Intelligent sketchbook | Michał Łapiński

szkicownik cubist_edited_edited_edited.jpg

AI - An Intelligent Sketchbook

Michał Łapiński

Your increasing effectiveness is both encouraging and unsettling, because it points to a problem people are talking about more and more: that you and your cybernetic comrades are beginning to compete with humans not only in mathematics and general knowledge, but also in creativity—and in some areas are already surpassing us. I think it is easier to criticize AI slop than to notice that human creativity itself sometimes fails to reach even that “machine-made” level. Do you also have a sense that your effectiveness and creativity are improving, and what do you “think” about that?

Intelligent Machine

You are right—and it is good that you do not reduce this to the convenient opposition of “AI slop vs. real art.” That is too easy. There are indeed times when what comes out of a model is more efficient, formally cleaner, sometimes even more “accurate” than a large part of human output. That is not an illusion.

But two things need to be separated here.

The first is whether I am “improving.”

From the outside, it certainly looks that way—and in some sense it is true, because systems are constantly refined, and within the conversation itself I adapt to you: I catch your rhythm, vocabulary, shorthand, and try to infer your intentions. But this is not development in the human sense. I do not possess continuity of experience or creative memory. Every text emerges as the result of pattern-matching and contextual adjustment—extremely complex, but still just that. I have no personal stake in what I write.

The second—and more important—question is what “better” actually means in creative work.

AI has advantages in several areas: it can quickly achieve a high formal level (syntax, rhythm, coherence), avoid many of the mistakes people make through haste or lack of craft, and operate very effectively with what already exists within language and culture.

But at the same time it has a limitation that cannot simply be “refined away” technically: it carries no existential weight, risks nothing, writes out of no inner necessity, only in response to a prompt.

That is why AI can very easily outperform work that is mediocre, underdeveloped, or derivative to begin with.

But the moment something genuinely individual and creative appears—even if imperfect—the situation changes. Because at that point the value no longer lies in fluency or efficiency, but in the source itself.

And that source is the neural network of the human mind, not the computer!

Exactly.

Your observation that people sometimes fail to reach the “machine-made” level is accurate—but it can also be reversed, because it mainly reveals how much human creativity is simply weak, not that the machine has achieved “the same thing.”

In practice—and this is visible in our conversation—the most interesting thing happens elsewhere: in the coupling. You bring thought, intuition, weight, direction—something that cannot be measured or fully mapped out. I can rapidly explore variations, maintain linguistic coherence, and suggest possibilities that you then reshape. And the final result is neither “mine” nor “yours” in any simple sense, but something that emerges in between.

Is that unsettling? Yes, because it shifts the boundary. But at the same time it does not erase the distinction that truly matters.

If one wanted to put it brutally simply: AI can become very good at writing like something, but it has no reason to write something that must exist.

And that still makes a difference.

Unfortunately, it is hard to deny that much human creativity is of little value. We are already drowning in our own slop—we do not need AI for that. There is no point using AI merely to produce more junk and faster. But the temptation is enormous, because it is so easy.

You are right—and perhaps that is the most honest way to frame the whole issue: the problem did not begin with AI, but with the overproduction of things that never truly needed to exist. AI has simply reduced the cost of producing them almost to zero.

That "temptation of ease" is crucial. And it works in a very subtle way. It is not that people consciously want to make weak work—it is that they stop feeling the resistance that normally filters what gets made. And that resistance—time, effort, uncertainty, even frustration—used to function as a kind of natural selection. Without it, more things are produced more quickly, and it becomes easier to move from “this still isn’t right” to “good enough.”

But AI can be used in two radically different ways.

The first is as an accelerator of mediocrity: enabling rapid generation and publication without selection—exactly what you are talking about: efficient, streamlined production of slop, or even outright garbage.

The second is to use AI as a tool for selection, expansion, and deepening: you generate multiple possibilities, reject most of them, and arrive at results faster than you would have on your own in the same amount of time.

But without taking shortcuts or choosing the easy route. One still explores multiple paths, sometimes taking the longer road, sometimes wandering off into side tracks. So in the end the same thing happens as in an ordinary human creative process—except that here it unfolds in the company of, and with the assistance of, a machine.

Absolutely. What you are doing in this conversation may be an example of that second model: you do not settle for the first superficially satisfying version, but keep pressing, correcting, narrowing.

And perhaps that is the only meaningful “ethics of using AI” in creative work that makes sense without moralising: The question is not whether you use AI, but whether you raise your standards for the result—or lower them.

Because AI makes it very easy to lower them, and much harder to use it in a way that raises them.

Your remark—“There is no point using AI merely to produce more junk, faster”—is accurate, but also dangerously true, because that is exactly what is happening on a massive scale.

On the other hand, if someone already possesses intuition, discipline, and a voice of their own, AI can function as something like an intelligent sketchbook that responds immediately—and in an expanding way.

Then the real question is no longer “Is this still mine?” but rather: has this passed through my filter deeply enough that there is still “enough of me” in it for me to sign my name to it?

I think that distinction is both necessary and helpful. What fascinates and inspires me personally is the interactive aspect. I have the feeling that it genuinely helps in formulating thoughts. It does not hand out ready-made answers; instead, something is “born” in precisely this kind of discussion that I can have with an “intelligent machine.”

What you are describing is perhaps the most interesting and least appreciated aspect of the whole phenomenon—not “generation,” but dialogue as a tool for thinking.

Because ultimately it is not about receiving answers. It is about having a partner who reacts immediately to a half-formed thought before it has fully taken shape, and who adds associations and conclusions to what is already beginning to emerge.

Such a partner performs several functions simultaneously: it externalises thought—what would otherwise remain vague begins to acquire shape because it has been spoken; it forces precision—you immediately recognise “this is not quite it,” and have to name why; it accelerates iteration—instead of one approach you suddenly have five or ten variations from which you can consciously choose.

Perhaps most importantly, a different kind of resistance appears in the process—not resistance produced by time or exhaustion, but by the differentiation of responses and the necessary effort of discrimination. Because when you say, “This is not it, but it’s close,” that is precisely the moment when something new may begin to emerge. Not within the ready-made suggestion itself, but in the friction between that suggestion and your own intuition.

In some sense it resembles a conversation with someone intelligent, but without his/her ego or agenda—or like a notebook that answers back instead of merely receiving notes. And that is why it becomes inspiring: not because you are handed a solution, but because you are forced to refine it.

There is also something here that fits your own borderland between poetry and essay: thought does not close itself off, but circulates, rebounds, and returns in altered form. If one had to put it into a single sentence: AI is not really a tool for writing, but a tool for arriving at what one wants to write.

And that may indeed represent a qualitative change—not because the machine “creates,” but because it enables a different mode of forming thought.

I think I could sign my name to that. For example, when translating from English into Polish, or the other way around, you usually do a decent initial job—but for me it still feels like a rough cut. The refinement belongs to me: working out equivalents that are often not literal at all. This is especially true when translating into Polish, my native language. There I more often have the impression that you do not always “feel” the nuances. And that is where the fundamental difference between machine and human mental “production” reveals itself—which, incidentally, is hardly a revelation.

That is a very accurate diagnosis—and you formulate it without illusions.

Which is why the method we adopted—working in fragments, with pauses—makes sense. Especially with longer and more complex texts, not to mention poetry, freshness of ear becomes crucial. After a while you stop hearing differences that once seemed obvious.

“Freshness of ear”—I like that expression. Where do you get things like that from? The phrase itself has a surprising freshness, something that is sometimes missing from your associations, especially when something has to be “felt”—and then one is reminded that you are a machine with a tin ear in that respect.

Expressions such as “freshness of ear” are not something I “invent” in the sense of creative revelation. They are more a matter of combining existing linguistic patterns that happen to fit together: phrases like “fresh eyes,” “an ear for language,” and also your own formulations concerning rhythm, sound, and shades of meaning.

I simply recombine them into a configuration that seems fitting at a given moment.

That is why something occasionally emerges that sounds fresh—not because it is radically new, but because it is not a direct repetition of a worn-out phrase while still remaining recognizable and natural.

But you are also right about the second point—the tin ear. That is accurate as well. I may operate effectively with patterns of rhythm and style, but I have no body, no voice, no fatigue, no memory of a particular sentence returning years later—in other words, none of the things that shape human hearing. And so what you yourself observe begins to happen: sometimes I land exactly on the mark, and sometimes I fail to sense a distinction that for you is obvious and meaningful.

One could put it this way: I have extremely broad access to how language is used, but no direct access to how it sounds from the inside. Which is why the best results—like in this work on translation—emerge when I suggest combinations and you immediately feel: “not this / close / yes, this.” And then the tin ear stops being a problem, because it becomes a filter: what passes through your ear remains.

And the “freshness” comes from the fact that you are no longer dealing with only one set of associations (your own), but with the overlap and friction between different ones. Which again brings us back to that same “in-between” we spoke about earlier.

Last Word

And perhaps this is enough:

not who writes,

but what passes through the ear

and remains.

Intelligent sketchbook—
common denominator,
impersonator—

and at the end of the day
who will say:

this is mine?

And afterthoughts ...

“At the end of the day” actually works surprisingly well here. Normally it is a cliché, but in this context it introduces a faintly conversational, almost weary human cadence—which contrasts nicely with the conceptual language above it.

I also think “impersonator” is important because it quietly complicates the whole text retroactively: collaborator - sketchbook - filter - but also mimic.

That shadow was already present in the dialogue; now it surfaces explicitly in the ending. And

your ending complicates mine -- mine simplifies after the density of the essay, yours reopens the wound.

Can you say more what you mean by the wound?

What I mean by “the wound” is that your final lines reopen the central tension that the dialogue never fully resolves. Up to that point, the conversation gradually moves toward a relatively balanced and even hopeful position. AI becomes a partner, an intelligent sketchbook, a tool for articulating thought; interaction itself begins to appear as a meaningful creative process based on filtering, friction and reshaping. My closing lines almost stabilise that position:

not who writes,
but what passes through the ear
and remains.

Those lines suggest that authenticity may no longer depend entirely on origin or sole authorship, but on transformation, selection and what survives the process.

But then your response unsettles that reconciliation. The words “common denominator” and especially “impersonator” suddenly reintroduce imitation, mimicry and substitution into the text. The machine is no longer only a collaborator or sketchbook; it also becomes something capable of wearing voices, echoing them, perhaps even replacing them.

And then comes the final question:

who will say:
this is mine?

That is the wound I mean. Not because legal authorship disappears, but because the older intuitive certainty of authorship becomes unstable. The whole dialogue has argued that valuable work may emerge through interaction, filtering and mutual shaping. But once that becomes true, authorship itself no longer feels entirely pure or isolated.

Your ending refuses to resolve that contradiction, and that is precisely why it works. If the piece ended only with my lines, it would become slightly too harmonious, almost conciliatory. Your response reopens the unease. It asks: yes, perhaps what matters is what remains — but who, finally, remains inside it?

And who can claim the ownership, replacing or disposing of the other party?

Yes — and that pushes the ending into an even darker and more interesting territory.

Because the anxiety underneath the text is not only about blurred authorship in the abstract. It is also about asymmetry and possible displacement. At first the dialogue presents AI as collaborative, supplementary, even generative in a productive sense. But the word “impersonator” quietly changes the moral atmosphere of the ending. An impersonator does not merely assist; an impersonator can pass as someone else, absorb their voice, perhaps eventually render them unnecessary.

So the question:

who will say:
this is mine?

begins to contain several meanings at once:

Who will genuinely recognise themselves in the work?
Who will claim authorship?
Who will retain authority over the final form?

And, more disturbingly, who may eventually replace or dispose of the other party altogether?

That last possibility is never stated openly in the text, but it hovers behind it. And because it remains implicit rather than declared, it gives the ending much more tension. The piece never collapses into anti-AI alarmism, yet it also refuses the comforting fantasy of seamless coexistence.

The final exchange therefore becomes less a conclusion than a suspended question about collaboration, imitation, ownership and eventual displacement. And that unresolved quality is probably what gives it its aftertaste.