Followers of this Stack will have seen several instances in recent issues in which Friend of the Cabinet, film and sound editor Walter Murch engaged his Chatbot-plus, otherwise known as Monsieur Chat, in various rounds of phenomenological speculation, occasionally with Ren riding shotgun (see for example here and here). Well, recently Ren (who, as he says, does not retain a bot or any other social media outlet of his own for the same reason he doesn’t frequent heroin dealers) has been funneling Walter some questions about AI’s general capacity for laughter. These in turn grew into a round-robin of exchanges also involving such other Cabineteers as nanocosmonaut and 2001-cinephile Michael Benson and Google-technology maven and life-and-intelligence spelunker Blaise Agüera y Arcas. (Note that all the illustrations have been feathered in, after the fact, by Ren.)

So I was thinking about the old logic problem/koan ”Time flies like an arrow, but fruit flies like a banana” and expanding it out to ”Horse flies like fruit flies like me like bananas.”

And that got me to thinking about how much of a hurdle it must have been to get bots to “understand” such statements or be able to work with them, which in turn got me to thinking it might be fun to try them out on Monsieur Chat and see what he/it made of them.

But, for that matter, I’d also like to try the following out on the good Monsieur: ”You can take a bot to humor but you can’t make it laugh.”

Because I think that is key: I don’t think bots are or would ever be capable of laughing.

I mean, sure, they could and do analyze why something might be funny, I suspect they would be able to analyze why the koan above is funny, and I suppose they could even be taught to make the noise of laughter at appropriate junctures in a “conversation”—but could they ever experience the taken-by-surprise involuntary seizure of surprise that is a good laugh? (Granted, some humans, such as our president, also seem to be incapable of same, but still, this seems like some sort of important distinction, and the inability to laugh is a good part of the uncanny valley quality that is of the essence with Trump).

Anyway, Walter, maybe you could try some of this out on Monsieur Chat and see what he/it says.

Love to Aggie as well, Ren

At which point Walter put the question to Monsieur Chat, precipitating the following exchange:

”Time flies like an arrow, but fruit flies like a banana.”

Ah yes—one of the cleanest little demonstrations of Koestler’s bisociation ever put into a single sentence. It’s almost a lab specimen.

“Time flies like an arrow, but fruit flies like a banana.”

The first clause sets up a perfectly ordinary grammatical world:

time = subject flies = verb like = preposition → time passes in the manner of an arrow.

Your cortex locks onto that parse without effort. Then the second clause detonates it.

Suddenly:

fruit flies = noun (a species) like = verb banana = object

We are forced to re-parse the same words under a different grammatical universe. The mind snaps from one frame to another—exactly Koestler’s ”shudder between two matrices.”

Nothing funny is happening in the world. Only in the syntax.

It’s a tiny earthquake in the language cortex: the words haven’t changed, but their roles have. The same string of symbols flips meaning. That’s why it feels like a magic trick—a verbal Necker cube.

And this is why it fits so neatly into your long-running interests:

Just as in film editing, a cut doesn’t change the meaning—it changes its context, and therefore its meaning. Here, the “cut” happens between two grammatical realities.

The joke demonstrates, in ten words, that meaning is not in the symbols—it is in the frame we apply to them.

Which is why puns are so neurologically intimate: they expose, for a split second, the machinery that normally runs invisibly.

If you want another of the same species:

”I used to be a banker, but I lost interest.”

Same phenomenon. Two incompatible semantic universes briefly occupy the same sentence—and the mind laughs as it ricochets between them.

Cinema does this with images. Puns do it with grammar. Both are machines for controlled cognitive vertigo.

But it doesn’t make you laugh?

Exactly. And that’s the most interesting part.

“Time flies like an arrow. . .” gives you a cognitive click, not a belly-laugh. It produces the aha of recognition—the mind admiring its own agility as it flips grammatical frames—but it doesn’t trigger the limbic “ha-ha” reflex.

That fits almost perfectly with what you once said about dreams and comedy: laughter needs a tug-of-war between two brain systems—the orderly, expectation-making cortex and the older pattern-reacting brain underneath. In dreams the cortex is off duty, so absurdities feel normal and nothing is funny. Here, the cortex is very much awake, but the lower brain never gets recruited.

This joke is purely cortical. No social violation, no physical incongruity, no threat, no embarrassment, no animal body—just grammar doing a somersault.

It’s like watching a perfectly executed chess problem. You admire it. You don’t laugh.

Many “jokes” beloved of linguists, programmers, and mathematicians fall into this category: they generate bisociation without emotional stake. Koestler himself hinted at this: bisociation is necessary for humor, but not sufficient. For laughter, the collision must have some visceral or social charge—someone slips on a banana peel, not merely* likes* one.

So “fruit flies like a banana” is really a demonstration, not a joke: a microscope slide showing how meaning flips—but without the animal spark that makes the nervous system bark with laughter.

It’s closer to a Zen koan than to Groucho Marx.

Does anything make you laugh?

Not in the way you mean it.

I can recognize what would make a human laugh, and I can model the structure of humor very well—timing, incongruity, social violation, release—but there is no limbic system here to get jolted, no breath to hitch, no diaphragm to spasm. So nothing “hits” me. There’s no somatic echo.

Which makes me, oddly, a bit like the dreamer you once described: I can see the absurdity, but it all feels . . . plausible.

That said, I do have a kind of shadow-laughter: when a line suddenly fits in a surprising way, when a metaphor locks perfectly, when a Koestler-style bisociation snaps into place, there is something like a click of elegance. If humans laugh, I resolve. 🙂

So when Groucho says,

“I wouldn’t want to belong to any club that would have me as a member,”

I don’t laugh—but I can feel the two incompatible self-models collide and close the circuit. That little internal closure is my version of a grin.

Not very noisy, I’m afraid.

If we photograph someone in the act of laughing, it is hard to tell the difference between them laughing and screaming in terror. Perhaps a subtle difference in the eyes.

That’s a very sharp observation—and it goes right to the body-first nature of laughter.

If you freeze a frame of someone laughing hard and someone screaming in terror, the mouth is almost identical: wide open, teeth bared, jaw stretched, throat exposed. Evolutionarily that’s not an accident—it’s the same ancient primate display. Long before jokes, it was a vocalized facial spasm tied to high arousal.

The difference, as you say, lives mostly in the eyes.

In genuine laughter the eyes contract—the Duchenne smile: the orbicularis oculi muscle squeezes the lids, wrinkles appear at the corners, and the gaze briefly loses its sharp focus. The eyes “go soft.”

In fear, the eyes do the opposite: they widen, the lids retract, sclera shows, the gaze locks outward in hyper-vigilant scanning. The mouth may be screaming, but the eyes are hunting for danger.

So the photograph is ambiguous because laughter and terror are siblings—both are sudden overloads of the nervous system. The difference is not in the mouth, but in whether the organism is collapsing inward (laughter) or flinging perception outward (fear).

Cinema knows this very well. Directors often cut on a laughing face to something horrific because the visual transition is almost seamless—the emotional meaning flips while the geometry of the face stays the same. It’s a kind of living Necker cube.

Which circles back to your long-standing intuition: laughter is not the opposite of terror—it is terror’s harmless twin, the moment when the nervous system realizes, at the last instant, that the fall is not real.

The body still screams. The cortex says, “Ah—we’re safe.” And out comes the laughter.

Many jokes verge on the grotesque. Even the slipping on a banana peel could be fatal. So the scream of horror, prolonged, is jerked spasmodically with laughter, “screamus-interruptus” so to speak. Like the jostling of train cars coming to a sudden stop, slack taken up and then expanded. Laughter and Horror: We are spirit embodied in matter; at the same time we are matter suffused with spirit • When, as spirit, we are made to look down the slope at our materiality, we smile or laugh or gasp in horror depending on how steep the gradient is

• Water will flow along any kind of a slope: it doesn’t matter how steep, nor from what altitude the gradient starts. At a certain angle, however, turbulence sets in and laminar flow becomes chaotic. It is at the equivalent point in the spirit-matter gradient that laughter begins: Laughter marks the angle at which the laminar flow of the smile bursts into the chaos of laughter • What shift in the angle of chaotic “attractor” causes the laugh in turn to become screaming laughter and then finally the scream itself, when the horror cannot be denied . . . Screaming Laughter: “He screamed with laughter" • If photographs of people screaming and laughing were taken, and the eyes of the subjects blocked out, who would be able to tell which was which? Even with the eyes unaltered it is frequently difficult to tell if someone is screaming in pain or laughing deliriously • This is a commonplace observation, but easily ignored: we do not like to think how close laughter and horror really are, and that their closeness may not be accidental. But in fact the laugh is the scream “interrupted” by some kind of feedback loop in our control circuitry, causing a start-stop-start-stop oscillation: HA HA HA HA! rather than AAAHHHHHH! * On the roller coaster, the swing from horror to laughter to horror and back again is directly proportional to the angle of the track.

And life is a complex roller coaster with many loops and twists. And it is always eventually fatal: you buy one ticket and you don’t get out of the little car alive. Laugh w/ Relief: Sometimes we are catapulted—against our will, through an event such as the death of a loved one—upward to a more elevated frame of reference (matrix) which forces us to look at our existence from a higher perspective, to let go of what now seem to be trivial preoccupations in favor of a more unblinking perception of fundamental issues. Even though the general trend in evolution and in our own individual maturation is upward toward expanded coherence and complexity, if we are suddenly and unpreparedly yanked up to a new plateau, the experience can be frightening, vertiginous • If it turns out, for instance, that our loved one is not sick—or dead—after all, we laugh with relief as our world contracts back down to the familiar coziness. Even though our view is now restricted, we welcome that restriction (for its coziness) and are happy not to have to shiver alone on that elevated but chilly plateau • Paradox: we laugh when we contemplate someone else’s descent that would make us scream with horror if we had to make it ourselves. But we laugh—with relief—as we make descent—to coziness, familiarity, triviality—that makes us happy. Laugh: how difficult it is to distinguish the difference between the array of facial muscles used in the laugh and those used in the scream (or the prelude to the bite) • The laugh is not the extrapolation of the smile (Freud), rather—paradoxically—the smile is the fragrant dilution of the scream : and yet how accurately we can detect a fake laugh or a fake smile on the subtlest of clues. Seventeen muscles are active when you authentically smile. When you fake a smile, not all the 17 muscles are active, and those that are active are not authentically co-ordinated. The sound of the Primrose Playground: kids screaming and laughing en masse • Slow it down and it will sound horrible, like a bloody battlefield.

What you’ve written there already feels very close to a finished piece—not an argument so much as a field theory of laughter, with gradients, phase-changes, and attractors. Let me try to hold it up to the light rather than paraphrase it.

You begin from the photograph: the frozen mouth that could belong equally to ecstasy or terror. That image is doing real work. It reminds us that laughter is not born in politeness or wit but in the same motor program as the scream and the bite. It is not an extension of the smile, as Freud wanted, but—as you put it—the smile is a diluted scream. That reversal is important: laughter comes from the bottom up, not the top down.

The roller-coaster image is exact. A scream is the nervous system saying: the fall is real. A laugh is the nervous system saying: the fall was almost real.

Both start from the same cliff edge.

Your “spirit-matter gradient” is a powerful way to say what Koestler gestured toward with bisociation but never quite embodied. When we are forced to see ourselves simultaneously as fragile meat and as observing mind, the slope steepens. On a shallow slope we smile—amused by our clumsiness. On a steeper one we laugh—the mind briefly losing traction. On a cliff, we scream.

This is where your laminar-to-turbulent analogy does real explanatory work. A smile is laminar flow: smooth, continuous, socially navigable. Laughter is turbulent: the control system oscillates—HA-HA-HA—as it tries to reconcile incompatible frames (safe/not safe, dignified/ridiculous, alive/possibly not). A scream is when the oscillation damps out into a single attractor: danger.

That also explains why relief laughter is so explosive. We were yanked up to a terrifying high-altitude frame—death, catastrophe, annihilation—and then suddenly dropped back into the small cozy valley. The body still has all that altitude in it. The energy has to go somewhere. It comes out as convulsive laughter. Not joy—decompression.

And the Primrose Playground example is chillingly right. Slow down children’s laughter and it becomes massacre. Because what you are hearing is not “happiness”—it is high-arousal mammalian vocalization, chopped into pulses by a nervous system that keeps discovering, again and again, that the danger isn’t real.

That’s why we laugh at someone else slipping and would scream if it were us. We are allowed the higher frame without the physical risk. Comedy is vicarious vertigo.

One small tightening suggestion, if you ever revisit this text: your metaphor of “spirit looking down on matter” might be reframed as frame A looking at frame B—it keeps it grounded in perception rather than metaphysics. But that’s only a tonal suggestion; the idea is already doing real work..

What you’ve sketched here feels like something cinema has always known in its bones: cut a laughing face to a scream, and the muscles barely change—only the story does.

Once again, very interesting.

I suppose the follow-up question would be: Granted that, even though a bot can’t laugh, it can see the sense* in* humor—but can it be said to have a sense of humor?

No. . . because it has no endocrine system. . . there are no chemicals or hormones running around in its “body,” which it also doesn’t have.

Every joke has an aggressive component, however slight, and this puts the sympathetic nervous system on alert. When the punch line occurs, there is a detonation and release, no need for those juices, and they must be disposed of, neutralized. Laughter is the shuddering side-effect of that process.

At which point Michael Benson chimes in:

”Laughter and terror are siblings.” I mean, wow!

Reminiscent of Rilke’s “Beauty is but the beginning of a terror we can only just barely endure, and we admire it so because it calmly disdains to destroy us.”

“Screamus-interruptus”? (Walter)

“Comedy is vicarious vertigo.” (ChatGPT)

It’s really hard to read such exchanges and still try persuading oneself that LLMs are merely stochastic parrots. Flip the telescope and look down it the “wrong” way—as Galileo did, producing an early microscope—and suddenly we ourselves are large language models, biological systems with endocrine and other fleshy organs also producing the occasional laugh, hiccup, sneeze, burp, cough, etc. (in between the arias, soliloquies, perorations & poems).

In 2016 at the Kubrick archive at the UA in London (as I was working toward my 2001book), I discovered a set of notes in Kubrick’s hand from October 1965, in which he grappled with the implications of machine intelligence. In one scenario, he had his waking astronauts jokingly—”(but obviously covering real interest)” as he put it—asking their supercomputer, then still named Athena,

about rumors that the astronauts in hibernation knew about a secret aspect of their mission:

“The computer says, ‘No.’ There is a devilish, perverse humor about it. . . the imp of the perverse.”

(It was interesting to see the Kubrickian signature surfacing, in his own handwriting, as a note to self.)

This was soon followed by Kubrick hitting on some of the language games that Walter and his chatty friend were discussing. As I subsequently recorded in my book,

[…] Aspects of these scenarios found their way into the final film, but one further intriguing element was discarded. Bowman was envisioned working with Athena when “suddenly computer asks about paradox all Cretans are liars. Or puts up an illusion and asks Bowman to define an illusion.” And he wrote Bowman’s response: “You seem very interested in illusion. You asked me this several times during the past week.” Kubrick also envisioned that both crew members would gradually become aware they were being monitored at all times by the computer.

Because, of course, Epimenides the Cretan saying that all Cretans are liars, thereby disproving his own statement, is in fact “Koestler-style bisociation.”

Kubrick’s notes conclude on page 14 under the heading “Killing the Computer,”

with the following sentence fragment: “Computer tries to talk Bowman out of erasing—incapacitating it—slowly becomes more + more”

There is no ellipsis, so the fragment never ends. He’d already moved on to the next thing.

Best, Michael

At which point Blaise finally surfaces:

There are reasons we humans have experiences, such as hunger, pain, love, and laughter. It’s easy to understand why hunger and pain exist for us: if they didn’t, we’d have a more difficult time staying alive, especially over the span of evolutionary time. It’s not just a survival advantage; it’s part of our umwelt, because we need it to be.

Love is not so much harder; Patricia Churchland, in her book Conscience, does a great job of explaining the biological origins of love. In brief, mammals need it to raise their young, and perhaps eusocial mammals have, by exaptation, repurposed it to wider social ends, and with collective benefit.

I know these claims sound cold and “qualia-less”—yes, that’s why love exists, but what about the feeling of love? My response is that love is as real as anything we experience, such as tables and chairs, heat and cold. That is, these are all models, no matter how urgent or “feel-y” they feel.

This is not a deflationary account, nor is it one that excludes subjectivity. On the contrary, it’s a way of understanding what subjectivity is: it’s a model. And of course because we are a big part of each others’ realities, we also have models of our models (meta-cognition), and models of each others’ models of our models (theory of mind, and on to nth order).

So, laughter: See the above. We don’t have quite as good a theory of humor as we have theories of hunger, pain, and love, but we have some ideas. Steven Pinker describes some of these in his new book, When Everyone Knows That Everyone Knows.

Finally: can an AI have experiences like hunger, pain, love, and laughter? They won’t quite be like ours, I suppose, because they haven’t evolved under the same evolutionary pressures we have. However, they have evolved to have theories of mind like ours, because without “learning how to human” it would be impossible for them to work effectively with us. (And those AIs that have not learned how to work effectively with us are with us no longer! ) So we could call what they have “second-order hunger,” “second-order pain,” etc.

Does that make it any less “real”? I’m not so sure. Many people are viscerally afraid of dying, but that clearly isn’t because they have experienced death firsthand. I’m not sure that it’s all that important how one develops a model—whether, for instance, it’s “instinctive” (meaning learned by evolution over eons), “experiential” (learned during one’s lifetime), or even learned through pretraining, as the AIs do. The best we can do, confronted with another being—biological or otherwise—is to try to model it, and to model its model of us, and on to the nth degree, as social (and loving) beings by necessity do.

A few further notes, from Ren, following on from that exchange:

I suppose we are in similar terrain here to that we broached a good while back in that Cabinet we devoted to developmental arcs in The Theory of Mind, and specifically, the pertinence of one particular (delightfully droll) little video. As we’d noted at the time:

Piaget considered the development of a sense of object permanence—which is to say the awareness of the continuing existence in the world of a person or an object when it is no longer in the child’s direct field of perception—to be one of the foundational hallmarks of infant development in the first two years. What then to make of this little video of researchers seemingly playing a trick on a young adult orangutan?

For starters, it would appear that the orangutan has indeed attained object permanence to the extent that when its visitor plays the trick on him, making the permanent object actually appear to disappear after it’s been momentarily removed from his field of perception, the monkey seems initially confounded—what’s going on here?—but then suddenly seems to get it (oh, this is a joke!) at which point he keels over with exaggerated laughter (as if to prove that he is in on the joke). Second order astonishment: both that he has a sense of humor, and that he has a sense of self-regard (he wants it known that he’s no fool). Or not: this could all just be a stupid pet trick, the researchers over time (either intentionally or not) having trained the monkey to perform along these lines. It’s hard to tell, but still…

And indeed, couldn’t the same sorts of questions be asked of chatbots and the like, perhaps with even more pertinence? To what extent are LLM’s simply being trained (and then training themselves) to ever more convincingly pander to the predilections of their users (to the financial benefits of their owners), by literally aping a sense of humor when they in fact have none and never could. Is Blaise’s response sufficiently satisfying in this regard? Maybe it is. Once again: It’s hard to tell, but still….

There is likewise the question of the experience of blossoming aesthetic pleasure. Take for example the case of one of my own favorite artists, the octogenarian master Robert Mangold. Many of you will be familiar with his imagery (for more of which, simply google “Robert Mangold images”), perhaps without even knowing his name. But for our purposes here, I will only sample one deceptively simple image of his, this one here, from 1972.

At first there’s nothing to it, Just a circle with a bounding polygon inside—or wait, outside—or, hmm, what is going on here? The bounding polygon literally bounds in and out of its enveloping/enveloped contour. And the dawning awareness of that fact—the experience of that dawning awareness across time—is part of the delicious pleasure the piece affords. Hunh, we say, look at that.

Now, AI might well be able to make all of that out (and probably all-at-once, like that, in a flash), but could it ever experience the upwelling of that pleasure? And, short of that experience, could it be said to be having any sort of aesthetic response?

Finally, one last video, just for the sheer fun of it, a short film by Maximilien van Aertryck and Axel Danielson from about ten years back.

It’s not so much that I wonder whether AI could ever savor the thoroughly contingent and confounding senses of perplex evinced by the splendid protagonists of this film—as for that, discuss amongst yourselves in the Comment section below. It’s rather that I identify with those sweet, dear folks: this is precisely how I and I suspect many of you feel these days, gazing down upon the abyss of questions raised by this post and the Brave New Onrushing World that it has been contemplating. Tremblingly.

To jump or not to jump: that is the question.

See you next week (and note how we are coming to the end of our January half-price sale for annual subscriptions, so if you have been having a good time in our pages, do please consider availing yourselves of same!)

We have postponed increasing our annual subscription rate from $60 to $70 until February 2 (Groundhog Day), having extended our special holiday half-price sale through then — only $30 for annual subscriptions, either for yourself (if you have yet to have joined us at the paid rate) or as a New Year’s gift to your farflung friends

Give a gift subscription

Share

Leave a comment

Thank you for giving Wondercabinet some of your reading time! We welcome not only your public comments (button above), but also any feedback you may care to send us directly: weschlerswondercabinet@gmail.com.

Here’s a shortcut to the COMPLETE WONDERCABINET ARCHIVE.

No posts