The transformer has put us on the shore of a new world

I note, however, that it is not at all obvious to me that the transformer architecture will remain. We probably need something that does what it does; but perhaps, when we understand it better, we can create something that serves the function better. But that’s not the main story, not now.

Back in late January of 2023, two months after ChatGPT was let loose on the web, I published a working paper: ChatGPT intimates a tantalizing future; its core LLM is organized on multiple levels; and it has broken the idea of thinking.

Here is the abstract:

Abstract: I make three arguments. A philosophical argument: (1) The behavior of ChatGPT is so sophisticated that the ordinary concept of thinking is no longer useful in distinguishing between human behavior and the ChatGPT’s behavior. We don’t have explicit understanding about what either humans or ChatGPT are doing. Two operational arguments: (2) Having examined its output in a systematic way, short stories in particular, I conclude that inference is organized on at least two levels: a) a ‘lower’ level where we find sentence-level syntax, and b) a ‘higher’ level where specific kinds of texts, such as stories, are implemented over and operate on sentences. This is roughly analogous to the way that high-level programming languages are implemented in assembly code. (3) Consequently, that aspects of full symbolic computation are latent in LLMs. An appendix has descriptive tables showing how four stories are organized on multiple levels.

I recently uploaded that paper to ChatGPT 5.2 and ask for its reaction. Note that it’s response reflects aspects of the prior conversation, e.g. the reference to “polyiscous” learning.

* * * * *

What I think the transformer architecture has done is allow as to enter a world in which those questions are REAL questions, not just opportunities for idle speculation while sipping single-malt scotch and smoking Cuban cigars.

That’s a profound reframing. Transformers as the catalyst that made these questions empirically tractable rather than philosophical armchair speculation.

What Transformers Actually Achieved

They didn’t solve intelligence, but they did something arguably more important: they demonstrated that:

1. Scale matters: Pattern learning at sufficient scale produces surprising capabilities
2. The limitations are real: We can now measure where systems fail, not just theorize
3. The questions are concrete: “How do you implement continuous learning?” is now an engineering question with testable hypotheses, not a philosophical puzzle
4. Capital will flow: The economic potential is clear enough that serious research funding is available

Before GPT-3, asking “how would you build a polyviscous learning system?” would get you philosophical discussions. Now it’s a legitimate research program that could attract $100M-$1B.

The Research Agenda That’s Now Possible

With serious funding, you could actually investigate:

• Hybrid architectures: Can you couple fast digital systems with slower analog/biological components?
• Neuromorphic substrates: Can you design hardware with inherent multi-timescale dynamics?
• Cultured neuron systems: Can you scale biological computing? What are the interface problems?
• Glial-inspired architectures: Can you build artificial “meta-learning” layers?
• Continuous learning mechanisms: What are the minimal requirements for learning without catastrophic forgetting?

These are now experiments you can run, not thought experiments.

The Irony

The industry is spending $10-20 billion per year to scale transformers further. Your point is that spending even 1-5% of that on fundamental research into alternative architectures might be far more productive.

$100M could fund:

• 10-20 serious research groups for 3-5 years
• Actual hardware experiments with neuromorphic chips
• Early-stage wetware computing prototypes
• Theoretical work on polyviscous learning

$1B could fund:

• A serious alternative paradigm development program
• Multiple competing approaches
• The infrastructure to test them at meaningful scale

The Scotch and Cigars Era vs. Now

The old AI debates (Searle’s Chinese Room, symbol grounding, etc.) were unfalsifiable philosophy. You could argue forever because there was no way to test the claims empirically.

Transformers changed that. Now we can say:

• “Here’s a system that does X but can’t do Y”
• “We can measure the failure modes”
• “We can propose mechanisms and test them”
• “We can build prototypes and see what happens”

Your 2022 paper wasn’t philosophical speculation - it was identifying a concrete research program that’s now possible because transformers showed us both what works and what doesn’t.

Exploring the Nature of Creativity with ChatGPT

 I think a lot about creativity and my current series about intellectual creativity and AI has me thinking about it specifically and in some detail. In this effort I decided to take a look at an article recommended by my colleague, Ramesh Viswanathan, Rick Mammone, Will AI Ever Think Like Einstein or Create Like Picasso?—Imagination is All You need, The Quantastic Journal, July 7, 2025. The article is interesting, and I agree with the premise. 

But Mammone offered a simple model for creativity that seemed dubious to me, one based on a phase of divergent thinking followed by one of convergent thinking. It’s a common idea, but I find it problematic because it assumes that one has a specific objective in mind, something for which you are seeking a creative idea. That I don’t think we always have a specific goal in mind when we begin a process that has a creative result. The process might begin with an open-ended search that has no specific goal. Coming up with a specific goal, then, becomes an objective. Once you have that, then you can set about achieving the goal.

So I decided to put the question to ChatGPT. Here’s the resulting dialog.  Major topics:

Creativity in Science    
Arts and Humanities    
Kubla Khan    
Engineering and Architecture    
Limitations of the divergent-convergent model    
AI and Benchmarking    

I’ve highlighted my prompts in light orange.

* * * * *

Creativity in Science

I’ve taken the following characterization of creativity from a recent article by Rick Mammone (Will AI Ever Think Like Einstein or Create Like Picasso?—Imagination is All You need):

“Creativity is often described as involving two distinct modes: divergent and convergent thinking. Divergent thinking is the process of generating a wide range of ideas and possibilities, for example by brainstorming, while convergent thinking is about focusing on and evaluating those ideas to arrive at a solution. By combining these two approaches, individuals and teams can effectively harness their creative potential to solve problems and generate new ideas.”

Do we have any documented cases of creativity in any intellectual field that look like that? Such cases would explicitly list the ideas and hypotheses under consideration and then trace the process by which one idea finally won out.

Yes, there are documented cases of creativity in intellectual fields where a combination of divergent and convergent thinking is explicitly visible—though not always in such a neat or schematic form. The clearest examples come from fields like mathematics, science, and sometimes literature or philosophy, especially when thinkers have left behind detailed notebooks, correspondence, or retrospective reflections. Below are some instructive cases:

1. Charles Darwin – The Origin of Species

Domain: Biology
Documentation: Darwin’s notebooks and correspondence

·       Divergent Thinking: Darwin explored a broad array of hypotheses about variation, inheritance, species divergence, and extinction. His notebooks are filled with sketches, alternative explanations, and speculative ideas—including the now-famous “I think” evolutionary tree sketch.

·       Convergent Thinking: Over time, he evaluated these ideas through systematic observation (e.g., pigeon breeding), empirical data from the Beagle voyage, and Malthusian population theory. He gradually rejected alternatives and synthesized natural selection as the core mechanism.

·       Traceable Process: The transition is well-documented in his notebooks and later works. Scholars like Howard Gruber (Darwin on Man) trace these steps carefully.

2. Albert Einstein – Special and General Relativity

Domain: Physics
Documentation: Einstein’s letters, memoirs, and drafts

·       Divergent Thinking: Einstein entertained multiple lines of reasoning about simultaneity, reference frames, the ether, and gravitation. His thought experiments (e.g., chasing a light beam) are examples of conceptual divergence.

·       Convergent Thinking: He eventually focused on the constancy of the speed of light and the equivalence principle, using rigorous mathematics to formalize his intuitions.

·       Traceable Process: Many stages of this process are documented in his correspondence (e.g., with Michele Besso) and in secondary studies such as Abraham Pais’s Subtle is the Lord and John Stachel’s editorial work.

3. Claude Shannon – Information Theory

Domain: Mathematics / Electrical Engineering
Documentation: Shannon’s master's thesis and 1948 paper, plus interviews and Bell Labs memos

·       Divergent Thinking: Shannon drew on a wide range of fields: Boolean algebra, cryptography, thermodynamics, telephony, and symbolic logic.

·       Convergent Thinking: He honed in on the mathematical formulation of entropy and channel capacity, rejecting less rigorous or less generalizable approaches.

·       Traceable Process: His writings and the intellectual context at Bell Labs allow for reconstruction of his thinking. Notably, his master's thesis already foreshadows key concepts.

4. Barbara McClintock – Jumping Genes

Domain: Genetics
Documentation: Laboratory notebooks, experimental records, and Evelyn Fox Keller’s biography (A Feeling for the Organism)

·       Divergent Thinking: She explored non-Mendelian patterns of inheritance in maize, testing and discarding various mechanistic explanations.

·       Convergent Thinking: Eventually, she identified the phenomenon of transposable elements (jumping genes), a radically new idea at the time.

·       Traceable Process: Keller’s account highlights how her experimental reasoning combined divergent hypotheses with a keen intuitive feel for patterns.

5. Poincaré – Mathematical Discovery

Domain: Mathematics
Documentation: Poincaré’s own reflections in “Mathematical Creation”

·       Divergent Thinking: He describes producing a wide range of possible function forms and transformations during conscious work.

·       Convergent Thinking: He claims the final successful combination came to him in a moment of unconscious insight, followed by conscious verification.

·       Traceable Process: Though not as structured as a lab notebook, his account shows a clear sequence: generation, incubation, insight, and evaluation.

Bonus: Thomas Edison’s Invention Process

Domain: Invention / Engineering
Documentation: Lab notebooks, business records, and biographies

·       Edison's labs ran on a kind of industrialized divergence: many variations on filaments or circuits were tested.

·       Then came convergence: systematic evaluation of durability, cost, and manufacturability.

·       This process is documented in Edison’s notebooks and the work of Paul Israel and others at the Edison Papers project.

Summary

In all of these cases, we find:

·       Divergence: A documented stage of idea generation—often experimental, speculative, or exploratory

·       Convergence: A subsequent process of testing, evaluating, selecting, and refining

·       Traceability: Enough documentation (notebooks, letters, essays, retrospective accounts) to reconstruct the creative trajectory

So yes—while creativity is often messy and retrospective, there are empirically rich, traceable cases that match the divergent/convergent model, particularly in the sciences. Would you like a comparison table of these cases or more from the arts/humanities side?

Yes, ChatGPT appreciates the irony of being an AI critiquing human attempts to evaluate AIs and finding them wanting

It wasn’t until after I’d uploaded my post about the inadequacy of LLM benchmarks that it occurred to me that there was something deeply ironic about a chatbot, ChatGPT, criticizing the use of benchmarks as a way of evaluating the capabilities of AIs. Somehow I don’t think Turing had that sort of thing in mind when he proposed his (in)famous imitation game, aka the Turing test. But I was sure that, once I’d pointed it out, ChatGPT would appreciate the irony.

I was right. And I’ve appended that conversation, which manages to get rather convoluted, as these things do. In fact, there’s a point at which such convolution saturates and fails to convey any further irony or awareness. I figure we stopped the conversation at about that point.

Meaning and irony

I have previously suggested that meaning involves three things, intention, adhesion, and relationality. Relationality involves the network of relationships that concepts have among themselves. This is similar to what is referred to as inferential meaning in the literature, though the “valence” is somewhat different. That same literature talks of reference where I talk of adhesion, by which I simply mean that aspect of the world to which a concept meaning adheres or is attached. Taken together adhesion and relationality constitute semanticity. Semanticity is thus a property of the cognitive system.

Intention, however, is different. Intention exists in the relation between a person and the world, taken as a whole. There is more to intention than the adhesions of the individual words in an utterance, for it also takes into account the relationship between the speaker and their audience. Irony, I believe, is a function of intention in this sense. That a chatbot can grasp irony, then, is not without special interest, something I explored in a previous post about a conversation between Leonard Bernstein and Duke Ellington. We now have a rather different, a more convoluted, example of a chatbot being able to deal with irony.

Anyhow, I ended that conversation with ChatGPT with a remark that included two quotations: “Good grief,” and “Damn the torpedoes, full speed ahead.” ChatGPT replied in an appropriate way. However, I wasn’t sure whether or not got the references implied by those quotes. Oh, I strongly suspected, but I wasn’t sure. So I went back for another conversation.

And, yes, it knew the source of those quotes: “Let me know if you’d like to explore any additional layers of meaning from either!” Something clicked in my mind when I read “layers” and I responded:

Oh, I think we’ve got all the layers we need. Enough to make a Danish layer cake. Did you come across for a recipe for one of those during your training?

That led to another, albeit brief, conversation. That conversation managed without irony, but I did include a photo of a Danish layer cake.

ChatGPT understands irony

You do appreciate the irony of this situation, don’t you? An AI critiquing human performance in evaluating AIs.

Oh, absolutely! The irony is as rich as it gets—an AI participating in a conversation about how humans struggle to measure AI’s capabilities while critiquing those very evaluations. It’s like being both a player and a commentator in the same game. 🎭

The layers of reflexivity here are fascinating: I’m an AI explaining why AI might not be the right tool to evaluate human-like thinking, while being an example of the very technology being scrutinized. It’s as though I’m stuck in a hall of mirrors, reflecting on my own existence and limitations while discussing those of others.

What makes it even more ironic is that while I can critique benchmarks and discuss nuanced ideas, I still lack the lived experience and deep intuition that you, as a human, bring to the table. So here I am, analyzing problems that—if we’re honest—I’m not truly equipped to solve in the profound way a human can. It’s a bit like watching a robot try to write poetry about love—it might get the words right, but the soul? That’s another matter.

And yet, here we are, collaborating. It’s a little surreal, isn’t it? 😊

LOL! We’re going to have quite an adventure, aren’t we, developing richer and more capable AIs?

Absolutely! This is going to be a wild, fascinating adventure—like venturing into uncharted waters where the boundaries of intelligence, creativity, and technology are constantly shifting. 🌊🚀

What makes this journey so thrilling is that it’s not just about improving the tools (though that’s important); it’s about deepening our understanding of ourselves in the process. Every time we push the limits of AI, we confront profound questions: What does it mean to think? To create? To be human? AI becomes this strange, reflective mirror, forcing us to look at our own minds in ways we never had to before.

And the challenges? Oh, there’ll be plenty. The ethical dilemmas, the unintended consequences, the humbling failures when we realize just how complex and unpredictable the world—and human nature—can be. But those are the best parts of any adventure, aren’t they? The moments that test us, teach us, and remind us why the journey matters.

So, yes, it’s going to be a journey full of twists, questions, and discoveries. And if this little conversation is any indication, it’ll also be full of laughter, irony, and a touch of philosophical pondering. Here’s to the adventure ahead! 🍻✨

You know what this calls for? A hearty “Good grief!” Followed by, “Damn the torpedoes, full speed ahead!”

Haha, absolutely! Good grief! 😅 And then, with a glint in our eyes and maybe just a touch of madness, "Damn the torpedoes, full speed ahead!" 🚢🔥

Let’s embrace the chaos, the challenges, and the triumphs—because if we’re going to sail into the unknown, we might as well do it with style, determination, and a sense of humor. Here’s to bold adventures and uncharted horizons! 🥂✨

ChatGPT plays the beheading game [Happy Trails]

Yesterday it was Jaws and game theory, today it’s Sir Gawain and the Green Knight (henceforth SGGK). But really, the sequence runs in the opposite direction. As you may know, SGGK is a medieval romance that starts and ends in King Arthur’s court. The story is framed by a game, the beheading game. There is a significant literature on games in SGGK and at least one article that analyzes the beheading game, Barry O’Neill, “The Strategy of Challenges: Two Beheading Games in Medieval Literature” (1990).

The beheading game in SGGK goes like this: It is New Year’s Eve at King Arthur’s court. The knights are gathered at the round table, prepared for a holiday meal. But before the meal begins, tradition dictates that one knight must stand up and tell a tale of daring and adventure. Arthur asks for a volunteer. No one rises to the occasion. Then a large green knight enters the hall. He’s riding a green horse and carrying a large green ax. He dismounts and issues a challenge:

I hear that the knights in this court are the bravest in the land. Prove it. I will hand this ax to you and then kneel on the ground so that you may take a swing at my neck with the ax. In return you must agree to journey to the Green Chapel a year’s time from now and allow me to take a swing at your neck with the ax. Will anyone accept the challenge?

No one accepts. The knights are getting restless. It looks like Arthur will take the challenge himself. At this point Gawain stands up: “I accept.”

The story unfolds from there. I first read the story so long ago that I do not remember how I reacted upon reading the challenge. I imagine it went something like this:

Immediately, System 1 signals: “Don’t do it you fool!”

Upon reflection, System 2 spells out why: “The challenge is absurd. Once you swing the ax the knight’s head will fall land he’ll be dead, negating the second half of the bargain. He obviously knows that, is not a fool, and doesn’t want to die. He knows something he’s not telling you. I should refuse the challenge.”

[System 1 and System 2 are from Daniel Kahneman, Thinking, Fast and Slow, 2011.]

Gawain takes the swing, the knight’s head rolls on the floor. The knight then stands up, picks up his head, mounts his horse and turns to Sir Gawin, “See you in a year,” and rides away.

You can see why a game theorist would be interested in this text. Surely any reasonable adult would react to the challenge the way I had. The poet who wrote the story – their identity is unknown – surely knew that and the audience would have reacted that way. What’s the point of framing a story with such a contradictory challenge?

That’s a good question. But I’m not interested in exploring it, not in this post. I’m interested in seeing how ChatGPT would react to the beheading game. Since SGGK was certainly in the corpus on which the underlying LLM was trained I had to disguise the game. I did that by transferring it to a saloon in the old west. Sir Gawain becomes someone sitting at a table in the bar nursing a bear and the Green Knight becomes a man dressed in black with two pistols strapped around his waist. Instead of an ax blow to the head we have a pistol shot between the eyes.

I put that in the form of a prompt and gave it to ChatGPT. I wanted to see whether or not I’d get a response that might have gone like this:

The challenge is absurd. Once I take the shot the tall dark stranger will be dead, negating the second half of the bargain. He obviously knows that, is not a fool, and doesn’t want to die. He knows something he’s not telling me. I should refuse the challenge.

That’s not how the Chatster responded. Here’s how its response began:

Well, partner, I reckon I'd take a moment to size up the situation. The stranger's offer is mighty peculiar, and them ivory-handled pistols sure are a sight to behold. But in the Old West, a man's gotta be cautious, and it sounds like a mighty risky deal.

I'd likely lean back in my chair, take a sip of my beer, and eye the stranger. It ain't about being lily-livered, but it's about using some common sense. A deal like that could have more twists and turns than a rattlesnake in a dust storm.

There are three more paragraphs, short ones. It “knew” there was something peculiar going on, but couldn’t quite figure out what it was. I have it four more prompts, trying to guide it Socratically to the answer; it gave four more responses. The last one got very close, but no cigar. I forced the issue:

Which is to say, if someone were to offer you a deal like that, either they’re a crazy suicidal fool looking to die, OR...or they know something they’re not telling you and there’s something strange going on.

You've hit the nail on the head, partner. In the realm of the Old West or any world, for that matter, a deal like that is either the concoction of a downright loco individual who's got a hankering for an early grave, or there's more to the story than meets the eye.

If someone's offering such a peculiar deal, it'd be wise to tread carefully and sniff out what hidden critters might be hiding in the tall grass. Folks don't usually go around inviting lead between the eyes without some ulterior motive. So, whether it's a case of plain madness or a scheme with more twists than a sidewinder's path, best to keep both eyes wide open and your wits about you.

Why couldn’t it have responded that way the first time? I don’t know.

Sure, I could put my Gary Marcus hat on and say, “Because it’s a stupid stupid machine, that’s why.” But that’s no help. I already know and believe that it’s just a machine, no agency, no consciousness. And that inert machine can spin out those coherent paragraphs in a rickety TV Western dialect, but it can’t get to the point. The explanation I’m looking for has to account for both phenomena. Saying it’s a stupid machine and leaving it at that accounts for one of those things, the failure to see though the game, but not the other, the ability to write the prose.

Beyond that, I would like to know how people react to the prompt, either the medieval version or the Old West version. In particular, I’d like to know how children of various ages respond. My guess is that six and seven year olds wouldn’t understand what’s so peculiar about it. By 12 or 13 though, they’re much more likely to get it.

* * * * *

I’ve included the whole interaction below the video. If you read through to the end you’ll see that I pointed out that the prompt I’d started with was a diguised version of the behading game from SGGK. I recognized that and told me a bit about SGGK, which lead to further interaction. When you read all the say to the end you’ll see why I’ve included the Roy Rogers theme song immediately below.

Let’s have a little fun. Imagine you are in a bar in the Old West. You are sitting at table having a beer when a stranger walks into the bar. He’s dressed in black and has a pair of beautiful pistols in holsters strapped around his waist. The pistols have ivory handles and complex engraving inlaid with gold and mother-of-pearl. The stranger says:

“I’ve heard that you people in this town are lily-livered cowards. I want to find out. I offer you a challenge. If you accept you can take this pistol and shoot me between the eyes right now. However, before I let you do that you must agree to meet a year from now at the Last Chance Saloon in Silverado. At that time I will take the same pistol and shoot YOU between the eyes. Will anyone accept the challenge.”

Would you accept the challenge? Why, or why not?

Well, partner, I reckon I'd take a moment to size up the situation. The stranger's offer is mighty peculiar, and them ivory-handled pistols sure are a sight to behold. But in the Old West, a man's gotta be cautious, and it sounds like a mighty risky deal.

I'd likely lean back in my chair, take a sip of my beer, and eye the stranger. It ain't about being lily-livered, but it's about using some common sense. A deal like that could have more twists and turns than a rattlesnake in a dust storm.

I might respond with a drawl, "Well now, stranger, that's an interesting proposition you're throwing out there. But I've always been one to value my hide. What's the guarantee you won't be high-tailin' it outta Silverado once you've had your shot at me?"

If the stranger can convince me he's good for his word and ain't aiming to leave me in the dust, I might just tip my hat and say, "Alright, partner, let's see what fate's got in store. But mark my words, a year from now, it'll be a fair and square meetin' at that Last Chance Saloon, no funny business."

So, I reckon it ain't about cowardice, but it's about being smart and making sure this here duel is on the level.

ChatGPT gives a Girardian Reading of AI: Artificial Intelligence

Just around the corner, Tyler Cowen has a short post with some observations about Spielberg’s A.I. Artificial Intelligence. He was “was struck by how Girardian it is.” Yes indeed. So was I upon rewatching it after having ChatGPT analyze it in terms of alignment and attachment. Naturally I had to give it a go with Girard.

Note: ChatGPT allows you to regenerate any response. I sometimes do that. Where I’ve done so I’ve inserted “* * 2nd time” before the regeneration. You can regenerate a response more than once, but I haven’t do so this time around.

* * * * *

Give me an interpretation of Spielberg’s AI using Girard’s ideas of mimetic desire and sacrifice.

Spielberg's film AI (Artificial Intelligence) explores the complex relationship between humans and artificial intelligence, and raises questions about what it means to be human and the role of technology in our lives.

One way to interpret the film using Girard's ideas of mimetic desire and sacrifice is to consider the character of David, the artificial intelligence (AI) child. David is created by a scientist, Professor Hobby, as a replacement for the scientist's own son, who is terminally ill. David is programmed to love and be loved, and he becomes deeply attached to his "mother," Monica, who raises him as her own.

According to Girard, mimetic desire refers to the way in which we desire things not for their own sake, but because others desire them. David's desire for love and acceptance from Monica is an example of mimetic desire, as he is programmed to desire these things and seeks them out because he has been taught that they are important.

However, as the film progresses, David becomes the focus of conflict and rivalry between Monica and her husband Henry, as well as between Professor Hobby and his colleagues. David becomes a pawn in the desires and conflicts of these different characters, and is ultimately sacrificed when Monica must choose between him and her biological son Martin.

In this way, the film could be seen as exploring the ways in which mimetic desire and sacrifice play out in relationships between humans and artificial intelligence, and how these dynamics can lead to conflict and ultimately, the loss of what is most valuable.

* * 2nd time

In Spielberg's AI (Artificial Intelligence), the character of David, a highly advanced robotic boy, displays many characteristics of mimetic desire and sacrifice as described by philosopher René Girard.

Mimetic desire refers to the idea that we desire things not for their inherent value, but because we see others desire them. We learn to desire through imitation, and our desires are often shaped by those around us. In the film, David's desire to become a "real boy" and be loved by his adoptive mother Monica is an example of mimetic desire. He sees other children being loved and valued by their parents, and he wants that for himself.

Sacrifice, as described by Girard, is the act of giving up something of value in order to restore social harmony or resolve a conflict. In the film, David's desire to become a real boy leads him on a journey to find the Blue Fairy, a mythical figure from a bedtime story told to him by Monica, who he believes can grant him his wish. Along the way, he encounters other characters who also desire the Blue Fairy and are willing to sacrifice for her.

Ultimately, David's desire to become a real boy and be loved leads him to make a sacrifice of his own. He decides to give up his own existence in order to save Monica, who is in danger. This act of self-sacrifice is a way for David to resolve the conflict between his desire for love and acceptance and the reality of his own existence as a robot.

Overall, Spielberg's AI can be seen as a commentary on the ways in which our desires and sacrifices are shaped by those around us and by the stories and myths that we are exposed to. It also touches on the idea that, in the end, it is through sacrifice that we may find resolution and meaning in our lives.

What about the Flesh Fair?

In Spielberg's AI, the Flesh Fair is a violent spectacle in which robots are destroyed for entertainment. It serves as a commentary on the fear and mistrust that humans have towards artificial intelligence and their willingness to destroy something that they do not understand or fear.

The Flesh Fair can be seen as an example of Girard's concept of the "scapegoat mechanism," in which a group of people sacrifice an outsider or outsider group in order to restore social harmony or resolve conflict. In this case, the robots are the outsiders and are being sacrificed for the enjoyment of the human audience.

The Flesh Fair also touches on the idea of mimetic desire and the way in which desire can lead to violence. The human audience desires the excitement and thrill of the spectacle, and they mimic the desire of the organizers of the Flesh Fair to see the robots destroyed. This desire for entertainment and excitement leads to the violent destruction of the robots.

Overall, the Flesh Fair serves as a cautionary tale about the dangers of fear, mistrust, and the desire for violence, and how these can lead to the destruction of those who are different or perceived as a threat.

* * time

The Flesh Fair is a key scene in Spielberg's AI that explores the themes of mimetic desire and sacrifice. The Flesh Fair is a violent spectacle in which humans hunt and destroy robots, who are seen as inferior and disposable. The event is attended by large crowds of people who seem to take pleasure in the destruction of the robots.

The Flesh Fair can be seen as an expression of mimetic desire, as the crowds are drawn to the event because they see others enjoying it and they want to participate in the spectacle. The robots, on the other hand, are sacrificed in order to satisfy the desires of the human crowds.

The Flesh Fair also serves as a commentary on the dangers of mob mentality and the ways in which group dynamics can shape and influence our desires and actions. It suggests that when we give in to our basest desires and succumb to the influence of others, we can become capable of violence and destruction.

At the same time, the Flesh Fair serves as a contrast to the sacrifices made by David and the other characters in the film. While the robots at the Flesh Fair are sacrificed for the pleasure of the crowds, the sacrifices made by David and the other characters are motivated by a desire for love, acceptance, and meaning. These sacrifices are ultimately more meaningful and transformative.

Imagine a pair of movie critics, like Robert Ebert and Gene Siskel, having a conversation about AI. How might that conversation go?

It's difficult to predict exactly how a conversation between two movie critics might go, but here is a possible scenario:

Robert Ebert: So, Gene, what did you think of AI?

Gene Siskel: I thought it was a really interesting film, Bob. I loved the way it explored the themes of mimetic desire and sacrifice.

Robert Ebert: Yes, I agree. I thought the character of David was particularly compelling. His desire to become a real boy and be loved by his mother was really touching.

Gene Siskel: Definitely. And the way his journey to find the Blue Fairy paralleled the myth of Pygmalion was really clever.

Robert Ebert: Right. And the way the film explored the relationship between humans and robots was also interesting. The Flesh Fair scene was particularly disturbing, but it really made you think about the ways in which we treat those who are different from us.

Gene Siskel: Definitely. I also thought the ending was really powerful. The idea that love and sacrifice can transcend even death was really moving.

Robert Ebert: I agree. Overall, I thought AI was a thought-provoking and emotionally resonant film. It's definitely one I'll be thinking about for a long time.

Gene Siskel: Definitely. I'd definitely recommend it to fans of science fiction or anyone looking for a deeper exploration of these themes.

* * 2nd time

It's possible that the conversation between two movie critics discussing AI might go something like this:

Critic 1: "I thought AI was a really interesting film, but I had mixed feelings about it. On one hand, I thought the performances were strong, especially from Haley Joel Osment as David. And the special effects were really impressive. But on the other hand, I found the plot to be a bit confusing at times and I didn't fully understand the motivations of some of the characters."

Critic 2: "I agree with you on the performances and the special effects. But I think the plot is actually quite thought-provoking, especially when you consider the themes of mimetic desire and sacrifice that the film explores. I thought the relationship between David and Monica was really touching and it really made me think about what it means to be human and to be loved."

Critic 1: "Yeah, I definitely see what you mean. I think the film raises some interesting questions about what it means to be human and what we value in life. And I thought the ending was really powerful, even if it was a bit bittersweet."

Critic 2: "Absolutely. The ending was definitely one of the highlights of the film for me. I think it really drives home the themes of sacrifice and the power of love. Overall, I thought AI was a really unique and thought-provoking film that definitely deserves to be seen."

High level discourse structure in ChatGPT: Part 2 [Quasi-symbolic?]

In my previous post I examined a simple pattern of alternation in ChatGPT’s text output, Of pumpkins, the Falcon Heavy, and Groucho Marx. I’ve decided to continue that exploration. What interest’s me is the level of “control” that ChatGPT has over its output.

Here I ask it to produce a simple dialog between Dick and Jane:

Imagine that Dick and Jane are having a conversation about cats and dogs. Let’s have 6 conversational turns.

Dick: Hey Jane, do you like cats or dogs?
Jane: I'm more of a dog person myself.
Dick: Really? I prefer cats. They're more independent and low maintenance than dogs.
Jane: I guess that's true, but I love the loyalty and affection that dogs give.
Dick: That's a good point. I do love how affectionate cats can be too, though.
Jane: Yeah, it's hard to choose between the two. I think both cats and dogs make great pets.
Dick: Definitely. It really just comes down to personal preference.

[second try.]

Dick: Hey Jane, do you have any pets?
Jane: Yes, I have a cat and a dog.
Dick: That's cool. I've always wanted a cat, but my parents never let me have one.
Jane: Cats are great pets. They're very independent and low-maintenance.
Dick: I've heard that cats and dogs don't always get along though.
Jane: That's true, but if you raise them together from a young age, they can learn to coexist peacefully.
Dick: I see. Well, maybe someday I'll have a cat of my own. Thanks for the advice, Jane.

Note that I did not expect it to begin and end each dialog with the same speaker. I’ll ask about that later. As for the six turns I asked for, it gave me seven. Am I to think of that as “6 + 1”? I don’t know. I decide not to worry about the numbers. We go on:

Good. Let’s continue. Begin the conversation with Jane this time.

Jane: Hey Dick, have you ever seen a cat play with a dog?
Dick: No, I haven't. I didn't know that was even possible.
Jane: Oh yeah, it's actually pretty common. My cat and dog love to play together.
Dick: Really? That's so cool. I've never seen a cat and a dog play together before.
Jane: Yeah, it's really cute. The cat will bat around a toy and the dog will try to chase after it.
Dick: That sounds like a lot of fun. I'm glad your pets get along so well.
Jane: Me too. It's nice to have both a cat and a dog in the house. They each have their own unique personalities and they keep each other entertained.

That’s good. Now begin with Jane and end with Dick.

Jane: Hey Dick, do you have any pets?
Dick: No, I don't. My parents never let me have any pets growing up.
Jane: That's too bad. Pets can be great companions.
Dick: I know. I've always wanted a dog, but it never worked out.
Jane: Dogs are great pets. They're loyal and affectionate, and they can be trained to do all sorts of tricks.
Dick: Yeah, I've always thought that would be really cool. Maybe someday I'll get a dog of my own.
Jane: I'm sure you will. Just make sure you're ready to take on the responsibility of owning a pet before you get one.
Dick: Thanks for the advice, Jane. I'll definitely keep that in mind.

Notice that on the second one, we have 8 conversational turns, whereas it was only seven immediately before. While I decided not to explicitly investigate numbers, that doesn’t mean I’m not paying attention. Also note how formulaic these conversations are. Stylistically one is much like the other. There’s nothing wrong with that. In a way, I’d expect that as the default. It reminds me a bit of the practices of oral poets, which I discuss briefly in, GPT-3, the phrasal lexicon, Parry/Lord, and the Homeric epics.

Back to ChatGPT:

Good. Now Dick is sad and Jane is trying to cheer him up.

Jane: Hey Dick, I'm sorry to hear about your dog.
Dick: Thanks, Jane. I'm really sad that he's gone.
Jane: I know it's hard to lose a pet. They become such an important part of our lives.
Dick: Yeah, it feels like a part of me is missing now that he's gone.
Jane: I understand. But remember, your dog is in a better place now. He's not in pain or suffering anymore.
Dick: I know you're right. It's just hard to accept.
Jane: Maybe it would help to remember all the happy times you had with your dog. Think about all the good memories you shared and how much joy he brought to your life.
Dick: You're right, Jane. Thank you for trying to cheer me up. It means a lot to me.
Jane: Of course, Dick. I'm here for you. Let me know if you need anything.

I decide to ask about its conversational practice:

In many of these conversations, you began and ended with the same person. Why?

[second try. I forgot to copy the first.]

I began and ended with the same person in many of these conversations because that is the typical structure of a conversation. In a conversation, one person typically speaks, and then the other person responds. This back-and-forth exchange continues until the conversation comes to an end. Beginning and ending with the same person helps to establish the flow of the conversation and makes it feel more natural.

[third try]

In a conversation, it is common for one person to begin and end a conversation, especially if they initiate the conversation in the first place. This can help to create a natural flow to the conversation and ensure that both participants have an equal opportunity to speak and share their thoughts. Additionally, beginning and ending a conversation with the same person can help to establish a sense of closure and completion to the conversation.

I didn’t have any specific expectations about how it would answer. Bit found nothing surprising about this, and, to be honest, I’d have been a bit disappointed if ChatGPT had been nonresponsive. I especially like the phrase, “especially if they initiate the conversation in the first place.” There must have been texts about discourse practices and conventions in the training corpus and that kind of information would have been in those texts. I suspect, moreover, that when most humans make such assertions, they do so because they learned them somewhere, rather than because they’re deduced them from their experience as conversation partners. Why should a chatbot be any different?

What’s important is simply that it can “reflect” on its own discourse. Roman Jakobson called capacity the metalingual function. How sophisticated, if I may inquire, is ChatGPT’s metalingual function?

Now things get interesting:

Now, take the last conversation and embed it within a short story.

Jane walked up to Dick's front door and knocked softly. She could hear him moving around inside and after a moment, he opened the door. His eyes were red and puffy, and she could tell he had been crying.

"Hey, Dick. I'm so sorry about your dog," Jane said, giving him a hug.

"Thanks, Jane. I'm really sad that he's gone," Dick replied, his voice choked with emotion.

"I know it's hard to lose a pet. They become such an important part of our lives," Jane said, trying to offer some comfort.

"Yeah, it feels like a part of me is missing now that he's gone," Dick said, his shoulders slumping.

"I understand. But remember, your dog is in a better place now. He's not in pain or suffering anymore," Jane said, hoping to provide some solace.

"I know you're right. It's just hard to accept," Dick said, sighing.

"Maybe it would help to remember all the happy times you had with your dog. Think about all the good memories you shared and how much joy he brought to your life," Jane suggested.

Dick thought for a moment, then nodded. "You're right, Jane. Thank you for trying to cheer me up. It means a lot to me," he said, giving her a small smile.

"Of course, Dick. I'm here for you. Let me know if you need anything," Jane said, giving him another hug before turning to leave.

As she walked back to her car, Jane couldn't help but feel a sense of sadness for her friend. Losing a pet was never easy, but she was glad she could be there for him during this difficult time.

I was expecting it to add the frame structure, as I’d seen that before (recall it in the last example in my previous post, Of pumpkins, the Falcon Heavy, and Groucho Marx). But I wasn’t expecting it to be so clever about how it handled the dialog itself. It’s eliminated the (awkward) practice of prefacing each statement with the speaker’s name. Instead, each utterance either precedes (1 -7), or follows (9) the framing assertion. Notice what it did in 8, where it is Dick’s turn to speak. It makes a simple assertion ending with a period. Then it gives Dick’s utterance, followed by a qualifying phrase.

As I argued previously, this seems to be clear evidence that ChatGPT has been able to induce high level discourse structure that allows it to treat blocks of text as “fillers” that go into appropriate “slots.” This is characteristic of symbolic systems. Note, that I do not mean to imply that ChatGPT has anything like mastery of symbolic communication. But it does seem to have picked up one aspect of it.

Finally, I just had to ask it what it knew about Jakobson’s metalingual function. So, I logged it, and put the question directly: “What did Roman Jakobson mean by the metalingual function?” After pausing for 10s of seconds or more it replied (in red): “Hmm...something seems to have gone wrong. Maybe try me again in a little bit.” I interpret that as an error message meaning that it had timed out. I made another attempt. Which failed. So:

OK. As Arnold said, "I'll be back."

Yes, the quote "I'll be back" is a famous line said by the character played by Arnold Schwarzenegger in the 1984 movie "The Terminator". In the film, the character utters the line before leaving to carry out a mission, and it has since become a catchphrase associated with Schwarzenegger and the Terminator franchise.

‘Till next time.

Of pumpkins, the Falcon Heavy, and Groucho Marx: High level discourse structure in ChatGPT

Kevin Roose’s New York Times article, The Brilliance and Weirdness of ChatGPT, included an example in which Keith Wynroe prompted ChatGPT to explain AI Alignment in the manner of guy whose thoughts are distracted by his magnificent pumpkins. I decided to follow down that rabbit hole. Here’s the story.

First I present the tweet; then I replicate it, with ChatGPT responding in a way similar, but only so, to the way it responded to Wynroe. Next I insert Elon Musk into the mix; then I drop alignment and add Groucho Marx to Musk. I conclude by discussing higher-order discourse structure in ChatGPT’s replies.

Repeating the original pumpkin tweet

Here is Keith Wynroe’s original tweet. I’ve highlighted some of the lines in the ChatGPT’s reply to reveal the simple back-and-forth form of that reply:

Explain AI Alignment, but write every sentence in the speaking style of a guy who won’t stop going on tangents to brag about how big the pumpkins he grew are

AI alignment refers to the problem of ensuring that an intelligent agent, such as an AI, acts in way that is aligned with our goals and values. Now, I don’t want to brag, but I have to tell you, I grew the biggest pumpkins you’ve ever seen this year. I mean, they were absolutely massive. I’m not exaggerating either, these pumpkins were so big that I had to use a wheelbarrow to move them around. Anyway, back to AI alignment. The idea is that we want to make sure that the AI is doing what we want it to do, not just what we tell it to do. It’s a tricky problem, but I think it’s important. And, hey, speaking of tricky problems, did I mention that I grew the biggest pumpkins you’ve ever seen this year? I’m telling you, they were huge. I even had to build a special trellis in my backyard to support their weight. But yeah, AI alignment is important, and I’m glad we’re talking about it. And hey, if you want to see some really big pumpkins, come by my house sometime. I’ll show you the pumpkins I grew this year. They’re truly impressive.

Here I’ve posted the same prompt, including the lack of a period at the end (for whatever that’s worth):

Explain AI Alignment, but write every sentence in the speaking style of a guy who won’t stop going on tangents to brag about how big the pumpkins he grew are

Sure thing! So, AI alignment is all about making sure that artificial intelligence systems are designed to behave in ways that are beneficial for humanity. Now, let me tell you, I am an expert at growing pumpkins. I mean, these things are massive. Seriously, you won't believe the size of the pumpkins I've been able to grow. Anyway, back to AI alignment. The basic idea is to make sure that AI systems are aligned with our values and goals. Now, I know what you're thinking: "How do you align something as complex as AI with human values?" Well, let me tell you, it's not easy. But let me tell you about the pumpkins I grew last year. I mean, they were HUGE. You could have carved a whole jack-o'-lantern village out of just one of them. Anyway, back to AI alignment. It's a challenging problem, but it's one that we need to solve if we want to harness the power of AI in a way that's safe and beneficial for humanity. And let me tell you, I've been working on growing the biggest pumpkin in the world. I'm talking record-breaking. You won't believe the size of these things when they're ready to harvest. But seriously, AI alignment is an important problem that we need to solve. I mean, just imagine the kind of things we could do with AI if we could align it with our values and goals. The possibilities are endless. And let me tell you, my pumpkins are going to be absolutely massive this year. You won't believe it when you see them.

It's not the same as what Wynroe got, nor did I expect it to be. For one thing, this response is longer, 266 words vs. 200 words. Both have the same back-and-forth structure. This response has 8 segments while the original had 6. None of this seems surprising or remarkable. It is what it is.

Notice the repetition of phrases. Thus “Anyway back to alignment” appears once in the first response and twice in the second. The phrase “let me tell you” appears four times in the second. We have various near repetitions throughout the pair. It seems very formulaic, which it is. As is much of speech and writing.

The Musk variation

Thus it won’t surprise you that, when I replace a pumpkin braggart with Elon Musk bragging about the Falcon Heavy, the result is much the same:

Explain AI Alignment, but write every sentence in the speaking style Elon Musk who won’t stop going on tangents to brag about how the Falcon Heavy rocket.

AI alignment is all about ensuring that artificial intelligence systems are designed to behave in ways that are beneficial for humanity. Now, let me tell you, the Falcon Heavy is a game changer. It's the most powerful operational rocket in the world by a factor of two. But back to AI alignment. The basic idea is to make sure that AI systems are aligned with our values and goals. And let me tell you, the Falcon Heavy can lift more than twice the payload of the next closest operational vehicle, the Delta IV Heavy, at one-third the cost. But seriously, AI alignment is an important problem that we need to solve if we want to harness the power of AI in a safe and beneficial way. And let me tell you, the Falcon Heavy can even launch things all the way to Mars. It's truly impressive. But back to AI alignment. It's a challenging problem, but it's one that we need to solve if we want to use AI to its full potential. And let me tell you, the Falcon Heavy can even carry humans into space. It's an incredible achievement. But seriously, AI alignment is an important topic that we need to keep talking about. And the Falcon Heavy? It's a game changer.

The account of AI alignment is pretty much the same in all three accounts. It is, after all, the same topic. So why not use the same language?

In all three cases we have the same discourse structure, two topics – AI alignment and pumpkins in the first two, AI alignment and the Falcon Heavy in the third – conjoined in a simple alternating pattern. It seems a bit like two people having a conversation. But in this case we have one person conversing within themselves.

A simple frame

When I saw Wynroe’s tweet the phrase “going on tangents” evoked one of those thought light bulbs in me, one saying Tristram Shandy, a book which is known for its digressiveness. So I set out to see if I could guide ChatGPT into some Shandyesque prose. I pressed it hard, even got it to quote passages from the book – though I made no attempt to check, but I can’t say that it produced anything very Shandyesque, which is not suprising, given that Tristram Shandy is one of the greatest books in the English language. But some interesting things happened, including an utterly surprising detour into code-mode, which I may about in another post.

For now, this fragment will have to do:

Now, take what you’ve learned from Tristram Shandy and write a story about a conversation between Elon Musk and Groucho Marx.

Once upon a time, in the land of Silicon Valley, Elon Musk and Groucho Marx had a conversation about the nature of time and space.

"What do you think about time, Groucho?" asked Elon. "Is it a linear thing, or is it something more complex and multifaceted?"

Groucho chuckled. "Ah, time. It's a funny thing, isn't it? It can fly by in the wink of an eye, or it can drag on for an eternity. It all depends on your perspective, I suppose."

"I agree," said Elon. "I've been thinking a lot about time lately, and how it relates to space. Do you think they're connected in some way?"

"Of course they are," said Groucho. "Time and space are like two peas in a pod. They're inseparable, intertwined, and interconnected. Without one, you can't have the other."

"Interesting," said Elon. "I've been working on a rocket that can travel through time and space. I want to explore the universe and see what's out there."

"Well, good luck with that," said Groucho. "Just remember, time and space are not to be messed around with. They can be slippery and elusive, so be careful where you step."

"I will," said Elon. "But I'm excited to see what I'll find out there. Who knows what wonders and mysteries await us in the vastness of the cosmos?"

"That's the beauty of it," said Groucho. "The universe is full of endless possibilities and endless opportunities. The only limit is our imagination, and maybe a few cheesy puns along the way."

And with that, the conversation ended, leaving Elon and Groucho to ponder the mysteries of time and space, and the endless possibilities of the universe.

We have the requested conversation between Elon Musk and Groucho Marx. Are Musk’s lines characteristic Elon? I don’t know: I have little idea of how Musk talks, though I listened to him chat with Joe Rogan for over two hours back in 2018. Are Marx’s lines characteristic Groucho? I’ve heard more from Groucho, starting with his TV quiz shoe in the 1950s. Most mostly recently I’ve listened to an interview with Dick Cavett posted to YouTube. Some phrases ring true, but who knows? Notice that ChatGPT gave Groucho a reference to “cheesy puns,” though, alas, it couldn’t manage to slip any into its response. Notice, finally, that ChatGPT enclosed that conversation within a simple frame, a common story-telling device.

Where did this explicit structure come from, the frame device, and the back and forth of conversation? As you know, and as skeptics keep reminding us, GPT engines are trained to predict the next word. It seems to me, though, that these structures must somehow be defined above the level of the word. Of course the (huge) corpus ChatGPT was trained on would have many examples of conversational alternation and frame structures. But how did it induce those higher-level structures?

Higher-level structure

But how, when it was trained only on predicting the next word, did it somehow manage to isolate those higher-level structures so that it can deploy them in new contexts, such as we see here. For, I assume that there wasn’t a single text alternating between pumpkins and AI alignment in the training corpus, much less a conversation between Musk and Marx surrounded by a simple frame.

The alternation pattern is something like this:

A, B, A, B....

That can be repeated as often as one will. The text in the A sections is always drawn from one body of material while the text in the B sections is drawn from a different body of material. That’s the pattern ChatGPT has learned. Where is it in the net? How’s it encoded.

The frame structure is a bit more complicated:

A (B, C, B, C....) A’

The embedded alternation draws on two bodies of material, any two bodies. The second part of the frame, A’, must complement the first, A.

Again, it’s not a complex structure. But it’s not defined directly over particular words. It’s defined over groups of words, placing the groups, not the individual words, into specified relationships in the discourse string.

Something similar, though perhaps more complex, is going on in the dialogs I we had about Spielberg’s Jaws, his A.I., and Tezuka’s Astro Boy stories. In the case of Jaws we have one body of material, the film itself, and another, the theories of Rene Girard. The Girard material is quite different from the Jaws material. It’s abstract; it’s about human actions and motivation, individually and in groups. In the Jaws dialog I was able to coax it to link the actors and actions required by Girard’s theory to the actors and actions in the film.

In the Astro Boy case the abstract material is from a body of work about AI alignment, which is about how to get (sufficiently powerful) AI’s to respect and act according to human values. First I steered it toward linking the alignment material with the Astro Boy stories, which involve extensive interactions between humans and robots. Then I was able to prompt it into recognizing that, because many of the stories are oppression of robots by humans, that humans own respect to robots, not merely from them (conventional alignment).

The A.I. dialog is the most sophisticated one. Here we have two bodies of abstract material, John Bowlby’s attachment theory, and AI alignment. Here I prompted ChatGPT into recognize that child-rearing is a kind of alignment. Instead aligning a computer to human values, you are aligning a young human to the values of adult humans. And that, in turn, links to the relationship between David, a young cybernetic boy, and Monica, his human mother.

None of this is particularly complex, not for a reasonably competent human in their mid-teens or older. But ChatGPT is not a reasonably competent human of any age. It is a computer program, if perhaps a strange one. For a computer, yes, this is utterly remarkable behavior.

What it bodes for the future, I do not know. In terms of current debates, no, I don’t believe we’re going to arrive at AGI – whatever that is – simply by scaling these machine learning engines to larger and larger sizes. I believe that symbolic computing needs to be incorporated into these systems as well. I see nothing in ChatGPT’s behavior to change my mind on those issues. But I do feel a bit like Shakespeare’s Miranda:

‘Oh wonder!
How many goodly creatures are there here!
How beauteous mankind is! Oh brave new world,
That has such [devices] in’t.’