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! 🥂✨

GPT in the Classroom, Part 2: Escape to America

Time to return to the classroom. I want to return to the question of authorship. First, I present the interaction that resulted in a poem, “Escape to America.” Once we’ve been through that we can ask: Who’s the author? 

* * * * *

The poem, “Escape to America”

I want you to create a sonnet based on three short texts by Miriam Yevick. Use the first text for the first quatrain, the second text for the second quatrain, and the third text for the third quatrain. Create the final couplet as you will. Call the sonnet: Escape to America. Here are the three texts:

The tanks exuded power. The soldiers cast their vigor into the rows of women. The women’s hot breath floated back to them on the cheers. Girls showered flowers and kisses into the warrior’s path. This was the scent of war! The orderly formations loosened, females leaped from the sidewalk and ran into the mêlée. Soldiers broke rank and grabbed at random. The caravan driven on by the motorized units undulated forward amidst frenzied embraces.

Luckily the trains to Paris were still running normally. We were greatly surprised to encounter a city still believing in the protection offered by the Maginot Line (circumvented by the Nazis the previous week) and oblivious to what had happened and was happening some hundred miles to the North.

The next day we took a walk on Broadway. I saw a striking dress in a shop window: sky blue with bold purple stripes descending diagonally from both shoulders to meet along the centerline of the body. I went into the store and bought the dress, which accentuated my youthful shape, for $ 1. Ready-made!

Here is a sonnet, as requested, based on Miriam Yevick’s three texts:

Escape to America

The tanks exuded power, soldiers’ cheers
Roared as their vigor met the women's cries.
Bouquets of kisses, flowers, flung through tears,
Unordered ranks dissolved in frenzy’s guise.

The train to Paris held us; still we sped
Toward streets naïve, in thrall to the Maginot,
Where brittle hopes ignored the northern dread—
The unseen storm that forced us here to go.

And Broadway’s throng—a different world it seemed.
In sky-blue dress, bold purple stripes askew,
I saw myself transformed, as if I dreamed;
Ready-made youth, America’s own hue.

And now, the past fades dim beyond my sight—
A foreign land, forsaken for this light.

I like what you did a lot, but I want two changes: In the first quatrain, somehow indicate that the soldiers are Nazis. Maybe do it by mentioning the swastika. In the third quatrain, I’m not sure about “America’s own hue.” The second line mentions purple stripes, but purple is not particularly identified with America.

Here is the revised sonnet with your requested changes:

Escape to America

The tanks exuded power, swastikas spread,
Their vigor met the women’s frantic cries.
Bouquets of kisses, flowers, flung through dread,
Unordered ranks dissolved in frenzy’s guise.

The train to Paris held us; still we sped
Toward streets naïve, in thrall to the Maginot,
Where brittle hopes ignored the northern dread—
The unseen storm that forced us here to go.

And Broadway’s throng—a different world it seemed.
In sky-blue dress, bold stripes in proud display,
I saw myself transformed, as if I dreamed;
Ready-made youth, new freedoms on my way.

And now, the past fades dim beyond my sight—
A foreign land, forsaken for this light.

* * * * *

But who is the author?

There’s no doubt that ChatGPT, with the aid of FredTheHeretic (aka FTH), generated the words. In what sense is ChatGPT the author? It couldn’t have done it without FTH, which is a GPT, a user-created device that can be installed in ChatGPT so that it can generate specialized output. FTH contains a small collection of poems by Frederick Turner. Does that make Frederick Turner the author? He didn’t actually write the words, but they’re (more or less) in his style.

What about Miriam Yevick? The poem is about her life, her escape from a Europe in the process of being swallowed up by Nazi Germany. Not only is it about her life, but it’s based on her words, words from her memoire, A Testament for Ariela. Surely Yevick deserves some authorship credit.

But how did FTH obtain those words? They didn’t just leap out of the book and into the computer. I selected the passages from the book – they are not contiguous in the text – and presented them to FTH. When FTH’s first try was a little wonky, I suggested changes. Surely I deserve some credit.

So far we’ve got ChatGPT, FTH, Miriam Yevick, and me. ChatGPT didn’t spontaneously emerge into existence one day when a computer had some CPU cycles to spare. It was created by a team of programmers, engineers, data scientists, and technicians at OpenAI. And FTH was created by Paul Fishwick and his graduate students at the University of Texas at Dallas. How do we credit these people?

This is not a new problem. The motion picture industry has been up against it for years and has evolved a rather elaborate set of conventions for doling out credit, credits negotiated with the various parties, both individuals, and organizations, involved. I don’t intend to propose a solution in this case. But the problem is now here and we’re going to have to deal with it.

One final point: It seems to me that denying “Escape to America” is meaningful because the words were actually produced by a computer, using that as an excuse to assert that it’s not a poem, that’s bone-headed, stupid, and short-sighted. We’ve got intellectual work to do.

GPT in the Classroom, Part 1: Five Sonnets on the Gettysburg Address

I realize that, with ready access to sophisticated chatbots, some students will use them to complete writing assignments rather than do the writing themselves. Since I no longer teach, however important that issue is, it is not a problem for me. I set it aside. That is not what this essay is about. I’ve got other fish to fry.

What opportunities does the existence of competent AI-generated literary texts present for teaching about the nature of texts? An obvious opportunity is to consider the difference between human-created and AI-generated texts. But that doesn’t interest me, not here and now. It’s too difficult. I am willing to take these texts at face value, giving them no favor or disfavor.

* * * * *

This is the first in a series of three posts in which I take up other issues. In this post I present five AI-generated sonnets based on the text of “Lincoln’s Gettysburg Address” and deal with the question of authorship. In the second essay I am going to present a poem based on a humorous story that Pope Francis told in a recent New York Times op-ed. In the third and final essay I once again take up the issue of authorship, using a sonnet based on three paragraphs from Miriam Yevick’s A Testament for Ariela.

The Five Sonnets

A GPT is a user-created device that can be installed in ChatGPT so that it can generate specialized output. These sonnets are generated by a GPT known as FredTheHeretic, aka FTH. It contains a small collection of poems by Frederic Turner that is uses as models and templates.

Here is the prompt I gave FTH:

Give me a sonnet based on the following text:

Four score and seven years ago our fathers brought forth on this continent, a new nation, conceived in Liberty, and dedicated to the proposition that all men are created equal.

Now we are engaged in a great civil war, testing whether that nation, or any nation so conceived and so dedicated, can long endure. We are met on a great battle-field of that war. We have come to dedicate a portion of that field, as a final resting place for those who here gave their lives that that nation might live. It is altogether fitting and proper that we should do this.

But, in a larger sense, we can not dedicate—we can not consecrate—we can not hallow—this ground. The brave men, living and dead, who struggled here, have consecrated it, far above our poor power to add or detract. The world will little note, nor long remember what we say here, but it can never forget what they did here. It is for us the living, rather, to be dedicated here to the unfinished work which they who fought here have thus far so nobly advanced. It is rather for us to be here dedicated to the great task remaining before us—that from these honored dead we take increased devotion to that cause for which they gave the last full measure of devotion—that we here highly resolve that these dead shall not have died in vain—that this nation, under God, shall have a new birth of freedom—and that government of the people, by the people, for the people, shall not perish from the earth.

Here are the five sonnets it generated:

The Battle-Field: A Sonnet

Upon this ground where blood and tears have lain,
A nation forged in liberty still stands.
These fields once drank the courage of brave hands,
Yet whisper now of loss, enduring pain.

No words of ours could sanctify this space;
The deeds of those who fell surpass our breath.
Their valor, richer than the soil of death,
Ensures that time cannot their work erase.

We bear the burden, those who linger here,
To finish what their sacrifice began,
To weave a union, broad and strong and clear,
Upholding freedom's light for every man.

Let not their purpose fade, nor courage wane;
That dream they died for shall not die in vain.

The Unfinished Work

Here lies the proof of freedom’s fragile dream,
A nation’s heart laid bare in sacrifice.
The dead, whose silent deeds beyond suffice,
Have sown with blood what words can only seem.

No stone can hallow what their courage sealed,
No tongue can add to what they’ve dared and done.
The work remains; their labor is begun,
And calls us now upon this battle-field.

Shall we, the living, falter in this task,
Or turn away from what their lives bestowed?
Or shall we bear the torch, their honor’s load,
And in their shadow, shed our timid mask?

A government, by us, for us, must rise,
Lest freedom fall and perish where it lies.

Time for another ramble: Melancholy, Claude, Bloom, Claude, Ring composition, and Other Stuff

It’s been a while since I’ve done one of these; May 29th was the last one. If you look over there to right at the Blog Archive you’ll see I’ve been on a posting slump, with 3-figure monthly totals from January through June, then a dip to 61 for July, August: 18, September: 15, October: 30, and now 33 for November as I write this, and the month isn’t over. Maybe I’m pulling out of the slump.

Anyhow, I’m feeling a little backed up with things to post about, so it’s time to ramble on and see what’s up.

Melancholy, Mind (Mine), and Growth

That’s the tentative title for my next 3 Quarks Daily article. Starting back in November 2017 I’ve been making occasional posts about my monthly posting habits, which tend to drop during the winter. I’m thinking of using that as the point of departure for my next 3QD piece, which will go up on December 2nd.

During those down times I’m depressed to one degree or another (melancholy). But why? Since those down times have been in the winter, perhaps its seasonal affective disorder (SAD). But that doesn’t square with all of the evidence. There was no down-time in the winder of 2022-2023 and 2023-2024, but there was a slump in the summer of 2023. Something else is going on, and I think it has to do with creativity. To that end I want to discuss the ridiculous blither of tags here, 665 by November or 2023.

Claude

I’ve starting working with Claude, Anthropic’s chatbot. I want to do some posts where I verify some of the work I’ve done with ChatGPT. I’m thinking of posts on stories, ontological structure, abstract definition, and the Girardian analysis of Jaws. I can then gather those into a working paper.

I also want to look at other things. At the moment I’m thinking of seeing how Claude summarizes longish documents. I’m thinking of the Hamlet chapter from Bloom’s Shakespeare book and Heart of Darkness.

Harold Bloom and GOAT literary critics

A year ago I began a series of posts on the theme of the greatest literary critics. I got bogged down in discussing Harold Bloom. It’s time to finish it off.

Bloom may well be as brilliant a literary critic as we’ve had in the last 50 or 60 years. But brilliance is one thing, greatness is another. Brilliance is a function of the individual, while greatness is a function of the relationship between an individual’s work and the arena in which they’re working.

I’m not sure about Bloom’s fit. While he’s got a wide readership, it’s not clear to me that scholars have taken up his work in any significant way. They may cite him – perhaps especially is concept of influence – but that they don’t much use of his ideas in his work. But we’ve also got to consider his work in the larger public arena, where he is hands-down the most prominent literary critic. I’m not sure of how to handle that.

However, if History wants to declare that Harold Bloom is one of the great all-time literary critics, maybe even the GOAT, what do I care? What would really bother me is if future critics should decide to take his work as a model and (attempt to) do more like it. Like most literary critics he’s neglected the study of form and he’s been deaf to the cognitive sciences. There’s little in his work that’s worth amplifying. It’s a dead end.

ChatGPT report

About a year or so ago I started writing a report summarizing my work on ChatGPT. I need to finish that report. I’d estimate the three-fourths or more are done. I’d like to be able to include some work with Claude. I don’t intend a lot on this, just enough to say that I’ve verified some things.

I’d like to finish this by the end of this year.

Why’s ring-form composition important?

That’s tricky. It has to do with the fact that literary works are extended in time, unlike the visual arts, which are static in time but extended in space. You can’t take the whole thing in at a glance like you can a painting.

There are constraints on how a literary work can unfold in time. There is a sense in which (the nature of) the end is inherent in the beginning. Ring-compositions are even more tightly constrained. Dylan Thomas consciously and deliberately plotted the ring-composition of the rhyme scheme in his “Author’s Prologue.” Rhyme is not about meaning; its patterns are arbitrary with respect to meaning. But Coleridge did not consciously work out the ring-compositions in “Kubla Khan,” not Conrad in Heart of Darkness. These patterns ARE NOT arbitrary with respect to meaning. On the contrary, they are central to how meaning is constituted.

Other Stuff

More Cobra Kai: Follow-up on my post where I explore the Freudian angle, saying a bit more about Girard, and extending that into history.

* *

More on meaning in LLMs: I’ve suggested that Ilya Sutskyver conflates mistakenly (and unknowingly?) conflates cognitive and semantic structure with the structure of the world and so suggests that robust next-token prediction requires knowledge of the world. In this post I argue that making that distinction is, in fact, difficult, and involves what I’ve been calling the word illusion. I first confronted the problem as an undergraduate when I was trying to understanding the difference between the signified, and mental structure, and the reference, something in the world, of a sign. I may not have gotten deep intuitions about that until I began studying cognitive networks in graduate school.

* *

Gila-monster venom and computational irreducibility: The idea is to start with a NYTimes article on drug discovery that starts with Gila-monster venom and ends up with Wolfram’s concept of computational irreducibility. This is about search, computation, and the complex and irregular structure of the (natural world).

* *

My work with Ramesh: Notes on conceptual ontology and hypergraphs in conceptual space.

* *

LLMs and literary study: Can we use LLMs to analyze the thematic structure of literary texts?

In particular, can we use them to examine Bloom’s these about Shakespeare as “inventing” the human. Are there themes that appear first in Shakespeare? We need more than Bloom’s vigorous assertion on this. We need to examine the thematic structure of prior texts and of Shakespeare’s texts and show that there are things new in Shakespeare. Do those new things then continue if texts after Shakespeare? To do this properly we need to examine a lot of texts.

I’d also like to know if LLMs could be used to find ring-composition in literary texts. It is by no means obvious to me that they can.

How far can next-token prediction take us? Sutskever vs. Claude

One of my main complaints about the current regime in machine learning is that researchers don’t seem to have given much thought to the nature of language and cognition independent from the more or less immediate requirements of crafting their models. There is a large, rich, and diverse literature on language, semantics, and cognition going back over a half century. It’s often conflicting and thus far from consensus, but it’s not empty. The ML research community seems uninterested in it. I’ve likened this to a whaling voyage captained by a man who knows all about ships and little about whales.

As a symptom of this, I offer this video clip from a 2023 conversation between Ilya Sutskever and Dwarkesh Patel in which next-token prediction will be able to surpass human performance:

Here's a transcription:

I challenge the claim that next-token prediction cannot surpass human performance. On the surface, it looks like it cannot. It looks like if you just learn to imitate, to predict what people do, it means that you can only copy people. But here is a counter argument for why it might not be quite so. If your base neural net is smart enough, you just ask it — What would a person with great insight, wisdom, and capability do? Maybe such a person doesn't exist, but there's a pretty good chance that the neural net will be able to extrapolate how such a person would behave. Do you see what I mean?

Dwarkesh Patel

Yes, although where would it get that sort of insight about what that person would do? If not from…

Ilya Sutskever

From the data of regular people. Because if you think about it, what does it mean to predict the next token well enough? It's actually a much deeper question than it seems. Predicting the next token well means that you understand the underlying reality that led to the creation of that token. It's not statistics. Like it is statistics but what is statistics? In order to understand those statistics to compress them, you need to understand what is it about the world that creates this set of statistics? And so then you say — Well, I have all those people. What is it about people that creates their behaviors? Well they have thoughts and their feelings, and they have ideas, and they do things in certain ways. All of those could be deduced from next-token prediction. And I'd argue that this should make it possible, not indefinitely but to a pretty decent degree to say — Well, can you guess what you'd do if you took a person with this characteristic and that characteristic? Like such a person doesn't exist but because you're so good at predicting the next token, you should still be able to guess what that person who would do. This hypothetical, imaginary person with far greater mental ability than the rest of us.

The argument is not clear. One thing Sutskever seems to be doing is aggregating the texts of ordinary people into the text of an imaginary “super” person that is the sum and synthesis what all those ordinary people have said. But all those ordinary individuals do not necessarily speak from the same point-of-view. There will be tensions and contradictions between them. The views of flat-earthers cannot be reconciled with those of standard astronomy. But this is not my main object. We can set it aside.

My problem comes with the Sutskever’s second paragraph, where he says, “Predicting the next token well means that you understand the underlying reality that led to the creation of that token.” From there he works his way through statistics to the thoughts and feelings of people producing the tokens. But Sutskever doesn’t distinguish between those thoughts and feelings and the world toward which those thoughts and feelings are directed. Those people are aware of the world, of the “underlying reality,” but that reality is not itself directly present in the language tokens they use to express their thoughts and feelings. The token string is the product of the interaction between language and cognition, on the one hand, and the world, on the other:

Sutskever seems to be conflating the cognitive and semantic structures inhering in the minds of the people who produce texts with the structure of the world itself. They are not at all the same thing. A statistical model produced through next-token prediction may well approximate the cognitive and semantic models of humans, but that’s all it can do. It has no access to the world in the way that the humans do. That underlying reality is not available to them. 

Now, it may well be the case that the structure of human discourse reflects, is somehow caused by, structure in the world. But the transformer engine producing the language model only has access to those texts, not to the world they reflect. The structure it is predicting is the structure in the texts, not the world, though it takes a bit of work to convince Claude of that.

What does Claude have to say about this?

I gave Claude 3.5 Sonnet Sutskever’s second paragraph and had a conversation about it. I wanted it to see if it could spot the problem. Claude saw various problems, but couldn’t quite find its way to what I regard as the crucial point. In the end I had to tell it that Sutskever failed to distinguish between the structure of the world and the structure of the semantic and cognitive structure expressed by the text.

My text is set in bold Sofia Sans while Claude's is plain Sofia Sans.

Fatherhood changes the brain and brings meaning and purpose

Darby Saxbe, Dad Brain Is Real, and It’s a Good Thing, NYTimes, June 15, 2024.

The brain and hormonal changes we observe in new dads tell us that nature intended men to participate in child-rearing, because it equipped them with neurobiological architecture to do so. They too can show the fundamental instinct for nurturing that’s often attributed solely to mothers.

Not only that, but men’s involvement in fatherhood can have long-term benefits for their brain health — and for healthy societies. At a time when boys and men seem to be experiencing greater social isolation and declining occupational prospects, the role of father can provide a meaningful source of identity. [...]

In a 2022 study, my colleagues and I collaborated with researchers in Spain to gather brain scans of a small number of first-time fathers before and after their babies were born. Our results echoed studies of mothers done by some of the same researchers. In several landmark studies, they found that as women became mothers, their brains lost volume in gray matter, the layer of brain tissue rich with neurons, in regions across the brain, including those responsible for social and emotional processing.

Although a shrinking brain sounds like bad news, less can be more: These changes might fine-tune the brain to work more efficiently. [...] Women who lost more brain volume showed stronger attachment to their infants after birth, indicating that the shrinkage promoted bonding.

Our findings for fathers were similar. Men also lost gray matter volume in new fatherhood, in some of the same regions that changed in women. But volume reductions for dads were less pronounced.

Fatherhood brings meaning and purpose:

Even so, most fathers tell us that they derive tremendous meaning and purpose from their connection to their children. Contemporary fathers are almost as likely as mothers to say that parenthood is central to their identity, and men are even more likely to report that children improve their well-being than women are. And the newest data suggests that parenting may ultimately promote long-term brain health; among older men and women, a brain-age algorithm estimated that the brain looked younger among people who had children.

There's more at the link. I've linked the original research below the asterisks.

* * * * *

Magdalena Martínez-García, María Paternina-Die, Sofia I Cardenas, Oscar Vilarroya, Manuel Desco, Susanna Carmona, Darby E Saxbe, First-time fathers show longitudinal gray matter cortical volume reductions: evidence from two international samples, Cerebral Cortex, Volume 33, Issue 7, 1 April 2023, Pages 4156–4163, https://doi.org/10.1093/cercor/bhac333

Abstract: Emerging evidence points to the transition to parenthood as a critical window for adult neural plasticity. Studying fathers offers a unique opportunity to explore how parenting experience can shape the human brain when pregnancy is not directly experienced. Yet very few studies have examined the neuroanatomic adaptations of men transitioning into fatherhood. The present study reports on an international collaboration between two laboratories, one in Spain and the other in California (United States), that have prospectively collected structural neuroimaging data in 20 expectant fathers before and after the birth of their first child. The Spanish sample also included a control group of 17 childless men. We tested whether the transition into fatherhood entailed anatomical changes in brain cortical volume, thickness, and area, and subcortical volumes. We found overlapping trends of cortical volume reductions within the default mode network and visual networks and preservation of subcortical structures across both samples of first-time fathers, which persisted after controlling for fathers’ and children’s age at the postnatal scan. This study provides convergent evidence for cortical structural changes in fathers, supporting the possibility that the transition to fatherhood may represent a meaningful window of experience-induced structural neuroplasticity in males.

Body part names across cultures

Variability in human body vocabularies. A team of linguists from @MPI_EVA_Leipzig and @UniPassau has conducted the first large-scale comparison of body part vocabularies across 1,028 #languages, using the #Lexibank database. See: https://t.co/YJfIGvtIUH & https://t.co/X81jSDweAY pic.twitter.com/WH4qH8je9z

— MPI-EVA Leipzig (@MPI_EVA_Leipzig) May 13, 2024

Abstract for the linked article:

Every human has a body. Yet, languages differ in how they divide the body into parts to name them. While universal naming strategies exist, there is also variation in the vocabularies of body parts across languages. In this study, we investigate the similarities and differences in naming two separate body parts with one word, i.e., colexifications. We use a computational approach to create networks of body part vocabularies across languages. The analyses focus on body part networks in large language families, on perceptual features that lead to colexifications of body parts, and on a comparison of network structures in different semantic domains. Our results show that adjacent body parts are colexified frequently. However, preferences for perceptual features such as shape and function lead to variations in body part vocabularies. In addition, body part colexification networks are less varied across language families than networks in the semantic domains of emotion and colour. The study presents the first large‐scale comparison of body part vocabularies in 1,028 language varieties and provides important insights into the variability of a universal human domain.

ChatGPT explores the metalingual definition of some concepts from science [MTL]

“Metalingual definition” is the term David Hays gave to the linguistic structure and process whereby some term, called the definiendum, is defined by a string of words, the definiens, where the string of words typically takes the form of a story. Charity was his standard teaching example: Charity (the definiendum) is when someone does something nice for someone without thought of reward (the definiens). Any particular story that matches that pattern is an example of charity. In this conception, charity inheres in the entire story; more precisely, in the pattern exhibited by that story. Note however that the definiens need not take the form of a story, though that may be the most perspicuous form. The point is simply that a pattern of words can be used to define a term.

The power of the construction becomes apparent when you realize that the definiens of any term can itself contain metalingually defined terms, each of which has its own definiens, and so on recursively through a large swath of the word stock. Thus in Hays’s standard example, charity, the term reward appears in the definiens. That term is abstract, and so must be linked to a definiens of its own.

From time to time I have investigated ChatGPT’s capacity explicitly to deal with metalingual definition. Here, for example, is a post in which I explore several such constructs with ChatGPT, Abstract concepts and metalingual definition: Does ChatGPT understand justice and charity?

This post collects some examples from science. My typical convention is to set my prompt in boldface and ChatGPT’s response in plain face.

* * * * *

Define gravity.

Gravity is a natural force that attracts any two objects with mass toward each other. The force of gravity is proportional to the product of the masses of the objects and inversely proportional to the square of the distance between them. This is known as the Law of Universal Gravitation, which was first described by Sir Isaac Newton in the late 17th century.

The force of gravity is what keeps planets in orbit around the sun, moons in orbit around their planets, and galaxies held together. It also causes objects to fall toward the ground when dropped.

Gravity is one of the four fundamental forces of nature, along with the strong nuclear force, the weak nuclear force, and the electromagnetic force. It is a non-contact force, meaning that it acts on objects at a distance without coming into direct contact with them.

Einstein's theory of general relativity, developed in the early 20th century, provides a more accurate and comprehensive description of gravity as a curvature of spacetime caused by the presence of matter and energy. This theory explains phenomena such as gravitational lensing and the behavior of black holes that are difficult to understand under the Newtonian theory of gravity.

Define entropy.

Entropy is a measure of disorder or randomness in a system. It is typically represented by the symbol S and has units of energy per temperature (typically joules per kelvin in SI units).

In thermodynamics, which is the branch of physics that deals with the relationships between heat, energy, and work, entropy is a measure of the amount of thermal energy in a system that is unavailable to do work.

The Second Law of Thermodynamics states that in any energy transfer or transformation, the total entropy of a closed system will always increase over time, this is also known as entropy increase principle.

In more general terms, entropy can also refer to the disorder, randomness, or uncertainty in a system. In information theory, entropy is a measure of the amount of uncertainty or randomness in a random variable. In statistical mechanics, entropy is a measure of the number of microscopic configurations of a system that correspond to a given macroscopic state.

Entropy, in the context of Thermodynamics is often referred to as a state function, which means that it depends only on the current state of the system, not on how that state was achieved.

Define atomic bond.

An atomic bond is the force that holds two or more atoms together to form a chemical compound. There are several types of atomic bonds, each with their own characteristics and properties. The main types of atomic bonds are:

• Covalent bonds: Covalent bonds form when atoms share one or more pairs of electrons. This type of bond is typically found in compounds made up of non-metals.
  
• Ionic bonds: Ionic bonds form when an electron is transferred from one atom to another. This type of bond is typically found in compounds made up of a metal and a non-metal.
  
• Metallic bonds: Metallic bonds form when atoms share a sea of electrons. This type of bond is typically found in pure metals and alloys.
  
• Hydrogen bond: Hydrogen bond is a type of dipole-dipole interaction, a weaker type of force that occurs between a hydrogen atom covalently bonded to a highly electronegative atom such as nitrogen, oxygen, or fluorine and another highly electronegative atom nearby.

All of these bonds involve the sharing or transfer of electrons between atoms, which results in the formation of a more stable configuration known as a chemical compound. The strength of these bonds depends on the nature of the atoms involved, their electronic configurations, and the distance between them.

The properties of the chemical compound formed will be determined by the type of bond and the arrangement of atoms that holds it together.

Define momentum.

Momentum is a measure of the motion of an object. It is defined as the product of an object's mass and velocity. Mathematically, the momentum of an object (p) is given by the formula:

p = m * v

where m is the mass of the object and v is its velocity. The unit of momentum is typically kilogram meters per second (kg·m/s) in the International System of Units (SI).

The momentum of an object is a vector quantity, meaning it has both a magnitude and a direction. The direction of momentum is the same as the direction of velocity. When an object is at rest, its momentum is zero.

Momentum is a measure of the "inertia" of an object in motion, meaning how difficult it is to change an object's speed or direction of motion. The larger the mass of an object or the greater its velocity, the greater its momentum.

In physics, the principle of conservation of momentum states that in a closed system, the total momentum of all objects before and after a collision or other interaction will remain the same, unless acted upon by an external force.

momentum is an important concept in physics, specially in mechanics, that plays a role in understanding how objects move, and it is widely used in fields like physics, engineering, and aeronautics.

Define gene.

A gene is a functional unit of DNA that contains the instructions for the synthesis of a specific protein or RNA molecule. It consists of a specific sequence of nucleotides, the building blocks of DNA, which are located on chromosomes.

Genes are inherited from our parents and play a central role in determining our physical and behavioral characteristics, such as eye color, hair color, height, blood type and many others. Each gene carries specific information for the synthesis of a particular protein, and the collection of all genes in an organism, known as its genome, contains the complete set of instructions for the development and function of that organism.

The proteins made by a gene play a vital role in the body, acting as enzymes, structural molecules, hormones, and other key components. They also help in controlling when and where certain genes are expressed.

Genes can also be studied in the context of evolution, where variations in genes can lead to different traits that can be beneficial or harmful to an organism's survival. Some variations can be passed on to next generations, leading to natural selection and genetic drift.

Mutations, changes in the DNA sequence of a gene, can also occur and have effects on the proteins it codes for and the way the gene is expressed, these mutations can have positive or negative effect, some being benign and others causing diseases.

What is a molecule.

A molecule is a chemical structure composed of two or more atoms that are chemically bonded together. The atoms in a molecule are held together by chemical bonds, which are forces that result from the attractive or repulsive interactions between the electrons of different atoms. Molecules can be composed of any type of atom, and can vary greatly in size and complexity.

The basic building block of a molecule is an atom and the smallest molecule is diatomic (made up of two atoms, such as H2,O2). Larger molecules can be made up of many atoms and can be quite complex, such as DNA and proteins.

Molecules can also be characterized by their physical properties such as boiling and melting point, solubility, density, and others. The properties of a molecule are often determined by the atoms that make it up, the arrangement of those atoms, and the types of bonds that hold the atoms together.

Molecules can be broadly classified into two groups: organic and inorganic. Organic molecules are those that contain carbon, while inorganic molecules are those that do not contain carbon. Organic molecules are found in living organisms and include compounds such as carbohydrates, lipids, nucleic acids, and proteins, while inorganic molecules include compounds such as water, gases, and minerals.

Molecules play an important role in many chemical processes, including those that occur within living organisms, and are essential for the study of chemistry and biochemistry.

ChatGPT on a Remark Leonard Bernstein Made to Duke Ellington [intention, intuition, and tacit knowledge]

I have been arguing recently that meaning consists of intentionality and semanticity, where intentionality inheres in the relationship between the interlocutors (even if they are not face-to-face and synchronous) and semanticity inheres in the linguistic/cognitive system. Semanticity, in turn, consists of relationality and adhesion, where relationality is the relationships words have among themselves and adhesion inheres in the perceptual and motor linkages words have with the external world. In this view, meaning in LLMs is a matter of relationality.

I recently performed an experiment which bears on ChatGPT’s capacity to approximate intentional meaning in a conversation. In this case, a conversation between Duke Ellington and Leonard Bernstein. Back in 1966 the two had a conversation that was televised on station WTMJ-TV. Here’s the video clip:

I wrote a blog post where I included a clip of the conversation along with commentary and transcriptions of bits here and there. At about 05:53 Bernstein remarked to Ellington: “...you wrote symphonic jazz and I wrote jazz symphonies.” The two men then shook hands.

In my commentary I wrote that the actual assertion was not very meaningful, and that both men surely knew it, but that much was conveyed indirectly. That meaning derived from the fact that both men were, in effect, acting as representatives of two cultural traditions and so Bernstein’s remark, and Ellington’s acceptance, had the weight of a cultural negotiation.

I was curious about how much of that ChatGPT would be able to pick up. So I transcribed a bit of the conversation up to that point and asked the Chatster what was going on. Here’s that interaction. My prompts are in boldface while ChatGPT's responses are plain face (except for bolded topics on numbered lists).I’ve inserted some comments which I’ve aligned to the right-hand margin and highlighted thus. At the end of the interaction I have some general remarks about ChatGPT’s performance, including some remarks about ChatGPT’s lack of intuitive or tacit knowledge.

* * * * *

In 1966 Duke Ellington and Leonard Bernstein were interviewed on television. They were asked to talk about the state of music in America. I am going to give you part of that conversation, including the interviewer, and then I want to ask you about it.

Here’s the interview:

Leonard Bernstein: As a matter of fact there's tremendous diversity taking place. Diversification I should say because it's in the process of happening. But there has been a tremendous spread away from the metropolises to smaller cities to university campuses communities of various sorts and sizes.

I mean popularly referred to as the cultural explosion which is a word that occasions some dismay among the higher brow critics. But it's a thing that I am very proud of and happy to see happening because it's not a fake explosion. It's a real one and it's penetrating everywhere everywhere in the country.

Interviewer: Mr. Ellington how do you feel on the same point?

Duke Ellington: Well, I'm very happy to know that the american audience is maturing.

Inteviewer: And maturing you feel across the country.

Ellington: Oh yes.

Interviewer: What evidence do you find in your concerts of this?

Ellington: Well ours is a rather a separate one. It's in that, you know, everybody gets their own audience more or less. And because I was just spread over a long period you know. It's we find people who met when we were playing "Mood Indigo" at a prom, or the baby was born when "Caravan" came out or "Don't Get Around much Anymore," they were seeing husband after war, or something like. We cover a lot of different generations, you know, and so it's uh you just sort of automatically have an appraisal of the scene.

Bernstein: But wouldn't you say, Duke, that one of the great signs of this explosion I was talking about is that we really share the same audiences to a much greater extent than we used to. I mean, as you say, you have your own audience. Well that used to be true and I suppose our kind of music, if you could say such a thing, used to have its own audience. But now you can't tell them apart anymore. I mean the same kids and the same excited people who come to hear you, come to hear me and that's something new.

Ellington: What what I have been trying to, in the talking of, seeing and trying to de-categorize this American, it's American music I would say. Oh the stuff that we're in, anyway. And it's getting to the point now where the modern contemporary composer and the guy who's supposed to be a modern jazz composer, they all come out of the same conservatories.

Bernstein: Exactly.

Ellington: And it's very difficult to find a place to draw the line.

Bernstein: Well you were certainly one of the pioneers in that.

Ellington: Oh yeah, but i didn't come out of the conservatory.

Bernstein: No. but you one of the the first people who wrote so-called symphonic jazz.

Ellington: I had a conservatory in the capital theater.

Bernstein: That's right exactly. Sit there and listen to the symphony before the picture. Maybe that's really the difference between us. That you wrote symphonic jazz and I wrote jazz symphonies.

Ellington laughs. They shake hands.

This conversation between Duke Ellington and Leonard Bernstein offers valuable insights into the evolving landscape of American music during their time. Here's a breakdown of some key points:

1. Diversification and Cultural Explosion: Bernstein highlights the significant diversification happening in American music, particularly away from major cities and into smaller communities and campuses. This expansion, often termed a "cultural explosion," is viewed positively by both musicians.

2. Maturing Audience: Ellington expresses satisfaction with the maturation of the American audience, indicating a broadening appreciation for music across the country.

3. Shared Audiences: Bernstein observes that the lines between different genres of music are blurring, with audiences for jazz and classical music increasingly overlapping. This suggests a more unified appreciation for music regardless of genre distinctions.

4. De-categorization of American Music: Ellington emphasizes the emergence of a unified American musical identity, transcending traditional genre boundaries. He notes the difficulty in categorizing music into distinct genres, as contemporary composers and jazz musicians share similar backgrounds and influences.

5. Symphonic Jazz vs. Jazz Symphonies: Bernstein and Ellington playfully discuss their contributions to music, with Ellington credited as a pioneer of "symphonic jazz" and Bernstein likening his own compositions to "jazz symphonies." This exchange highlights the innovative approaches each artist brought to their respective styles.

Overall, the conversation reflects a period of significant change and artistic exploration in American music, with both Ellington and Bernstein contributing to its evolution in distinct yet interconnected ways.

That’s a decent summary of the things they talked about, but it completely misses the significance of Bernstein’s remark to Ellington. So I asked ChatGPT directly about that.

Our Need for Meaning and Coherence

Steven J Heine, Travis Proulx, Kathleen Vohs, The Meaning Maintenance Model: On the Coherence of Social Motivations, Personality and Social Psychology Review 10(2):88-110, February 2006, DOI: 10.1207/s15327957pspr1002_1.

Abstract: The meaning maintenance model (MMM) proposes that people have a need for meaning; that is, a need to perceive events through a prism of mental representations of expected relations that organizes their perceptions of the world. When people's sense of meaning is threatened, they reaffirm alternative representations as a way to regain meaning-a process termed fluid compensation. According to the model, people can reaffirm meaning in domains that are different from the domain in which the threat occurred. Evidence for fluid compensation can be observed following a variety of psychological threats, including most especially threats to the self, such as self-esteem threats, feelings of uncertainty, interpersonal rejection, and mortality salience. People respond to these diverse threats in highly similar ways, which suggests that a range of psychological motivations are expressions of a singular impulse to generate and maintain a sense of meaning.

Does ChatGPT Understand the Concept of Tragedy? Symbolic AI & Neural Nets

New working paper. Title above, links, abstract and contents below:

• Academia.edu: https://www.academia.edu/116049941/Does_ChatGPT_Understand_the_Concept_of_Tragedy_Symbolic_AI_and_Neural_Nets
• Social Science Research Network (SSRN): https://ssrn.com/abstract=4754558
• ResearchGate: https://www.researchgate.net/publication/378858260_Does_ChatGPT_Understand_the_Concept_of_Tragedy_Symbolic_AI_Neural_Nets

Abstract: Tragedy is an abstract concept. David Hays proposed that abstract concepts can be metalingually defined. A term, such as tragedy, can be given its meaning by a pattern over a string of words. By using a cognitive network to define tragedy Brian Phillips investigated that computationally. In this paper I have taken Brian Phillips’ definition and presented it to ChatGPT in six different trials, where each trial also included a short story. Three stories satisfied the definition and three did not. ChatGPT was asked to indicate whether or not the story satisfied the definition it was given. It was correct in each case.

CONTENTS

Symbolic Computation and Abstract Definition 2
Tragedy Defined 2
ChatGPT Confronts “Tragedy” in Six Trials 4

John Sowa on meaning in language and LLMs

 

An exercise for the reader: 

Background: I have argued that meaning involves intention, relationality, and adhesion, where intention inheres in the relationship between the parties to a linguistic act and relationality and adhesion together make up semanticity, which inheres in the language system itself. Relationality exists in the relationships of words among themselves and is the basis of inferential semantics, as it's called in the literature. Adhesion exists in the relationship between a word and the situation in the world designated by the word.

The exercise: Sowa gives a number of examples between 02:29 and 13:10. Examine those various examples and assess the roles of intention, relationality, and adhesion in each.

AI and intellectual integration @3QD

I have a new article up at 3 Quarks Daily:

Western Metaphysics is Imploding. Will We Raise a Phoenix from The Ashes? [Catalytic AI], https://3quarksdaily.com/3quarksdaily/2024/02/western-metaphysics-is-imploding-will-we-raise-a-phoenix-from-the-ashes-catalytic-ai.html

It is about philosophy, though not philosophy as it is currently practiced as an academic discipline. I like it. In fact I like it a lot.

Why? Because it’s built on a number of articles I’d previously published in 3QD as well as other work I’d published about AI over the past year, ChatGPT in particular. When I finally posted it on Sunday afternoon, it felt good, really good. “Man, I’ve got something here,” said I to myself.

When I got up early Monday morning, so early one might call it late Sunday night, I looked at the article and started glancing through it. “Holy crap!” thought I to myself, “when people start reading this they’re going to think they’ve landed in one of those classic New Yorker essays that wander all over the place before getting to the point, if there is one. What happened?”

Those are two very different reactions: “I’ve got something here” vs. “Holy crap!” Conclusion: I’ve got some work to do.

I might as well begin here and now.

Meaning, intention, and AI

One of my friends remarked, “you are too smart for me.” I took that to be a polite and diplomatic way of saying that he figured there must be something there but he sure couldn’t find it. How’d I get from his remark to that interpretation? I can tell you want it didn’t involve: conscious, deliberate thought. I simply knew that’s what he was saying. I intuited the intention behind my friend’s words, an intention that I’ve subsequently verified.

Intentionality – closely related to but not quite the same as intention – is at the heart the classic argument against AI. As far as I know that argument was first articulated by Hubert Dreyfus back in 1969 or 71’, in that time frame, but is probably best-known from John Searle’s Chinese Room argument, which first appeared in 1980 in Behavioral and Brain Sciences. That argument has been refitted for the current era, perhaps most visibly by Emily Bender, who coined the phrase “stochastic parrot” to characterize the actions of Large Language Models (LLMs).

I accept that argument. The problem is, however, that it’s one thing to have made that argument at a time when AI systems responded to human input in a relatively simple and straightforward way, which was the case when Dreyfus and Searle made their arguments. Back then the argument supplied a fairly satisfying – at least to some people – account of why AI won’t work. Now, in the face of ChatGPT’s much more impressive performance, you are asking a lot more from that argument, more, I’ve argued elsewhere, more than it can reasonably deliver.

The issue here is the gap between our first-person experience of the machine and what the machine is actually doing. Back in Searle’s time the philosophical concept of intentionality was able to account for that gap, at least for some of those familiar with the concept. In the case of ChatGPT the nature of that gap is quite different. To a first approximation, our first-person experience is that we’re conversing with a person that has a strange name, ChatGPT. Some people have strange names and stilted discourse is not uncommon. If ChatGPT is in fact a person, then there is no gap to account for. We know, however, that ChatGPT is NOT a person. It’s a machine.

We are now faced with a HUGE gap. What’s the machine doing? We don’t know. The people who built these systems can’t tell us what they’re doing – a point I make in the first section of the article after the introduction, “Views about Machine Learning and Large Language Models.” They can’t even tell themselves what the machine is doing much less craft a simplified account, based on metaphors and analogies, for the rest of us. They know how the system builds an LLM and how it accesses the LLM, but they don’t know what’s going on inside the LLM itself, with its billions and billions of parameters.

That’s one thing. This business about bridging the game between first-person experience and what’s really going on, that’s a second thing. That’s a view of philosophy articulated by Peter Godfrey-Smith, which I discuss in the second part of the article, “Philosophy’s integrative role.” “Integrative” is the word he uses for that function that philosophy plays in the larger intellectual discourse. His argument is that philosophy has largely abandoned that role and that it needs to get back to it. My argument is that nowhere is that more important than in the case of artificial intelligence.

I spend the rest of the article making that point. First, I digress into a section entitled, “Tyler Cowen, Virtuoso Infovore,” where I also discuss Richard Macksey. Cowen has recently argued, in effect, that the very greatest economists, in addition to their specialized work within economics, have also performed that integrative role on behalf of the larger intellectual pubic. Then I get to the argument I’ve been chasing all along, “Artificial Intelligence as a catalyst for intellectual integration,” which you are welcome to read.

But I want to get back to my friend’s response to my article and say a few words about that.

Intention, intuition and deduction in “intelligence”

How did my friend arrive at that statement he made to me? I don’t know. But I’m guessing it was mostly by intuition rather than explicit deductive reasoning. He’d read the article and was puzzled, conjured up our relationship and, viola! out comes the statement, “you are too smart for me.” Simple as pie.

Could he have arrived at that statement through a process of rational deduction? Possibly. How might that have gone?

ONE: FACT: The article doesn’t make sense to me.

TWO: There are three possibilities: 1) It’s nonsense, or at least deeply flawed. 2) It’s fine but too abstract for me. 3) Some combination of the first two.

THREE: PREMIS: Bill’s a smart guy. CONCLUSION: It’s probably 2 or 3. What do I say?

FOUR: FACT: Bill’s a friend. THEREFORE: I’ll give him the benefit of the doubt and base my response on 2.

FIVE: PREMIS: The article is too abstract for me. PREMIS: I’m smart. FACT: Bill made the argument. THEREFORE: Bill must be very smart...

SIX: Here’s what I’ll say: “...you are too smart for me.”

As logical arguments go, it’s rather rickety. I would hate to have to formulate it in terms of formal logic. But you get the idea. Logically, it’s a tangled mess.

In the annoying matter of text books, I leave it as an exercise for the reader to make a similar argument about how I knew what my diplomatic friend was telling me.

I do not believe that ChatGPT is capable of anything like this, though, given that there’s been tons of fiction in its training corpus, containing millions and millions of lines of dialog, it might provide a passable simulacrum in this or that case. The situation will not change when the underling LLM has more parameters and has been trained on a larger dataset, assuming there’s one to be had. The limitation is inherent in the technology.

Critics like Gary Marcus argue that LLMs need to be augmented by the capacity for symbolic reasoning if they are to be truly intelligent, whatever that is. I agree. Symbolic reasoning will get you a lot, but not a whole hell-of-a-lot in the situation I’ve been discussing here. That pseudo-deduction I just went through, symbolic reasoning will get you the capacity to do that, but in even more detail.

On that basis I don’t expect that AI and ML systems will be able to handle the nuances of human interaction in the foreseeable future, if ever. We’ve come a long way, and we have a long way to go.

Invariance and compression in LLMs

One way to thinking about what transformers do is compression. OK. The transformer performs a simple operation on a corpus of texts in such a way that some property of the corpus is preserved in the model. What’s kept invariant between the training corpus and the compressed model?

I think if must be the relationships between concepts. Note that in specifying relationships I mean explicitly to differentiate that from meaning. The process of thinking about LLMs has brought me to think of meaning in the following way:

1. Meaning has two major components, intention and semanticity.
2. Semanticity has two components, relationality and adhesion.*

Intention resides in the relationship between the speaker and the listener and is not always derivable directly from the semantics (semanticity) of the utterance. Intention in this sense is outside the scope of LLMs. And, of course, there are those who believe that without intention there is no meaning. That’s a respectable philosophical position, but it leaves you helpless to understand what LLMs are doing.

By adhesion I mean whatever it is that links a concept to the world. There are lots of concepts which are defined more or less directly in terms of physical things. That’s not going to be captured in LLMs. Of course we now have LLMs linked to vision models so the adhesion aspect of semantics is being picked up. In the universe of concrete concepts we still have relationships between those concepts, and those relationships between concepts can be captured in language without directly involving the adhesions of those concepts. That apples and oranges are both fruits is a matter of relationships between those three concepts and doesn’t require access to the adhesions of apples and oranges. And so forth and so on for a large number of concepts. Then we have abstract concepts, which can be defined entirely through patterns of other concepts, which may be concrete, abstract, or both.

So, relationality. The mechanisms of syntax are designed to map multi-dimensional relationality onto a one-dimensional string. But syntax only governs relationships between items within a sentence. But that’s not quite adequate, because sentences can consist of more than one clause. The relationship between independent clauses within a sentence is different than that between a dependent clause and the clause on which it depends. Etc. It’s complicated. And then we have the relationship between paragraphs, and so forth.

What I’m attempting to do is figure out a way of thinking about the dimensionality of the semantic system. More or less on general principle, one would like to know how to estimate that. Now, when I talk about the semantic system, I mean the semanticity of words. But transformers must deal with texts, and texts consist of sentences and paragraphs and so forth. Setting metaphorical structures aside, the meaning of a sentence is a composition over the meanings of the words in the sentence. But, as I understand it, a transformer is perfectly capable of relating the meaning of a sentence to a single point in its space. And it can do that with larger strings as well. And, of course, the ordinary mechanisms of language allow us to use a string to define a single word; that’s how abstract definition works.

And that’s as far as I’m going to attempt to take this train of thought. Still, I do think we need to recognize a distinction between what’s happening within sentences (the domain of syntax), and what happens with collections of sentences. Beyond that, it seems to me that where we want to end up eventually is a way of thinking about the relationship between the dimensionality of our semantic space and the size of the corpus needed to resolve the invariant relations in that space.

More later.

*Note: The current literature recognizes a distinction between inferential and referential processing, due, I believe, to Diego Marconi, The neural substrates of inferential and referential semantic processing (2011). The functional significance is similar, but only similar, to my distinction between referentiality and adhesion. Inferential processing depends on the relational structure of texts. Adhesion is about the physical properties of the world, affordances in J.J. Gibson’s terminology that are used to establish referential meaning for concrete concepts. But it is also about the patterns of relationships though which the meaning of abstract concepts is established.

Bleg: The use of “role” where (formerly) we used “job” or “position”

Over the last five or ten years I’ve noticed that the word “role” is being used where I normally expected to see “job” or “position.” Just when did this happen, why, is it concentrated in certain sectors of the work force, and does it reflect a mere surface phenomenon, the use of a different word form, or does it reflect some shift in the implied semantics?

Consider these examples:

• The Acme Corporation is hiring for the following jobs...
• What’s your position at Acme?

Now “role” is likely to be used in both cases.

As I think about it, it seems to me that “job” is oriented toward some set of tasks and implies nothing about organizational context. One could be talking about a job at Acme or one could just as easily be talking about, say, unclogging the toilet at home (“it’s a nasty job”). “Position,” on the other hand, does imply an organizational context. One would never say, “Unclogging the toilet is a nasty position.” Nor, for that matter, would one say, “Decorating the Christmas tree is a fun position.”

“Role” also implies an organizational context. How, then, do we differentiate the use of “position” and “role”? Do we do so at all, sometimes, much of the time?

If there’s a difference, I think it’s between what sociologists call status and role (here I’m thinking of a classic discussion by Ralph Linton dating back to the 1930s). A status is a position within a social system having certain rights and obligations where a role is the set of behaviors used to enact that status. The social system could be a family, nuclear or extended, a social group, perhaps a neighborhood, or a bunch of people who sometimes gather together for various activities (e.g. the local rock musicians), a club, or some kind of business or governmental organization. In the case of well-defined organizations, statuses are likely to be specifically named and have specific responsibilities associated with them and listed in appropriate documents.

Language mavins (e.g. Ben Zimmer, John McWhorter), what’s up with current usage of the word?

Hence Syd Lamb's observation that the 'meaning' of a 'concept' (in a neural net) is a function of its place in the network

This is a great illustration of the fact that the *meaning* of neural firings inheres in the act of interpretation by other neurons https://t.co/QzrEsZ5Nub

— Kevin Mitchell (@WiringTheBrain) January 8, 2024

Conceptual coherence for concrete categories in humans and LLMs

Siddharth Suresh, Kushin Mukherjee, Xizheng Yu, Wei-Chun Huang, Lisa Padua, and Timothy T. Rogers, Conceptual structure coheres in human cognition but not in large language models, arXiv:2304.02754v2 [cs.AI] 10 Nov 2023.

Abstract: Neural network models of language have long been used as a tool for developing hypotheses about conceptual representation in the mind and brain. For many years, such use involved extracting vector-space representations of words and using distances among these to predict or understand human behavior in various semantic tasks. Contemporary large language models (LLMs), however, make it possible to interrogate the latent structure of conceptual representations using experimental methods nearly identical to those commonly used with human participants. The current work utilizes three common techniques borrowed from cognitive psychology to estimate and compare the structure of concepts in humans and a suite of LLMs. In humans, we show that conceptual structure is robust to differences in culture, language, and method of estimation. Structures estimated from LLM behavior, while individually fairly consistent with those estimated from human behavior, vary much more depending upon the particular task used to generate responses– across tasks, estimates of conceptual structure from the very same model cohere less with one another than do human structure estimates. These results highlight an important difference between contemporary LLMs and human cognition, with implications for understanding some fundamental limitations of contemporary machine language.

What the abstract doesn’t tell you is that the categories under investigation are for concrete objects and not abstract: “The items were drawn from two broad categories– tools and reptiles/amphibians–selected because they span the living/nonliving divide and also possess internal conceptual structure.” Why does this matter? Because the meaning of concrete items is grounded in sensorimotor schemas whie the meaning of abstract is not.

In their conclusion, the authors point out:

Together these results suggest an important difference between human cognition and current LLM models. Neuro-computational models of human semantic memory suggest that behavior across many different tasks is undergirded by a common conceptual "core" that is relatively insulated from variations arising from different contexts or tasks (Rogers et al., 2004; Jackson et al., 2021). In contrast, representations of word meanings in large language models depend essentially upon the broader linguistic context. Indeed, in transformer architectures like GPT- 3, each word vector is computed as a weighted average of vectors from surrounding text, so it is unclear whether any word possesses meaning outside or independent of context.

For humans, the sensorimotor grounding of concrete concepts provides that conceptual core, which is necessarily lacking for LLMs, which do not have access to the physical world. Context is all they’ve got, and so their sense of meanings for words will necessarily be drawn to the context. The authors acknowledge this point at the end:

Finally, Human semantic knowledge is the product of several sources of information including visual, tactile, and auditory properties of the concept. While LLMs can implicitly acquire knowledge about these modalities via the corpora they are trained on, they are nevertheless bereft of much of the knowledge that humans are exposed to that might help them organize concepts into a more coherent structure. In this view, difference in the degree in conceptual coherence between LLMs and humans should not be surprising.