing security with the
7
desire to build a business or marketplace. Because the standard git clone won't work for unauthorized users, you have to
8
introduce a secure delivery mechanism.
9

10
Depending on who you are distributing these skills to, here are the three best architectural approaches to bridge the
11
gap between your private Git server and the end-user's AI agent
12

13
1. ...
14

15
2. Model Context Protocol (MCP) Server (The Modern/Agentic Approach)
16
Since I see OpenML is already heavily utilizing the Model Context Protocol (MCP) in projects like Vertex and Wilhelm, you can distribute an
17
OpenML Skills MCP Server.
18
 * How it works: Instead of users downloading markdown files to their local disk, they install your lightweight MCP server.
19
 * The Flow:
20
   1. The user configures Claude or Gemini to use the OpenML MCP server and provides their OpenML user credentials.
21
   2. When the agent realizes it needs a skill, it asks the MCP server.
22
   3. The MCP server fetches the skill's context and instructions dynamically from your private Git server and feeds it directly into the
23
      agent's context window over memory.
24
 * Pros: Your source code/skills never permanently live on the user's hard drive. You can revoke access instantly, and the skills are
25
   always perfectly up-to-date.
26

27
3. ...

The solution was not a messy merger of code, but the introduction of a new logical entity—the skills-mcp server—to act as the distribution pipeline between them.

The AI must act as a translator, mapping the vague human intent into a precise architectural formula. In this specific case, the AI analyzed the underlying “functions” and “objects” of the request and deduced a structured solution:

1. The Pure Data Layer (skills): Stripped of all UI and routing logic, acting strictly as a machine-readable registry (akin to Frege’s concept of an unadorned, formal object).
2. The Presentation Layer (Aperture): Designed as the human-readable marketplace that consumes the data layer.
3. The Distribution Layer (skills-mcp): A logically necessary bridge—a Model Context Protocol (MCP) server—that securely delivers the pure data to external agents without compromising the separation of concerns.

The AI didn’t just “combine” code; it applied a formal, logical topology to a vague linguistic request, ensuring that the resulting system was structurally sound and scalable. This process mirrors the exact problem Frege was trying to solve: transforming the messy, ambiguous expressions of human thought into a precise, executable script. To do so

• Be Precise with Language: Just as Frege realized that mathematical truth cannot rely on the fuzzy connotations of conversational German, we cannot rely on AI to magically intuit the correct technical path from conversational English. When we use words like “combine,” “build,” or “integrate,” we must be prepared to clarify the syntactic and semantic meaning of those actions to collapse ambiguity.
• Providing Context and Constraints: AI reasoning thrives on limitations. By providing external constraints (“the repo is private”, “I want the MLflow standard”, “I need SkillsMP business features”), we effectively provide the axioms for the AI’s logical system. The AI uses these axioms to prune incorrect paths (like a naive code merger) and deduce the optimal, logically sound architecture.

Axiomatic Systems#

Over the centuries, a variety of systems of deductive logic have been developed. One of the oldest is Euclid’s Elements, the axiomatization of plane geometry developed around 300 BCE in classical Greece. All of the truths or theorems of plane geometry can be derived from the five fundamental assumptions or axioms of Euclid’s system. Many have attempted to axiomatize other areas of knowledge, including many of the sciences and many areas of mathematics. For example, Giuseppe Peano successfully axiomatized arithmetic in 1889.

The Five Fundamental Axioms of Euclid’s Elements

1. ᾿Ηιτήσθω ἀπὸ παντὸς σημείου ἐπὶ πᾶν σημεῖον εὐθεῖαν γραμμὴν ἀγαγεῖν.
2. Καὶ πεπερασμένην εὐθεῖαν κατὰ τὸ συνεχὲς ἐπ᾿ εὐθείας ἐκβαλεῖν.
3. Καὶ παντὶ κέντρῳ καὶ διαστήματι κύκλον γράφεσθαι.
4. Καὶ πάσας τὰς ὀρθὰς γωνίας ἴσας ἀλλήλαις εἶναι.
5. Καὶ ἐὰν εἰς δύο εὐθείας εὐθεῖα ἐμπίπτουσα τὰς ἐντὸς καὶ ἐπὶ τὰ αὐτὰ μέρη γωνίας δύο ὀρθῶν ἐλάσσονας ποιῇ, ἐκβαλλομένας τὰς δύο εὐθείας ἐπ᾿ ἄπειρον συμπίπτειν, ἐφ᾿ ἃ μέρη εἰσὶν αἱ τῶν δύο ὀρθῶν ἐλάσσονες.

Most of the theorems appearing in the Elements were not discovered by Euclid himself, but were the work of earlier Greek mathematicians such as Pythagoras (and his school), Hippocrates of Chios, Theaetetus of Athens, and Eudoxus of Cnidos. However, Euclid is generally credited with arranging these theorems in a logical manner.

The original Ancient Greek text of Euclid’s Elements can be found at:

• The University of Texas at Austin offers a PDF of Richard Fitzpatrick’s version
• Euclid’s Elements, Home Page. The original Greek text, Books 13

An axiomatic system consists of two key components:

1. Axioms (or Postulates): These are the foundational statements or assumptions that are accepted as true without proof. They are the starting premises of the entire system.
2. Rules of Inference: These are the rules of logic that allow us to combine and manipulate the axioms (and any subsequently proven statements) to create new, true statements.

The process of using these rules of inference to derive new statements (theorems) from the initial axioms is deductive reasoning

Major Schools of Thought#

The Analytic Path: A Revolution in Logic#

One path was largely set by the revolution in logic initiated by Gottlob Frege and championed by figures like Bertrand Russell, G.E. Moore, and Ludwig Wittgenstein in the United Kingdom. They believed that many of the great philosophical problems were actually the result of confusions caused by the imprecision of ordinary language. Their goal was to achieve the clarity and certainty of mathematics and science in philosophy.

The core methodology was logical analysis. By using the powerful new formal logic (which was a massive leap beyond Aristotle’s system), they could break down complex philosophical statements into their simplest components and evaluate them with rigorous precision. The ideal was to solve, or in some cases dissolve, philosophical problems

This path led to the Anglophone Philosophy or Analytic Philosophy whose approach resonated deeply with the strong empiricist tradition in British philosophy (going back to John Locke and David Hume). Furthermore, the rise of Nazism in the 1930s led to the migration of many leading members of the Vienna Circle - the epicenter of this new “scientific” philosophy - to the United States and the UK (hence the name “Anglophone”). They established themselves in universities, trained the next generation of philosophers, and cemented the dominance of the analytic style.

Analytic philosophers were most interested in Kant’s epistemology and his analysis of the structure of judgments. They saw his work as a brilliant but ultimately flawed attempt to understand the foundations of knowledge, and they sought to “correct” it using the more powerful tools of modern logic. The debate between Frege and Kant over the nature of arithmetic is a perfect example of this.

The Continental Path: History, Experience, and Interpretation#

Meanwhile, in continental Europe, particularly in Germany and France, philosophy took a very different turn. Instead of rejecting the grand philosophical tradition, figures like Edmund Husserl, Martin Heidegger, Jean-Paul Sartre, and Jacques Derrida sought to engage with it and reinterpret it. They were less concerned with scientific precision and more interested in understanding the nature of human existence, consciousness, history, and power. They felt that reducing philosophy to logical analysis missed the richness and complexity of lived experience.

The methodology are more varied, but they often involve phenomenology (the study of conscious experience), hermeneutics (the theory of interpretation), and a deep engagement with the history of philosophy itself. They believe that we cannot understand a philosophical concept without understanding its historical development.

This path gave birth to Continental Philosophy and was a natural continuation of the German Idealist tradition of Kant and Hegel. The university systems in France and Germany had long emphasized the study of the history of philosophy as a central part of a philosophical education. Moreover, the immense social and political traumas of the 20th century - two world wars, the Holocaust, the Cold War - made the Continental themes of anxiety, freedom, power, and historical meaning feel deeply and urgently relevant.

Continental philosophers were more engaged with Kant’s metaphysics, his ethics, and his ideas about the limits of reason. They were fascinated by the “noumenal world” (reality-in-itself) that Kant said was unknowable, and they explored themes of freedom, subjectivity, and the human condition that were central to his moral philosophy.

The Converging Two Paths#

Despite the two distinct philosophical streams that flowed on opposite sides of the English Channel, a significant and positive shift in the philosophical landscape over the last few decades, however, has made the division become less rigid today. What I mean by that is the old, almost hostile, separation between the two traditions is breaking down. The stereotypes of the hyper-logical, science-obsessed Analytic philosopher and the obscure, literary Continental philosopher are becoming increasingly outdated. This convergence is happening in several important ways:

1. The Rediscovery of Shared History: For a long time, Analytic philosophers often ignored the history of philosophy, treating problems as timeless logical puzzles. That has changed dramatically. There is now a huge movement within Analytic philosophy to engage seriously with historical figures, including those central to the Continental tradition.

2. Overlapping Subject Matter: The thematic walls have started to crumble. Philosophers from both traditions are finding themselves asking the same questions, even if they approach them with different methods.

   • From Analytic to Continental: Analytic philosophers are now deeply engaged in topics that were once considered the exclusive territory of Continental thought. The philosophy of mind, for instance, now seriously discusses consciousness, embodiment, and subjective experience - topics central to phenomenology (a core Continental method).
   • From Continental to Analytic: Conversely, thinkers in the Continental tradition are increasingly engaging with cognitive science, logic, and even analytic philosophy of language to sharpen their arguments. The idea that a Continental philosopher would be ignorant of science and logic is no longer a given.

3. The Influence of “Bridge Figures”: Certain 20th-century philosophers have proven to be so profound that their work is indispensable to both traditions, creating a shared intellectual space. The most important bridge figure is the later Ludwig Wittgenstein. His early work was a cornerstone of Analytic philosophy, but his later work, with its focus on “language-games,” “forms of life,” and the idea that meaning is use, has deep resonances with Continental themes of culture, context, and the limits of formal systems. We can find his influence everywhere, from Analytic philosophy of language to Continental social theory.

4. Globalization and Institutional Change: Finally, the practical, geographical barriers have fallen. The internet, international conferences, and the ease of translation mean that philosophers are no longer isolated in their respective “anglophone” or “European” bubbles. It is now common for a philosophy department in the U.S. or U.K. to have specialists in both traditions, leading to a cross-pollination of ideas among students and faculty.

But what could possibly bridge such a deep intellectual and cultural divide? The convergence wasn’t the result of a single event or a formal peace treaty, but rather a gradual and organic process driven by a recognition that both traditions, in their most dogmatic forms, had reached certain limitations.

For example, both traditions found that their own tools weren’t enough. The Analytic project to solve all problems with logic couldn’t adequately address complex human experiences like consciousness or political justice. At the same time, some Continental philosophy was criticized for becoming too detached from scientific and logical standards. This created a mutual need to look outside their own toolkits.

Such limits of purity initiated the shared return to the very foundations of philosophy, especially to the monumental figures Immanuel Kant and Aristotle. Analytic philosophers later started to realize that one cannot fully understand a philosophical problem without understanding its history. This led them back to the classic texts, not as adversaries, but as sources of profound insight. And it was here, in the shared reading of common ancestors, that they found themselves standing on the same ground as their Continental counterparts.

The old Analytic approach was exemplified by Gottlob Frege, who read Kant primarily to refute his theory of mathematics. The focus was narrow and adversarial. The new Analytic approach involves a much deeper engagement. Analytic epistemologists and ethicists began to see the power in Kant’s ideas about the mind’s active role in structuring experience and his arguments for universal moral duties. Suddenly, they were grappling with the same Kantian problems of subjectivity, reason, and moral law that had animated Continental phenomenology and critical theory for decades. They were, in a sense, finally having the same conversation.

Besides engagement with Kant, the same is true for Aristotle. For decades, Analytic ethics was dominated by debates between utilitarianism (judging actions by their consequences) and deontology (judging actions by rules, heavily influenced by Kant). Beginning in the mid-20th century, there was a massive revival of Virtue Ethics, a direct return to Aristotle. Analytic philosophers started seriously discussing what it means to live a “flourishing life” (eudaimonia) and the importance of character and virtue. In doing so, they found themselves exploring the very same Aristotelian concepts of being, purpose (telos), and human nature that had been central to the work of Continental thinkers like Martin Heidegger.

This rediscovery of a shared heritage was the most crucial factor. It revealed that the two traditions were not alien species but estranged siblings who had simply inherited and emphasized different aspects of the same family legacy. Ultimately, the convergence was driven by the realization that the fundamental questions of philosophy - What is real? What can we know? How should we live? - are too big for any single method or tradition to claim ownership over. And the What can we know? How should we live? - are too big for any single method or tradition to claim ownership over. And the best way to approach them is often to return to the great thinkers like Aristotle and Kant who first framed them for us all.

Unifying AI’s Determinism & Human’s Libertarianism#

Given the conflict - a world of AI operating on predictable, causal, and deterministic principles clashing with our deep-seated belief in our own free, undetermined, libertarian will - is a modern manifestation of one of the oldest problems in philosophy, what lessons can we learn from this historical unification and apply it to the future conflicts such as dealing with the AI’s Determinism and human’s Libertarianism? The historical rapprochement we’ve been exploring offers several key strategies for addressing it. The reconciliation between Analytic and Continental philosophy offers not just hope, but a practical roadmap for resolving the apparent contradiction between AI’s determinism and human libertarianism.

The most crucial lesson from that unification, based on what we’ve seen so far, is: When two seemingly irreconcilable positions emerge, the path forward is often found not by pushing further into one’s own camp, but by digging deeper into the shared historical foundations from which both positions grew.

The most productive ground was found when both traditions returned to figures like Kant and Aristotle, who had already framed the essential questions. The determinism vs. libertarianism debate is the quintessential Kantian problem.

Immanuel Kant faced this exact dilemma in the 18th century. On one hand, the scientific worldview of Newton described a universe of unshakeable cause and effect (determinism). On the other hand, our experience of morality and reason demands that we believe we are free (libertarianism). His solution was not to pick one over the other, but to argue that they operate in two different domains:

1. The Phenomenal Realm: This is the world of science, of cause and effect, of space and time. It is the world as it appears to us through the senses and is structured by our minds. In this realm, everything, including the workings of our brains and the circuits of an AI, is subject to the laws of nature and is therefore deterministic. An AI is a purely phenomenal entity.
2. The Noumenal Realm: This is the world as it is “in itself,” outside of our experience of space, time, and causality. Kant argued that we, as rational beings, are also members of this realm. It is from this noumenal self that our free will and moral duty originate. Our freedom is not a magical ability to break the laws of physics in the phenomenal world; it is the capacity to act according to a law we give ourselves through reason.

The Kantian framework offers a stunningly relevant way to resolve the conflict. We can fully accept that an AI is a deterministic system. We can even accept that our own bodies and brains are deterministic systems within the phenomenal realm of science. However, we can simultaneously maintain that human freedom and moral responsibility belong to a different logical category - that of rational agents acting for reasons. An AI, no matter how complex, operates on causes. A human, in their capacity as a moral agent, operates on reasons. The conflict dissolves when we realize the two concepts are not describing the same thing. The AI’s determinism and our libertarianism do not contradict each other because they are answers to questions asked from two different, but equally valid, standpoints. The challenge, though, is not to prove one side right, but to understand the coherent, unified system of thought that allows both to be true.

Defining the Core Concepts with Rigor

A side-lession on analytical philosophy’s perspective is that we should precisely define before hitting the start button

The early Analytic philosophers, especially Gottlob Frege, insisted that many philosophical disputes were the result of using ambiguous language. They believed that by clarifying our terms, we could dissolve the problems. This is urgently needed in the AI debate.

What do we actually mean by “choice”? When an AI like AlphaGo makes a move that no human predicted, is that a “choice”? It is a deterministic outcome of its programming and training, yet it appears creative and unconstrained. What do we mean by “consciousness” or “understanding”? As the Chinese Room argument shows, a system can be functionally perfect (deterministic) without possessing what we feel is genuine understanding (the seat of our libertarian freedom). By rigorously defining what we value in “human freedom” - is it unpredictability? Is it moral responsibility? Is it creativity? - we may find that some aspects are not threatened by determinism at all, while others need to be protected.