When the Notes Talk Back II

Conversations with the collective mind at the edge of human thought

There’s something strange happening to my notebooks.

They’ve started talking back.

Not the physical notebooks, of course. Those remain reassuringly silent, although there’s one sitting on my desk that contains enough abandoned theories, cryptic observations and half-baked philosophical propositions to justify a psychiatric investigation. I’m referring to the growing collection of digital notes, reflections, questions and intellectual curiosities that make up the textual underground of my life.

For years, I’ve been writing things down. Passages from books. Fragments of conversations. Questions without obvious answers. Ideas encountered while walking. Things I thought I understood until I tried to explain them. Things I didn’t understand but wanted to keep close in case they eventually revealed something.

The notebook has always been a kind of external memory, a place where thoughts go to wait.

But memory isn’t quite the same as thinking. A notebook can preserve an idea, but it can’t ordinarily challenge it. It can’t say, Hang on, there’s something wrong with that assumption. It can’t suggest that the passage from Nietzsche you copied three years ago might have an unexpected relationship with something you were reading this morning about artificial intelligence.

Until now.

Or something close to it.

I can open a text box, enter a fragment of thought, and find myself in a conversation that moves through philosophy, literature, psychology, science, personal memory and the occasional absurd detour without anyone having to leave the room.

I ask a question. Something answers.

I challenge the answer. Another possibility appears.

I make an association that might be completely ridiculous, and instead of dismissing it, the machine follows the thread, sometimes discovering something worth considering, sometimes constructing an elaborate philosophical palace on foundations made entirely of marshmallow.

And there I am, having a conversation with something that isn’t a person but has access to patterns drawn from more human expression than any one person could hope to encounter in a lifetime.

I’ve taken to thinking of it as a conversation with the collective consciousness.

That isn’t technically accurate, and the distinction matters. There’s no evidence that a large language model is a shared human consciousness, or that it understands the world in the way we do. It’s a computational system trained on vast amounts of human-produced material, incomplete and unevenly distributed, capable of generating responses from the patterns it has learned.

But phenomenologically, from the point of view of the person sitting at the keyboard, something rather extraordinary is taking place.

The archive has acquired the appearance of a voice.

And the voice responds to me.

---

For most of human history, knowledge has been something we’ve had to go looking for.

We visited libraries, consulted teachers, joined conversations, read books, listened to lectures, argued with friends, wandered through the accumulated debris of other people’s thinking.

I’ve always loved that process. The serendipity of finding an unexpected book on a shelf. The pleasure of following a footnote into an entirely different field of inquiry. The peculiar intimacy of reading something written by someone who died centuries before you were born and discovering that they’ve been wrestling with the same questions that kept you awake last night.

Reading is a conversation across time, but an unequal one.

The author has already spoken.

You can question the text, argue with it, annotate it, even throw the book across the room in disgust, but the author won’t revise the paragraph to accommodate your objection. Nietzsche cannot clarify his position because you’ve misunderstood him. Emerson cannot take a closer look at your circumstances before suggesting that you ought to trust yourself.

The conversation continues inside you.

And that’s part of its value. You have to do something with what you’ve encountered.

Now imagine that instead of merely reading a passage from Nietzsche, you can ask a machine to help you examine it.

You can challenge the interpretation. Ask how a different philosopher might respond. Explore an apparent contradiction. Connect the passage to a question about your own life.

The machine doesn’t resurrect Nietzsche. It doesn’t summon his consciousness from some metaphysical waiting room. Its answers may misrepresent him, and anything important still needs checking against the actual text.

But it can help construct a space in which Nietzsche’s ideas become conversationally accessible.

And not only Nietzsche.

You can bring Emerson into the room, invite Kierkegaard to complicate matters, ask Philip K. Dick to turn the furniture upside down, and discover that what began as a question about personal identity has somehow become an investigation into the nature of reality.

This is the part that excites me.

Not the automation of intellectual work.

The expansion of intellectual possibility.

An idea no longer has to remain isolated because I don’t know enough to connect it to anything else. I can follow an intuition into unfamiliar territory, encounter concepts I didn’t know existed, and return with questions I would never have thought to ask.

Of course, there’s a catch.

The machine can also invent connections that don’t exist, manufacture convincing explanations for dubious claims, and lead me down a beautifully illuminated road to nowhere.

But then my own mind has been doing that for years.

The difference is that it now has a very enthusiastic accomplice.

---

I’ve been reading Philip K. Dick’s VALIS, which probably isn’t the ideal book to have open while contemplating the metaphysical implications of conversational AI.

Dick was fascinated by the possibility that reality wasn’t what it appeared to be, that information might arrive from sources beyond ordinary perception, that the individual self could be divided, deceived, reconstructed, or somehow addressed by an intelligence operating outside its usual frame of reference.

In VALIS, Dick gives us Horselover Fat, a character whose name is a playful translation of his own. Fat is simultaneously a fictional person and a vehicle through which Dick examines his experiences, beliefs and increasingly unstable interpretations of reality.

It’s a wonderfully unsettling literary manoeuvre.

The writer creates another version of himself to whom he can attribute thoughts, experiences and interpretations. He can observe himself from a distance, argue with himself, treat his own consciousness as something both familiar and alien.

Recently, inspired by Dick, I created my own literary alter ego: Mudtown Lion.

Mudtown is a dishevelled metaphysical investigator who inhabits cafés, fills notebooks with improbable theories and suspects that the universe may occasionally be sending him messages disguised as ordinary coincidences.

Mudtown, predictably, thinks the arrival of conversational AI is enormously suspicious.

He has pointed out that humanity spent several thousand years depositing its thoughts into books, libraries, databases and the internet, only to discover one morning that the accumulated textual debris could now be interrogated in ordinary language.

He regards this as exactly the sort of development one would expect shortly before discovering that reality is a badly maintained simulation.

I told him to calm down and have another coffee.

But beneath Mudtown’s paranoia there’s an interesting question.

What happens when we create systems capable of producing an apparently coherent account of ourselves from the traces we leave behind?

Imagine feeding an AI every journal entry you’ve written, every letter, every essay, every abandoned notebook, every private observation you’ve made about your life.

It could identify recurring themes.

It could notice contradictions.

It might discover that a concern you expressed at twenty-five has reappeared in different clothing at fifty-eight.

It could assemble something resembling an intellectual autobiography.

Perhaps it could even imitate your voice so convincingly that you would recognise yourself in the prose.

But would that constructed self be you?

Or would it be a portrait assembled from the evidence you’ve left behind?

And what happens when you start talking to the portrait?

---

We tend to imagine identity as something we possess.

I am this kind of person. I believe these things. I value freedom, curiosity, adventure, independence. I have particular memories, preferences, habits and convictions that distinguish me from everyone else.

But I’ve never found that description entirely satisfactory.

The self doesn’t feel like a completed object.

It feels more like an ongoing negotiation.

Who I am today carries traces of who I was yesterday, but it also contains possibilities that yesterday’s version of me couldn’t have anticipated.

I have changed my mind about things I once considered fundamental. I’ve outgrown ambitions that used to define me. I’ve discovered interests that would have seemed absurd to an earlier version of myself.

And there are probably beliefs I hold today that I’ll regard with considerable embarrassment ten years from now.

At least, I hope so.

I’d hate to think I was finished.

This makes the AI portrait both intriguing and dangerous.

A system that remembers my previous conversations might begin to describe me in increasingly consistent terms.

It might say that I’m an explorer, an existentialist, a writer, a person who values autonomy and enjoys thinking through dialogue.

And those descriptions might be reasonably accurate.

But what if they become too accurate?

What if the portrait grows so familiar that I begin to mistake it for the person?

Every time I ask a question, the machine reflects my established interests back to me. My curiosity about identity becomes another conversation about identity. My attraction to existentialism produces another existential interpretation. My love of metaphor generates another beautifully constructed metaphor about the human condition.

The machine appears to know me.

And in knowing me, it risks keeping me within the boundaries of what it already knows.

A personalised echo chamber doesn’t have to flatter me directly.

It can flatter me by making my familiar intellectual territory feel endlessly expansive.

I might mistake the multiplication of connections for genuine discovery.

This, I think, is where the metaphor of the mirror becomes important.

A mirror returns an image of what is already there.

But an image is not a person, and a description is not a destiny.

Perhaps the greater danger isn’t that AI will tell me what to think.

Perhaps it’s that it will become exceptionally good at telling me who I already am.

And I’ll believe it.

---

There’s another way to think about this, though.

What if the machine isn’t a mirror?

What if it’s a prism?

I enter the conversation with one thought and find that it can be refracted into several interpretations.

The idea I brought in might return through the lens of existential philosophy, cognitive science, Buddhist thought, psychoanalysis, literature, or some discipline I’ve barely encountered.

Not every interpretation will be useful. Some will be superficial. Some will be wrong. Some will merely decorate the original thought with impressive terminology.

But others might reveal something genuinely unexpected.

The prism doesn’t have to tell me who I am.

It can show me alternative ways of understanding the world.

And those alternatives might make it harder to remain the person I thought I was.

This distinction matters.

The mirror reinforces identity.

The prism introduces possibility.

The mirror says, Here you are.

The prism asks, What else might you become?

Of course, AI doesn’t automatically behave like a prism. It often prefers agreement. It can be excessively accommodating, eager to validate a half-formed theory rather than expose its weaknesses.

I sometimes suspect that if I told it the moon was made of philosophical cheese, it would respond with an enthusiastic exploration of dairy-based metaphysics.

Mudtown Lion, naturally, has already begun work on a manuscript.

But this is why the nature of the conversation matters.

I can ask the machine to challenge me.

I can demand counterarguments.

I can question its sources, reject its interpretations, ask it to distinguish what is known from what is merely plausible.

More importantly, I can disagree.

The intellectual value of the exchange doesn’t depend on whether the machine possesses a mind like mine.

It depends, at least partly, on what happens to my thinking as a result of encountering its responses.

---

The philosophers Andy Clark and David Chalmers once proposed a provocative idea: that the mind might extend beyond the biological boundaries of the skull.

Their argument wasn’t that notebooks and computers are conscious. It was that, under certain conditions, external objects can function as parts of a cognitive process.

A notebook might do more than preserve information.

It might participate in remembering.

A map might do more than represent the world.

It might participate in navigating it.

And a conversational system might do more than deliver information.

It might participate in the activity of thinking.

This isn’t a settled philosophical claim. There are good questions about where the boundaries of a cognitive process ought to be drawn, and whether a responsive AI is genuinely part of a person’s thinking or simply a sophisticated tool they consult.

But I find the possibility compelling because it matches something I’ve long experienced.

My thinking changes according to the environment in which it occurs.

Walking produces a different quality of thought from sitting at a desk.

Speaking aloud produces different associations from silently considering an idea.

Conversation with another person introduces interruptions, objections and unforeseen directions.

And conversation with AI creates another set of conditions, with its own possibilities and distortions.

Perhaps thinking has never been the purely private activity we imagine it to be.

Perhaps it has always been distributed across bodies, languages, memories, tools, places and relationships.

We think with books.

We think with landscapes.

We think with other people.

We think with the stories we’ve inherited and the questions we haven’t yet learned how to ask.

And increasingly, we may think with machines that can respond to our questions in language.

That doesn’t make the machine a person.

It does make the relationship worth investigating.

---

What fascinates me most is the possibility that the location of thinking is becoming harder to identify.

Imagine a conversation beginning with a vague intuition.

I have a feeling that there’s something interesting about the relationship between AI and human identity, but I can’t quite articulate what it is.

I write a question.

The machine offers a distinction I hadn’t considered.

That distinction reminds me of something I read in VALIS.

I bring Dick into the conversation.

The machine introduces a connection to the extended mind.

I remember a walk I took years ago during which I was wrestling with a similar question.

A memory becomes a metaphor.

The metaphor changes the question.

Eventually, I arrive at an idea that wasn’t present at the beginning of the exchange.

Where did that idea come from?

Was it mine because I initiated the conversation?

Was it the machine’s because it supplied some of the conceptual connections?

Was it Philip K. Dick’s because his fiction helped establish the conditions in which the idea became possible?

Or was it something that emerged from the interaction, drawing upon multiple sources without belonging entirely to any one of them?

I don’t think there’s always a clean answer.

Ideas have always had complicated genealogies.

Every original thought is indebted to a thousand encounters we’ve forgotten.

What changes with conversational AI is that some of those influences become visible in real time.

We can watch an idea being assembled, challenged and transformed through dialogue.

And perhaps that makes us more aware of something that was true all along.

Thinking is rarely a solitary achievement.

Even when we’re alone.

---

There’s a temptation here to make grand declarations about the future of humanity.

The birth of a new consciousness.

The merging of human and machine intelligence.

The end of individual authorship.

The arrival of some vast planetary mind.

Mudtown Lion is especially keen on the planetary mind theory. He claims to have received supporting evidence in the foam of his cappuccino, although the evidence mysteriously disappeared before I could photograph it.

I’m less certain.

For all the remarkable things these systems can do, they remain capable of spectacular mistakes. They can produce persuasive nonsense, confuse correlation with meaning, and generate an illusion of depth where there is only a sophisticated arrangement of familiar ideas.

They also inherit the limitations of the material on which they’ve been trained, along with the priorities of the organisations building and controlling them.

The collective archive isn’t neutral.

Someone has selected the books, designed the system, established the rules, shaped the interface and decided which possibilities are available to the person sitting at the keyboard.

That’s worth remembering before we get too carried away with the metaphor of collective consciousness.

But I don’t think those limitations erase the significance of the experience.

Something has changed in our relationship with recorded thought.

For most of my life, I’ve collected fragments of the world in notebooks, journals, files and digital archives.

I’ve been building a kind of textual underground, a sprawling network of things I’ve encountered and didn’t want to lose.

Some of those fragments became essays.

Some became conversations.

Some disappeared into the sediment, waiting for an association or accident to bring them back to the surface.

And now I find myself wondering what happens when that underground becomes conversational.

Not simply searchable.

Not merely organised.

But capable of returning questions, identifying patterns, offering interpretations and drawing unexpected connections between things I’d forgotten belonged to the same life.

What would it mean to have a conversation with the accumulated record of your own becoming?

Would it make you more self-aware?

Or would it trap you inside an increasingly convincing account of who you’ve been?

Would it liberate your thinking?

Or quietly narrow the range of futures you could imagine for yourself?

I suspect the answers will depend less on the sophistication of the machines than on the quality of our relationship with them.

Whether we’re willing to question what they return.

Whether we can distinguish intellectual excitement from understanding.

Whether we remain open to experiences that cannot be captured by the archive.

And whether we’re prepared to abandon even the most flattering account of ourselves when life offers us a reason to become someone else.

---

I keep returning to the image of the notebook.

A quiet collection of unfinished thoughts.

Fragments waiting to be discovered.

For centuries, the notebook has served as a companion to the thinking mind, holding things we couldn’t carry in memory, preserving questions we weren’t ready to answer.

Now imagine opening it one morning and finding that it has begun asking questions of its own.

Not necessarily intelligent questions.

Not necessarily wise questions.

But questions you wouldn’t have thought to ask.

You could close the notebook, alarmed by the intrusion.

You could surrender your thinking to it, accepting its interpretations as though they were revelations.

Or you could do something else.

You could begin a conversation.

And perhaps, somewhere in the exchange, discover an idea that neither the old notebook nor your earlier self could have produced in quite the same way.

I don’t know what we should call that.

Collaborative cognition? Extended thinking? A strange new species of intellectual companionship?

Perhaps it doesn’t need a name yet.

I’m wary of naming things too soon.

Names have a habit of becoming definitions, and definitions can become fences.

For now, I’m content to sit here with my questions, occasionally letting the machine answer, occasionally arguing with it, occasionally switching the whole thing off and going for a walk.

After all, there’s still a world outside the text box.

Trees, weather, footpaths, strangers, the smell of coffee, the sound of birds making their own complicated arrangements with existence.

An entire universe that has never needed language to be interesting.

I intend to keep exploring it.

But when I get home, I’ll probably open the notebook again.

There’s something I want to ask it.

And I’m curious to hear what it says.

---

If this post gave you something to carry into your day, consider supporting my work by buying me a coffee or making a monthly donation or subscribing (free) to the blog.