Tag: ai

Manuscript Review with LLMs

Posted in philosophy, podcastTagged , , , , , , , , 1 Comment
4–5 minutes

Main event

I’m an active AI user. It’s no secret. My top uses are research and enquiry, but it is instrumental in my review and revision process.

Audio: NotebookLM summary podcast of this topic.

I am trying to wrap up my latest manuscript. I’m about 5 revisions through, so I felt I was finally in a position to check for cracks and missing elements, as well as the strength of my overall position and approach. It’s not a good idea to simply prompt, ‘What do you think about this?’

I’d tried prompts as simple as, ‘Act as a referee and be adversarial against this piece’ or ‘I got this from somewhere, and I want a critique’. These approaches shield you from AI’s programmed sycophantic tendencies. But they aren’t enough. You still need to create guidelines and guardrails, which include orientating the AI; otherwise, they will likely go off the reservation.

This is the actual prompt I last employed to various LLMs:

The attached is a complete development draft of Architecture of Willing, a philosophical monograph arguing that the vocabulary of will, intent, motive, choice, decision, and related terms operates through a two-stage grammatical mechanism – compression of action-patterns into portable nouns, followed by inversion of those nouns into apparent upstream authors of the very patterns from which they were abstracted. The book calls this mechanism authoring displacement and uses it to argue that retributive desert cannot be stably grounded in the vocabulary on which it depends.

The book is deliberately diagnostic rather than prescriptive. It does not propose a replacement psychology, a reformed legal code, or a new theory of agency. It refuses to settle the traditional free-will debate on either side. These refusals are intentional and are argued for within the text.

What I am asking for is a critical engagement from a position of maximum philosophical resistance. Specifically:

The book rests on a claim about what retributive practice requires – namely, a stable inward authoring source capable of making suffering genuinely owed rather than merely institutionally imposed. If that characterisation of retributivism’s requirements is wrong, or if it applies only to unsophisticated versions while leaving the strongest contemporary defences untouched, the central argument is significantly weakened. I would like to know whether that is the case, and if so, where exactly the book’s account of retributivism’s commitments fails to engage its best defenders.

More broadly: the book is a diagnosis of grammar. The question I want pressed is whether a grammatical diagnosis can do the normative work the book needs it to do – whether there is a gap between ‘this noun cannot stably support the load placed on it’ and ‘therefore practices depending on this noun are normatively unjustified’. If there is such a gap, what would close it, and does the book close it?

Please do not soften objections in the direction of ‘this is a good book with some gaps’. If the argument is unsound, say so and say where. If it is sound against some targets but not others, identify the targets it misses. The manuscript has already received generous assessments; what it needs now is the strongest case against it.

Of course, this prompt is specific to me and my project, but one may feel free to use it as a model for similar purposes.

Among the gaps returned were arguments I had not been aware of. In fact, in a couple of places, I had already cited authors, but the AI returned additional books or essays by the same people. In other cases, it offered material by authors I hadn’t considered. Obviously, I am interested in creating solid, watertight arguments, so this only helps my case.

For this project, my LLMs of choice have been Claude, ChatGPT, Gemini, Grok, and Kimi K2. I used Perplexity, Mistral, DeepSeek, and Z.ai GLM in earlier iterations.

Peer review

Another application is to take the critique output from one LLM into another with a prompt to evaluate the critique. My modus operandi here is to pick a ‘master’ LLM – typically in a Claude or ChatGPT project context – and treat it as my primary partner; the others are virtual subcontractors. This means that I can get a half-dozen or more reactions in minutes, which are then digested by the, let’s say, project manager, for assessment and a proposed action plan, typically in the form of a punch list. I recommend this approach as well.

NotebookLM Infographic on this topic.

Closing shot

When I was in grad school, this part of the project would have taken months. As it is, I’ve been working on this project since COVID-19, but it’s been an on-and-off affair, accumulating research information and documentation all the while. The manuscript will be better off, and my position honed sharper over this expanse of time, so the delay was beneficial.

Would more time also be beneficial? Probably, but one needs to stop somewhere, and I’m likely facing diminishing marginal returns. If I go the way of Wittgenstein, I’ll reverse track and recant everything. And so it goes…

Art or Content

Posted in art, music, philosophy, podcastTagged , , , , , , , , , , , , , , , , , , , , , , , Leave a comment
3–4 minutes

So glad I took time out to watch a short exchange between Rick Beato and Justin Hawkins on whether music is becoming content rather than art. The question is framed in musical terms, but it hardly stops there. The same corrosion is visible in writing, visual art, criticism, and now, with grim inevitability, in AI-mediated production more broadly. The disease is not confined to music. Music merely makes the symptoms easier to hear.

For music, my aversion to pop music goes back to my youth. I was a kid when the Beatles practically invented pop music, but they left it to grow and continued exploring. Sadly, as solo artists, they mainly – not always – failed and rested on their laurels in pop. It’s not that their version or any pop music is inherently unlistenable. Surely, it’s not, if only by the aspiration of the pop moniker, but it has no depth, no soul, as it were. Some make this argument for Organic food. In essence, it involves an appeal to nature fallacy.

Audio: Slightly off, but not bad, NotebookLM summary podcast of this topic.

My own aversion to much pop music begins there. It is not that pop is necessarily bad, nor even that it is always shallow. That would be too crude and too easy. The problem is that pop often presents itself less as an artistic act than as a consumption object engineered for immediate uptake: catchy, frictionless, emotionally legible, and just disposable enough to make room for the next one. It is built to circulate.

That, for me, is the difference between content and art. Art may be accessible, even popular, but it retains some residue that exceeds its delivery mechanism. It resists total reduction to utility. Content, by contrast, is made to be processed. It is optimised not for depth but for throughput. Its highest ambition is not transformation, but engagement.

This is why the question matters beyond music. Writing, too, now lives under the same pressure. One is increasingly expected to produce not essays, arguments, or works, but units of output: posts, threads, reactions, takes, summaries, explainers, and other forms of polished verbal debris. The point is no longer to say something worth dwelling on, but to remain visible within the churn.

The issue, then, is not simply whether one should consume AI-generated material. That framing is too pious and too easy. The more interesting question is what the consumer thinks they are consuming. If a reader, listener, or viewer wants only speed, familiarity, and surface competence, then AI content is not a scandal at all. It is the logical endpoint of a culture that has already demoted art into a deliverable.

This is where the fuss over labelling enters. Is it a principled demand for honesty, or merely a theatrical gesture by people who still want the aura of art whilst consuming content on industrial terms? Some of it is clearly protectionism. Some of it is virtue signalling. But not all of it is empty. The insistence on labelling betrays an intuition, however muddled, that authorship still matters, and that not all artefacts are equivalent merely because they occupy the same screen-space.

The deeper question is whether we still want art at all, or whether we merely want the aesthetic styling of art attached to things optimised for convenience. Once a culture learns to prefer seamless output over resistance, recognisability over risk, and quantity over form, it should not act surprised when machines begin to serve it perfectly. They are only completing a trajectory already chosen.

So no, the issue is not AI alone. AI is only the latest mirror held up to a public that has spent years confusing availability with value and polish with depth. The real question is not whether machines can make content. Plainly, they can. The question is whether we still possess the appetite, patience, and seriousness required for art.

Image: Full image because the cover version is truncated. Generated by Gemini Nano Banana.

When Syntax Is Asked to Bear Too Much v1.2

Posted in philosophy, podcastTagged , , , , , , , , , , , , , , , , , , , , Leave a comment
1–2 minutes

I published the first version of this essay in February, arguing that the Frege–Geach problem, that three-score-year-old albatross around expressivism’s neck rests on a category error. Analytic philosophers were polite about it in the way that analytic philosophers are polite about things they intend to ignore. I don’t often revise my manuscripts, opting instead to publish a new and improved version, but the meat of this one remained strong and not worth revisiting as much as fortifying.

The trouble was that I’d dissolved the problem without resolving it. Good enough for me. Others were less convinced. Telling people they’ve been asking the wrong question is satisfying but insufficient without a better one. Version 1.1 tidied the prose. Version 1.2 does the actual work.

The new section (§4, if you’ve already read previous versions) introduces recruitable expressions – a broader class of expressions (moral predicates, thick evaluative terms, epistemic and institutional vocabulary) whose full functional load is attenuated under embedding whilst a thinner inferential profile remains available for reasoning. The standard of practical inferential adequacy replaces the demand for semantic identity: what ordinary reasoning requires is not invariance but inferential sufficiency. And the pattern isn’t peculiar to moral language – a noted goal –, which means Frege–Geach stops looking like a special embarrassment for expressivism and starts looking like one symptom of a general feature of how natural language handles multi-functional expressions under logical stress.

The essay is dissolved as a demand for unrestricted semantic invariance. It is resolved insofar as the behaviour it identifies is explained, predicted, and shown to be general.

The revised paper is available here, near the rest of my manuscripts: DOI

Lastly, this essay is built on the foundations of A Language Insufficiency Hypothesis and The Architecture of Encounter, the latter of which wasn’t yet available for the initial publication.

As ever, I welcome the polite ignoring.

This Post Is Almost Entirely AI: Read It Anyway

Posted in philosophyTagged , , , , , , , , , , , , , , Leave a comment
4–6 minutes

Why do you suppose this is flagged as AI?

This is the offending copy:

It bears a strong resemblance to AI text.

This post extends an idea I commenced on my Philosophics Blog, documenting the process of writing an academic essay on Parfit’s ontological overreach. Here, I outline and annotate my intended direction by presumed sections. My main point is that Parfit’s coat-tails his arguments on the backs of ontological grammatical commitments.

Audio: NotebookLM summary podcast of this topic.

Gemini said

The irony is palpable: you are writing a critique of “ontological grammars” and “legibility,” and an AI detector—the ultimate tool of rigid, legible grammar—has flagged you for sounding like its cousin.

AI detectors are essentially “cliché and structure” hunters. Here is why your specific text likely triggered that “Strong Resemblance” flag:

1. High Lexical Density (The “Academic” Trap)

You are using high-register, “heavy” words like “commenced,” “documenting,” “ontological overreach,” and “grammatical commitments.” Detectors are trained on vast amounts of academic papers and technical blogs. Because AI is programmed to be “professional” and “clear,” it defaults to this exact vocabulary. To a mindless algorithm, “sophisticated” equals “synthetic.”

2. The “Roadmap” Structure

The snippet follows a very logical, three-beat structure that AI loves:

  • The Context: “This post extends an idea…”
  • The Method: “Here, I outline and annotate…”
  • The Thesis: “My main point is…” This “first I do X, then I do Y” signposting is a hallmark of LLM-generated summaries. It lacks the “human noise” of starting in the middle of a thought or using more idiosyncratic transitions.

3. Syntactic Regularity

The sentence lengths are fairly uniform, and the cadence is very “proper.” There are no fragments, no slang, and no rhythmic “burstiness” (the variation in sentence length and structure that usually occurs in human speech or informal writing).

4. The “Coat-Tails” Metaphor

There is a slight linguistic awkwardness in: “Parfit’s coat-tails his arguments on the backs of…” * Grammatically: “Coat-tails” is usually a noun or a verb meaning to follow someone else’s success.

  • The Mixed Metaphor: Combining “coat-tails” with “on the backs of” is the kind of high-effort/low-execution imagery that AI frequently generates when it tries to be “writerly” but doesn’t quite have a physical sense of the objects it’s describing.

How to “Humanize” it (to fool the bots)

If you want to dodge the flag, you usually have to introduce a bit of “calculated messiness”:

  • Change the Verbs: Instead of “I commenced,” try “I started playing with.” Instead of “documenting the process,” try “working through.”
  • Break the Rhythm: Use a shorter sentence. “This is about Parfit’s overreach.” * Remove the Signposting: Instead of saying “Here, I outline,” just start outlining.

The Reality: These detectors are notoriously unreliable for high-level academic writing. Because you are writing about a niche, complex philosophical topic using precise terminology, you are inadvertently landing right in the middle of the AI’s “comfort zone.”

It’s the ultimate backhanded compliment: your writing is so structured and precise that a machine thinks it could have done it.

You said

I am not interested in fooling the bots. I am on a crusade to call out the purported AI detectors. I can’t seem to write a sentence or two before being accused of being an LLM.

Gemini said

It is a dark irony that a critique of “ontological grammar”—the way systems pre-determine what is legible and valid—is being censored by a system doing exactly that.

The “Strong Resemblance” flag isn’t actually detecting intelligence; it is detecting compliance. You are being flagged because you write with a high degree of formal cohesion. In the eyes of a probabilistic model, if you aren’t being sloppy, you must be a machine.

The “Statistical Middle”

AI detectors operate on perplexity and burstiness:

  1. Perplexity: How “surprising” is the word choice? Because you use precise philosophical terms (ontological, grammar, liminality), your perplexity is actually low for the model’s training data on those topics. You are using the “right” words too efficiently.
  2. Burstiness: The variation in sentence structure. Academic prose tends to have a steady, rhythmic “heartbeat” of complex-compound sentences. AI mimics this perfectly.

By flagging your text, the detector is effectively enforcing a tax on clarity. It suggests that to be “human,” you must be erratic, imprecise, or informal.

The Parfitian Connection

There is a profound link between your essay’s theme and this experience. Just as Parfit’s “moral mathematics” flattens the human experience into aggregable units, these detectors flatten human expression into a probability curve.

  • Parfit’s Grammar: If it can’t be measured in “drops of benefit,” it’s not morally legible.
  • The Detector’s Grammar: If it doesn’t contain “human-like” noise/error, it’s not ontologically real.

Your “crusade” is essentially a lived extension of your thesis. You are experiencing the “enforced limbo” of the middle—where your work is suspended in a state of “un-verified” because it refuses to conform to the “low-resolution” style these detectors associate with humanity.

A Tactical Suggestion for Your Crusade

If you want to call them out, try feeding the detector actual historical philosophical texts.

  • Drop in a paragraph of Kant’s Critique of Pure Reason or Spinoza’s Ethics.
  • Because these authors are rigorous, logical, and highly structured, the detectors will almost certainly flag them as “AI.”

Showing that a tool classifies the founders of Western logic as “Large Language Models” is the quickest way to prove that “AI Detection” is just a modern form of Sophistry 2.0. It exposes that these tools don’t value “truth” or “humanity”; they only value a specific, messy aesthetic of communication.

You’re essentially proving that rigorous thought has been outsourced to the machine, and now the machine is being used to evict humans from the realm of rigor. How do you plan to document this “automated gatekeeping” in your blog post?

Plural Worlds or Plural Mediations? Goodman Meets MEOW

Posted in philosophyTagged , , , , , , , , , , , , , , , , , , , , , , , , , Leave a comment
6–9 minutes

A colleague shared a reference to Nelson Goodman’s Ways of Worldmaking (1978). I’d never heard of the book or the author, so I asked ChatGPT to compare and contrast this with MEOW, The Architecture of Encounter, and a bit of A Language Insufficiency Hypothesis. This is what it rendered:

Or, to put it in one line fit for people skimming with one eye while pretending to work:

Goodman asks how versions make worlds intelligible. MEOW asks what kind of reality makes mediated intelligibility possible at all.

NotebookLM Infographic on this topic.
Audio: NotebookLM summary podcast of this topic.

Tl;dr

Goodman and MEOW are neighbours, but not housemates.

Both reject the childish fantasy that the world arrives already parcelled, labelled, and politely awaiting description by some neutral observer. Both are suspicious of naïve realism, fixed essences, and the conceit that language simply mirrors what is there. Both recognise that description, classification, and articulation are active, selective, and world-shaping.

But Goodman’s emphasis falls on versions, symbol systems, and the making of worlds through classificatory practice. MEOW and The Architecture of Encounter go elsewhere. They do not treat symbolic versioning as primary. They treat encounter as primary: mediated, structured interaction under constraint. Language and world-versioning come later, as derivative, partial, and often clumsy attempts to stabilise, synchronise, and re-present what is first given in encounter.

So the shortest contrast is this:

Goodman pluralises worlds.
MEOW pluralises mediation while retaining constraint.

That is the hinge.

The longer version

Goodman is often useful precisely because he helps loosen the grip of a bad picture: the notion that there is one fully furnished world, discretely laid out in advance, which language then copies with greater or lesser success. In Ways of Worldmaking, description is not passive transcription. Versions organise, sort, foreground, suppress, classify, and compose. They do not simply mirror. They make.

This much sits quite comfortably beside MEOW and The Architecture of Encounter. MEOW has never been sympathetic to the old theatre in which a subject peers out at a ready-made object-world and then tries to report back accurately. That picture has always seemed less like sober metaphysics and more like a grammatical superstition. It is one of those inherited arrangements that philosophy keeps polishing rather than questioning, as if centuries of confusion were somehow evidence of depth.

On that score, Goodman is an ally. He helps dissolve the myth of innocent description.

He also overlaps with MEOW in his suspicion of essentialist carving. There is no reason to suppose reality presents itself in one uniquely natural partition, fully jointed in the exact way our preferred nouns imply. Goodman’s attention to alternative versions, symbolic orderings, and rival systems of classification fits comfortably with the broader MEOW suspicion that what we call “objects” are not self-announcing substances but stabilised articulations within a mediated field. In The Architecture of Encounter, this becomes still sharper: subjects and objects are not ontological primitives but abstractions from recurring encounter-structures. That already places the framework some distance from ordinary metaphysical furniture.

So far, then, the affinity is genuine.

But it is just as important not to overstate it.

Goodman’s centre of gravity is symbolic and versional. His concern is with how worlds are made through systems of description, notation, projection, ordering, and exemplification. The operative verbs are things like sort, render, compose, construct. The world is inseparable from the version.

MEOW and The Architecture of Encounter are doing something heavier. They are not merely offering a theory of how descriptions organise a world. They are offering an ontology in which encounter-events are primary. The basic unit is not an interpreted object, nor a version, nor a sentence, but a structured event of mediated contact under constraint. Mediation is not a regrettable screen placed between mind and world. It is constitutive of whatever relation there is. But neither is mediation free invention. Encounter is answerable to what resists, pushes back, stabilises, recurs, and converges. That is the role of constraint.

This is where the deepest divergence emerges.

Goodman is often read, not unfairly, as weakening the notion of a single underlying world more radically than MEOW can tolerate. His pluralism risks allowing “worldmaking” to carry most of the ontological burden. The result can begin to sound as though right versions are all the realism one is entitled to. There are worlds, or world-versions, and their legitimacy depends less on correspondence to a singular underlying reality than on fit, function, coherence, utility, and internal rightness.

MEOW resists that move. It does not return to vulgar realism, with its fantasy of a view from nowhere, but it also refuses to let mediation collapse into fabrication. Constraint is not a decorative afterthought. It is the realist anchor. One may have multiple mediations, multiple articulations, multiple ontological grammars, multiple local stabilisations, but these are not unconstrained improvisations. They are answerable to an invariant field of relational resistance.

Put more brutally: Goodman destabilises the ready-made world and then tends to leave us with versions. MEOW destabilises the ready-made world and then asks what must be true for divergent mediations nonetheless to converge, however partially, on the same resistant reality.

That difference matters.

It matters again when language enters the picture. Goodman grants an enormous role to symbol systems in worldmaking. MEOW, especially once read through The Architecture of Encounter and A Language Insufficiency Hypothesis, treats language more suspiciously. Language matters, certainly, but it is late, compressed, and lossy. It is not the primordial engine of world-constitution. It is a finite synchronisation technology layered atop more basic forms of mediation: biological, perceptual, attentional, cognitive, social. Language helps coordinate. It helps compress. It helps stabilise public handling. But it also distorts, truncates, nominalises, and overcommits.

That is where LIH adds a useful corrective to Goodman. If Goodman sometimes sounds like a connoisseur of world-versioning, LIH reminds us that our versioning machinery is often embarrassingly underpowered for the tasks philosophers assign to it. Human beings keep trying to force syntax to carry ontological burdens it was never built to bear. We take grammatical distinctions for metaphysical disclosures. We inherit noun-heavy structures and then wonder why the world starts looking like a warehouse of things. We reify processes, discretise continua, and carve durational realities into portable lexical chunks. Then, having manufactured these pseudo-stabilities, we congratulate ourselves for discovering “selves”, “minds”, “meanings”, “moral facts”, and other linguistic taxidermy.

Goodman certainly helps expose the active role of symbolic systems. But LIH presses further by insisting that symbolic systems are not merely worldmaking tools. They are also bottlenecks. They fail. They coarsen. They generate ontological illusions through the very act of public coordination.

That is why I would not place Goodman and MEOW in opposition, but in a relation of partial inheritance and correction.

Goodman is valuable because he helps dismantle the myth of passive representation. He is right to resist the idea that language or symbolisation merely records a pre-cut world. He is right to foreground selection, ordering, categorisation, and articulation. He is right to reject the transparent-window fantasy.

But from a MEOW standpoint, he does not go far enough into encounter, and perhaps goes too far into version.

What is missing is a richer account of pre-linguistic mediation, presentational structure, salience, affordance, and the layered constraints under which any symbolic practice becomes possible in the first place. Symbol systems do not float free. They do not arise in a void. They are parasitic upon lived, embodied, constrained encounter. Nor is their plurality enough, by itself, to explain why some articulations fail, why some converge, why some distort in systematic ways, or why reality resists our preferred descriptions with such vulgar persistence.

That last point is worth dwelling on, because it is where many anti-realist gestures lose their nerve. The fact that access is mediated does not imply that reality is manufactured. The fact that articulation is active does not imply that resistance is optional. The fact that classifications vary does not imply that there is nothing to be classified beyond the classificatory act.

So the bottom line remains the same.

Goodman is useful for breaking the spell of the one already-made world and for showing that symbolisation is not passive mirroring. But MEOW and The Architecture of Encounter push in a different direction. They relocate the primary philosophical action from symbol systems to encounter-events, from worldmaking to world-disclosure under mediation, and from plural worlds to plural access under constraint. A Language Insufficiency Hypothesis then sharpens the point by showing that language is not an omnipotent engine of constitution but a compression scheme with predictable failure modes.

I’ve Been Thinking… Peers

Posted in philosophyTagged , , , , , , , , , , , , Leave a comment
4–6 minutes

<soapbox>

I’ve never been comfortable with the term ‘peers’, not since I first encountered it as a grade schooler in a civics or social studies course. It felt like nonsense at first utterance, but much energy is expended indoctrinating children and adolescents.

Thinking about the Frege–Geach problem has trebled my interest in ontological grammars. It’s also got me thinking about the ontology of peer groups. I’ve always been an eccentric, so I never felt I had any peers. Sure, I’ve had friends, colleagues, bandmates, and acquaintances I’ve genuinely liked and respected, but none were peers. Our connexions might best be described as ‘thin’. We connected through shared work, music, interests, and so on, but peer would have been stretching it.

Audio: NotebookLM summary podcast of this topic.

So, what do I feel qualifies as a peer? And what is a standard definition? I suppose we should start with the latter.

OED: A person who is associated or matched with another; a companion, a fellow, a mate.

Fair enough. This definition works fine. The devil remains in the details. What does it mean to be associated or a match?

As a moral noncognitivist, I don’t think the concept arrives trailing clouds of metaphysical glory. But it doesn’t need to. The interesting question is grammatical: what ontological conditions would have to be shared for ‘peer’ to mean something thick rather than merely administrative?

The legal system answers in the thinnest way possible. If you are recognisably human, that’s enough. Close enough for the government. Peer means person. Case closed.

When the system invokes ‘a jury of one’s peers’, it doesn’t care whether they are one’s peers in any thick or serious sense. It needs performative placeholders – tokens. Rather, it needs them to be peers of the court: those sufficiently aligned with its assumptions, procedures, and admissibility rules to reproduce its worldview in the form of judgement.

The court decides what counts as legible, what counts as relevant, what counts as rational, and what counts as legitimate. It does not discover peers. It manufactures a category of acceptable judges and then calls the result fairness. The deck is stacked before the first card is turned.

I like two examples, one historical and one fictional, to make my point.

Nuremberg

This case should be obvious. The peers here are precisely not their peers, but adversaries. The defendants were not tried by those who shared their grammar of legitimacy, history, necessity, authority, or even the relevant category boundaries. They were tried by agents operating within a rival grammar – one that had already classified the defendants’ framework not as a competing ontology, but as criminal pathology.

The Nazi grammar was effectively annulled. Not refuted, not outargued – annulled. And as with more typical civil and criminal courts, symmetry was never the goal. The institution ruled by fiat. I call this ontological imperialism in a yet unpublished manuscript. The dominant system merely declares the adversarial grammar invalid and inadmissible.

The standard legitimation story for Nuremberg is natural law: there exist moral facts so fundamental that they transcend positive law and sovereign authority. ‘Crimes against humanity’ was coined precisely to name offences no ontological framework could render legitimate. The phrase does the work – against humanity, not against a particular legal code or polity, but against the species as such. It presupposes exactly the universal semantic accessibility that the philosophy of language has shown to be unavailable.

Man in the High Castle

Now switch venues to a fictional universe. Philip K. Dick asks what would have happened had the Axis won the Second World War. The answer, structurally speaking, is: practically nothing — except that a different ontological grammar would now be dominant.

That is the value of the thought experiment. It doesn’t change the species, the cognitive architecture, or the capacity for deliberation. It changes the constitutive act – the moment at which a grammar gets installed as the world’s grammar. And everything downstream shifts with it. In Dick’s world, the inhabitants don’t experience their moral order as imposed or artificial. They navigate it as the background of intelligibility, the way things simply are. The I Ching functions for Tagomi the way human rights discourse functions for a postwar liberal – not as a choice, but as the grammar within which choices become possible.

The counterfactual is devastating because it is structurally symmetric. Had the Axis won, there would have been trials. Those trials would have applied retroactive categories – perhaps ‘crimes against racial destiny’ or ‘crimes against civilisational hygiene’. Allied leaders would have been the defendants. And the verdicts would have felt, to the inhabitants of that world, exactly as self-evidently correct as Nuremberg’s feel to us.

I don’t secretly wish the Axis had won. But the dialectic is worth consideration, and the discomfort it produces is itself the datum. Not evidence that the examination is wrong – evidence that the grammar is working.

So when modern institutions speak reverently of ‘a jury of one’s peers’, I hear not a triumph of fairness but a legitimating fiction. The phrase conceals the fact that institutions do not seek the defendant’s peers. They seek their own. They seek judges formed within the same order, obedient to the same grammar, and willing to mistake its categories for universal reason.

A peer, in any meaningful sense, would have to share enough ontological grammar with me that the same things register as real, salient, and intelligible in roughly the same way. By that standard, peers are rare. Institutions know this perfectly well. Which is why they do not look for them.

They appoint their own and call the matter settled.

</soapbox>

The Demise of Frege–Geach?

Posted in philosophyTagged , , , , , , , , , 2 Comments
4–5 minutes

Journal Entry

I published an essay on the Frege–Geach problem in February. I published an update yesterday. I still wasn’t satisfied, so I engaged with several LLMs. This was my approach.

The involved LLMs were:

  • Claude
  • Grok
  • ChatGPT
  • Gemini
Audio: NotebookLM summary podcast of this topic.
(This summary misses the mark in some ways, but it brings up some interesting observations along the way.)

First, I fed them some documents in no particular order, my goal being to share my own knowledge and position on the purported problem.

I started with Gemini. This was my prompt:

I am interested in resolving the Frege–Geach problem, but it seems I can only dissolve it. This doesn’t appear to be adequate for some analytical philosophers. How might I get closer to resolving it? My main argument is that they are assuming that language is stronger than it is, and they don’t agree with my argument.

As the prompt notes and by design, many analytical philosophers are reluctant to grant the degree of insufficiency I take to be constitutive of natural language, especially where logical embedding is concerned. Evidently, that counts as my not wanting to play their game. From my perspective, they are committed to a different ontological grammar. What this means practically is that I need to present my solution proposal in their terms. This doesn’t mean their terms are right; problems are only relevant in their dialect, even though my argument is that all dialects are lossy – mine included.

Part of the challenge is that formal logic was invented precisely because ordinary language is imprecise, yet its standards are often retrofitted back onto natural language as though they revealed what language must have been doing all along.

Without sharing the entire play-by-play of the transcripts, I established my course of action. I’d dissolved the problem, but I hadn’t yet resolved it.

My initial intuition of several years ago was to argue that they were expecting too much from grammar. I’ll use a well-worn example. Follow these statements:

  1. IF ‘Murder is wrong.’
  2. THEN ‘If murder is wrong, then getting your brother to murder is wrong.’
  3. SO ‘Getting your brother to murder is wrong.’

According to them, the embedded ‘murder is wrong‘ doesn’t make sense. Here’s their logic:

According to Ayer, moral statements are simply emotive. When one utters, ‘murder is wrong‘, they are really saying ‘Boo, murder‘ – ‘I don’t like murder‘.

If ‘murder‘ is defined as ‘killing disallowed by the state‘, then murder is wrong might be translated into ‘killing disallowed by the state is wrong’ or ‘what the state declares is wrong is wrong’, but we also know that the state makes many pronouncements, many of which carry no moral weight and others which are counter to expected moral positions – law does not equal moral, and vice versa. Let’s move on and revisit our statements:

  1. IF ‘Boo to Murder is wrong.’
  2. THEN ‘If boo to murder is wrong, then boo to getting your brother to murder is wrong.’
  3. SO ‘Getting your brother to murder is wrong.’

My intuition was that the embedded clause does not perform the same linguistic act as the standalone assertion, even if the lexical material is repeated. We’re committing a category error. More crucially, the category it belongs to doesn’t exist, so it’s unspecified. It needs to be invented.

Although I struggled to find apt nomenclature, I settled on performance-sensitive expressions.

A parallel challenge is that the solution can’t be a simple carve-out for moral language. Whilst I feel that moral language does use its own grammar and semantics, I don’t expect analytical philosophers to accept this assertion, so the solution should be more generalisable. I’d need to demonstrate where else this conditional logic fails in the same manner.

ChatGPT had this to say in response to a draft:

The comparative-cost section is good, but one sentence should be added to pre-empt the “your account also uses theory-laden notions” objection.
A critic may reply: your own terms, like “coherence zone” and “synchronisation protocol,” are also theoretical machinery. True enough. The difference is that your machinery is independently motivated and not introduced solely to patch Frege-Geach. You imply this already, but it would help to say so directly.

It offers clearer language:

Embedded moral predicates are not semantically identical to their unembedded counterparts, but neither are they inert; they are performance-sensitive expressions whose full evaluative load is attenuated under embedding while a thinner inferential profile remains available for reasoning.

In any case, I am still polishing the essay, dotting Is and crossing Ts. I think I’ve got the main argument and some examples. One of my weaknesses may be that I rely heavily on my own theories, but these are published and debatable on their own merits.

A real challenge of some problems and paradoxes is that they are self-inflicted. Faith that language is truth apt and lossless leads down many blind alleys.

100% Human Content

Posted in philosophy, podcastTagged , , , , , , , , , , , , , Leave a comment
3–4 minutes

Perhaps not 100% because I’ve just spent hours chatting with several LLMs, complaining about the spate of purported AI detectors that tell me ‘this content shows a high similarity to AI-generated content’ or somewhat.

Audio: NotebookLM summary podcast of this topic.

If I weren’t already familiar with the AI tells, I am now, as Claude reluctantly shared this:


Lexical tics. The ones you’ve already flagged are real: “delve,” “robust,” “nuanced,” “multifaceted,” “landscape” (used metaphorically for any domain whatsoever), “tapestry,” “leverage” (as verb), “facilitate,” “underscores.” There’s also a tier of slightly subtler ones — “notably,” “crucially,” “it’s worth noting that,” “this is particularly important because” — which function as false-emphasis markers, lending rhetorical weight where the content doesn’t actually earn it.

Structural patterns. This is where it gets more diagnostic than any single word choice. LLM prose tends toward a very particular architecture: a thesis-restating opening sentence, three to five parallel elaboration points (often with suspiciously balanced treatment), and a synthesising conclusion that circles back to the opening. The ratio of setup-to-substance is high. Everything gets contextualised before it gets said. There’s a compulsive both-sides-ism, a hedging reflex — “while X is certainly true, it’s also important to consider Y” — that reads less like genuine intellectual caution and more like a conflict-avoidance subroutine.

Tonal tells. LLM text tends to be relentlessly encouraging. There’s an affective flatness masquerading as warmth — everything is “great question,” everything “raises important considerations.” The prose never gets genuinely irritable, confused, bored, or reckless. It maintains a kind of supervisory friendliness, the register of a middle manager who’s been on a coaching course. Relatedly, there’s almost never a genuinely surprising analogy or a metaphor that risks falling flat — the figurative language is always safe, always legible, always serviceable.

Punctuation and rhythm. The em-dash point is real but often overstated. More telling is sentence-length uniformity — LLM prose tends to hover in a narrow band of 15–25 words per sentence, rarely producing either the genuinely terse or the genuinely complex-subordinated. Paragraph length is also strikingly regular. The prose breathes at metronomic intervals.

The absence of mess. Perhaps the deepest tell. Human writing — even polished human writing — contains asymmetries: a paragraph that runs long because the writer got interested, a point that gets dropped without resolution, a transition that’s slightly rough, a joke that doesn’t quite land but stays in anyway. LLM prose is clean in a way that signals not craft but indifference — nothing is there because the writer cared about it, so nothing is disproportionate.

Add to this grammar-checkers attempting to drive writing back to middle school formats – not to mention competing grammar checkers hypercorrecting each other. The struggle is real.

NotebookLM Infographic on this topic.

Of course, a solution might be just to review the content and undo these inclusions. Almost interestingly, I use Autocrit (but not often because it is too orthodox), but it does serve somewhat as an anti-AI agent, assessing rhythm and pace, plus the usual copyediting functions.

I prefer to draft in a stream-of-consciousness style. The AI organises my messes, so if I asked it to, it would remove or relocate my functional parenthetical about Autocrit. But I’m leaving it just to prove I’m human. Or did I add it to an AI-scripted piece? 🧐

Whilst I considered that I could either overdo AI or join the 54 per cent of Americans who write at a third-grade level. Grok suggested something even more sinister – Friggin Musk. It suggested that I double down on the AI likeness and make my content into an AI parody factory – overpopulate it with em-dashes, deving, and tapestry. Evidently, Carole King was AI before Suno.

In any case – and AI might suggest moving this to the top – the problem is that I now have an additional layer that interrupts my flow and process. It’s disconcerting, and I resent it. My psyche is disturbed to appease witchhunters. And it’s bollox.

The question is whether to succumb to the moral suasion or ignore the moral posturing.

This post contains no sugar, salt, fat, carbohydrates, protein, or fibre. No animals were harmed in the production of this blog. All proceeds will be donated to the Unicorn Recovery Foundation.

I Am a Language Model.

Posted in philosophyTagged , , , , , , , , , , Leave a comment
1–2 minutes

I was writing a video script, and a warning popped up. I first noticed this warning a couple days ago, but I didn’t think anything of it.

‘Strong resemblance to AI text.’

I am creating some videos to promote my latest book. I realise that it is academically dense, so I want to break down some of the subject matter.

As usual, my writing is on language and its insufficiencies.

But why was I being warned that my writing resembled AI? What does that even mean?

I reread the sentence:

What if the biggest trick language played on you is convincing you that the world is made of things?

I was looking for a hook to open the short clip. I altered it slightly:

What if the biggest trick language played on you is convincing you that the world is made of things, of objects?

This appeased the AI detector. I had to dilute the message by adding ‘of objects‘. As I write this, it reminds me of Margaret Atwood’s A Handmaid’s Tale and the naming of the women as property of their masters – OfFred, and so on. Of objects.

The AI police are annoying to say the least. Profiling: Minority Project. A 1984 thought crime.

I’m hopping down off the soapbox, down off my high horse, but I’m miffed by bollocks.

How does AI summarise it? Find out here:

Audio: NotebookLM summary podcast of this topic.

EDIT: Does it get worse?

When I checked into LinkedIn to share this post, I was distracted by another thread chatting about Emotivism. I’ll spare you the entire thread, but now Grammarly wants me to write in German. Was ist los?

Truth, Subjectivity, and Constraint

Posted in philosophyTagged , , , , , , , , , , , , , , , , , , , Leave a comment
3–5 minutes

I like this bloke. Here, he clarifies Rorty’s perspective on Truth. I am quite in sync with Rorty’s position, perhaps 90-odd per cent.

Allow me to explain.

Audio: NotebookLM summary podcast of this topic.

I have written about truth several times over the years, 1, 2, 3, and more. In earlier posts, I put the point rather bluntly: truth is largely rhetorical. I still think that captured something important, but it now feels incomplete. With the development of my Mediated Encounter Ontology of the World (MEOW) and the Language Insufficiency Hypothesis (LIH), the picture needs tightening.

NotebookLM Infographic on this topic.

The first step is to stop pretending that ‘truth’ names a single thing.

Philosopher Bernard Williams helpfully distinguished between thin and thick senses of truth in Truth and Truthfulness. The distinction is simple but instructive.

In its thin sense, truth is almost trivial. Saying ‘it is true that p’ typically adds nothing beyond asserting p. The word ‘true’ functions as a logical convenience: it allows endorsement, disquotation, and generalisation. Philosophically speaking, this version of truth carries very little metaphysical weight. Most arguments about truth, however, are not about this thin sense.

In practice, truth usually appears in a thicker social sense. Here, truth is embedded in practices of inquiry and communication. Communities develop norms around sincerity, accuracy, testimony, and credibility. These norms help stabilise claims so that people can coordinate action and share information.

At this level, truth becomes something like a social achievement. A statement counts as ‘true’ when it can be defended, circulated, reinforced, and relied upon within a shared framework of interpretation. Evidence matters, but so do rhetoric, persuasion, institutional authority, and the distribution of power. This is the sense in which truth is rhetorical, but rhetoric is not sovereign.

NotebookLM Infographic on this topic. I prompted NotebookLM to illustrate a 4-layered model that shows how removed language is from encounter, attention, conception, and representation of what we normally consider to be reality. This view is supported by both MEOW and LIH.

Human beings can imagine almost anything about the world, yet the world has a stubborn habit of refusing certain descriptions. Gravity does not yield to persuasion. A bridge designed according to fashionable rhetoric rather than sound engineering will collapse regardless of how compelling its advocates may have been.

This constraint does not disappear in socially constructed domains. Institutions, identities, norms, and laws are historically contingent and rhetorically stabilised, but they remain embedded within material, biological, and ecological conditions. A social fiction can persist for decades or centuries, but eventually it encounters pressures that force revision.

Subjectivity, therefore, doesn’t imply that ‘anything goes’. It simply means that all human knowledge is mediated.

We encounter the world through perception, language, culture, and conceptual frameworks. Every description is produced from a particular standpoint, using particular tools, within particular historical circumstances. Language compresses experience and inevitably loses information along the way. No statement captures reality without distortion. This is the basic insight behind the Language Insufficiency Hypothesis.

At the same time, our descriptions remain answerable to the constraints of the world we inhabit. Some descriptions survive repeated encounters better than others.

In domains where empirical constraint is strong – engineering, physics, medicine – bad descriptions fail quickly. In domains where constraint is indirect – ethics, politics, identity, aesthetics – multiple interpretations may remain viable for long periods. In such cases, rhetoric, institutional authority, and power often function as tie-breakers, stabilising one interpretation over others so that societies can coordinate their activities. These settlements are rarely permanent.

What appears to be truth in one era may dissolve in another. Concepts drift. Institutions evolve. Technologies reshape the landscape of possibility. Claims that once seemed self-evident may later appear parochial or incoherent.

In this sense, many truths in human affairs are best understood as temporally successful settlements under constraint.

Even the most stable arrangements remain vulnerable to change because the conditions that sustain them are constantly shifting. Agents change. Environments change. Expectations change. The very success of a social order often generates the tensions that undermine it. Change, in other words, is the only persistence.

The mistake of traditional realism is to imagine truth as a mirror of reality – an unmediated correspondence between statement and world. The mistake of crude relativism is to imagine that language and power can shape reality without limit. Both positions misunderstand the situation.

We do not possess a final language that captures reality exactly as it is. But neither are we free to describe the world however we please. Truth is not revelation, and it is not mere invention.

It is the provisional stabilisation of claims within mediated encounter, negotiated through language, rhetoric, and institutions, and continually tested against a world that never fully yields to our descriptions. We don’t discover Truth with a capital T. We negotiate survivable descriptions under pressure.