Serendipity in the Wild: Three Cases, With remarks on what computers can’t do

New working paper. Title above. URLs, abstract, table of contents, and abstract below.

Academia.edu:
https://www.academia.edu/145860186/Serendipity_in_the_Wild_Three_Cases_With_remarks_on_what_computers_cant_do
SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6043814
ResearchGate: https://www.researchgate.net/publication/399584810_Serendipity_in_the_Wild_Three_Cases_With_remarks_on_what_computers_can't_do

Abstract: This paper examines intellectual creativity as it occurs in ordinary, open-ended scholarly practice, through three detailed case histories drawn from my own work in the humanities. Each case traces how a line of inquiry emerged without a predefined problem, method, or endpoint, and how a vague sense of interest gradually crystallized into a focused intellectual project.

Introduction: Serendipity in Mind

The first case reconstructs the process by which an unplanned viewing of Jaws developed into a Girardian interpretation of the film, centered on the recognition of Quint as a sacrificial victim. The second follows an exploratory investigation of the term “Xanadu” on the early web, where a surprising search result led, through low-cost probing, to the identification of distinct cultural clusters. The third describes the discovery of a previously unrecognized center-point structure in Conrad’s Heart of Darkness, originating in the noticing of a minor narrative anomaly and pursued through opportunistic quantitative checks.

Across these cases, creative work proceeds through hunches, comparative wandering, sensitivity to salience, and decisions shaped by opportunity cost, rather than through the execution of well-defined tasks. Only late in each process does a clear problem boundary emerge, enabling more systematic reasoning.

The paper uses these cases to clarify a limitation of contemporary large language models and related AI systems. While such systems can operate effectively once a problem frame is supplied, they do not participate in the open-ended, exploratory processes by which humans discover what is worth investigating in the first place. The cases thus support a model of human–AI complementarity in which problem-finding remains a distinctively human contribution, and AI serves as a powerful tool once direction has been established.   

Table of Contents

Introduction: Serendipity in Mind 3
Intellectual creativity: Case 1, Interpreting Jaws 4

Watching Jaws 5
The game is afoot 6
I’m in! 6
What kind of a process is that? 7

Intellectual creativity: Case 2, The Xanadu meme 9

Cultural evolution 10
The cybernetic cluster appears 11
But still, what got me started? 12
Opportunity cost 13

Intellectual creativity: Case 3, Center-point construction in Heart of Darkness 15

Story and plot, Center-point construction 15
What would an AI do? 19
Is this real? How do we know? 21

Appendix 1: Summary and Evaluation by Claude 23

Summary 23
Evaluation 23
Conclusion 24

Appendix 2: Summary and Evaluation by ChatGPT 26

Summary 26
Overall Assessment 27
Human–AI Complementarity: Lessons from Three Cases 28

Appendix 3: Seven Brief Examples of Serendipity 30    

Introduction: Serendipity in Mind

In the three years since ChatGPT was released on the web the sophistication of chatbots has increased a great deal, leading some to assert that it won’t be long before chatbots will be able to outperform humans on all intellectual tasks. I’m not so sure. Why? Because the problems they solve, the tasks they’ve been set, all seem to be bounded tasks in well-explored universes. As Johnson, Karimi, and Bengio have remarked in a recent article, many important intellectual tasks require “the ability to navigate intractable problems - those that are ambiguous, radically uncertain, novel, chaotic, or computationally explosive.” They talk of wisdom as the quality missing in current machines. I’m not sure what I think about that word, “wisdom”; it carries a lot of baggage.

I’m more comfortable talking about serendipity, not as an attribute of minds, or of situations, but as, well, whatever it is that serendipity characterizes. In the third appendix I present seven brief cases that I elicited from ChatGPT with a prompt, but the burden of my argument rests accounts of three problems that I’ve worked on: an interpretation to Steven Spielberg’s Jaws, the distribution of the term “Xanadu” on the web (the “Xanadu meme”), and the discovery of center-point structure in Joseph Conrad’s Heart of Darkness. I present these cases, not because I think that they’re somehow special. My thinking in these cases doesn’t seem to me to be particularly abstract. It’s not “rocket science” as the saying goes. I present those cases simply because they are mine and I know them in detail. I know what I did, when I did it, and why. What I did depended on experiences and “hunches” built up through experience. Those don’t strike me as the kinds of things amenable to current computational techniques.

Once I’d completed drafting those cases I presented them to both ChatGPT and Claude. I’ve included summary excerpts from those interactions as appendices. I’ll let Claude conclude this introduction:

Benzon's case studies effectively demonstrate that the most interesting intellectual work often begins not with problem-solving but with problem-finding. His examples show how experience, intuition, and willingness to follow hunches create opportunities for genuine discovery. While current AI systems are powerful analytical tools, they lack the kind of embodied experience and open-ended curiosity that drives human creativity. The work suggests that the future of AI-assisted scholarship may lie not in replacing human insight but in creating powerful partnerships where humans excel at boundary-setting exploration and AI excels at systematic analysis within those boundaries.

A historical problem solved by a human but that is beyond chatbots

Elon Danziger, ChatGPT Will Never Beat Indiana Jones, NYTimes, Dec. 22, 2025.

Across from the Florence Cathedral in Italy stands a much older church, the Baptistery of San Giovanni. It is a beloved center of religious life, where many Florentines are baptized to this day. Staid columns and lively arches hug its eight sides, half-camouflaged in patterns of green and white marble. Without the baptistery’s emulation of the architecture of ancient Rome, it’s hard to imagine Florence birthing the architectural Renaissance that changed the face of Europe. Yet for centuries, there has been no compelling solution as to who built it and when and for what reasons. Decades ago, I gave tours of the baptistery and came to revere it, and in the early 2020s I began delving into its origins.

After years of poring over historical documents and reading voraciously, I made an important discovery that was published last year: The baptistery was built not by Florentines but for Florentines — specifically, as part of a collaborative effort led by Pope Gregory VII after his election in 1073. My revelation happened just before the explosion of artificial intelligence into public consciousness, and recently I began to wonder: Could a large language model like ChatGPT, with its vast libraries of knowledge, crack the mystery faster than I did?

So as part of a personal experiment, I tried running three A.I. chatbots — ChatGPT, Claude and Gemini — through different aspects of my investigation. I wanted to see if they could spot the same clues I had found, appreciate their importance and reach the same conclusions I eventually did. But the chatbots failed. Though they were able to parse dense texts for information relevant to the baptistery’s origins, they ultimately couldn’t piece together a wholly new idea. They lacked essential qualities for making discoveries.

There are a few reasons for this. Large language models have read more text than any human could ever hope to. But when A.I. reads text, it’s merely picking up patterns. Peculiar details, outlier data and unusual perspectives that can influence thinking can get lost. Without eccentric or contrarian ideas, I never would have made my discoveries. [...]

Synthesizing so many pieces of medieval history into a new interpretation required stepping back and reconsidering their importance and how they relate to one another. A.I. may be able to optimize the process of collecting those pieces, but discovery means drawing new connections — something far beyond current A.I. capabilities, as the tests I did confirmed to me.

This is consistent with a series of posts I did on my own work: One about interpreting Jaws, one about investigating the "Xanadu" meme, and one about ring-composition in Heart of Darkness.

Intellectual creativity, humans-in-the-loop, and AI: Part 4, Centerpoint construction in Heart of Darkness

This case involves aspects of the first two cases we’ve considered. Like the analysis of Jaws, it began with an open-ended examination of a particular text, Joseph Conrad’s Heart of Darkness. That open-ended process leads to a specific moment that is like the beginning of my hunt for the history of the Xanadu meme. In this case I noticed that, at a certain point, the narrative gets ahead of itself. I then noticed that the paragraph in which that occurred was extraordinarily long. “I wonder,” thought I to myself, “if that’s the longest paragraph in the text?” It didn’t seem like a very interesting or important question. So events get a bit out of order in the paragraph, so what? The opportunity cost of answering the question was low, however, so I set out to answer it. When I charted the distribution of paragraph lengths in text order I saw a surprising pattern. That changed the way I thought about the text.

This case brings up another issue as well, one that didn’t exist in the other two: Is it real? I discovered a pattern of words in Heart of Darkness that is of a kind not recognized by literary scholars. What follows from that? What kinds of questions does it raise and how are they to be answered?

Story and plot, Center-point construction

I think the thing to do is to run through the decisions I made and why I made them.

1. Read Heart of Darkness

I had finished a series of posts about Apocalypse Now and decided that it was time to read Joseph Conrad’s Heart of Darkness. Why? Because the movie was loosely passed on Conrad’s book, which I’d never read. So I took it from the library and started reading it. At some point I decided I might actually want to work on the text so I downloaded it from Project Gutenberg. Why? Mostly so I could annotate it. At some point I noticed that temporal anomaly I mentioned above.

I mentioned that anomaly in the first post I did on the book, Heart of Darkness: Narration and Temporal Displacement (July 10, 2011).

2. Examine a temporal anomaly

As you know, Heart of Darkness is a story about a pilot named Marlow taking a boat up the Congo River to a trading station whose proprietor, a man named “Kurtz,” had gone silent. I was at a point in the story when we had not yet reached that trading station. All of a sudden I found myself reading about something that obviously had happened at the station. “That’s odd,” I thought. Writer’s sometimes do that, I know, and one is supposed to notice it. I fished around in the text to verify that, yes, we’d skipped ahead in time, and in this process noticed that this particular paragraph, which was a precis of Kurtz’s life, seemed extraordinarily long. Kurtz is one of two central characters (Marlow is the other); This paragraph is the first time we learned much about him.

By this time I had either started blogging about the book – something I do, blog about books in the process of reading – or I had decided to do so. I thought that, in the process of writing about this paragraph I would remark on the length of the paragraph, suggesting that it was possibly the longest paragraph in the text.

3. Check the length of that paragraph

At that point I realized that that’s something I could check easily enough. As I had an electronic text, I wouldn’t have to count the words by hand. Rather, I could enlist the help of Microsoft Word, which has a function that counts the number of words in a chunk of text that one has selected. Sure, I’d have to do that for every paragraph in the text, and I’d have to number the paragraphs to keep track, but that’s easy to do, though tedious. So I did it. One by one I counted the number of words in each paragraph and entered the total into a spread sheet. When I was done verifying that that particular paragraph, number 103 in the Gutenberg text, was easy; just sort the spreadsheet entries in order. Once that was done paragraph 103 was at the top of the list, with 1531 words.

I mention that in my third post, Closure, Attachment, and Abstract Objects in Heart of Darkness (July 12, 2011), where I also give the length of the 2nd, 3rd, and 4th longest (1129, 1103, and 865 words respectively). Otherwise I nothing about paragraph lengths.

On July 16 I did a post in which I asserted that paragraph 103, which I was now calling the nexus, was the structural center of the text: The Heart of Heart of Darkness. That suggested ring-composition, which interests me a great deal. Ring-composition foregrounds the structural center of the text. As I was not sure that Heart exhibited all the characteristics I coined the term “center point construction,” after the characteristic it did exhibit. [For what it’s worth, the late Mary Douglas brought ring-composition to my attention. Her last book was about the subject: Thinking in Circles (Yale 2007).]

In that post I also pointed out that, not only was the nexus paragraph the structural center, but that it was very strongly marked. At the point in the story Marlow’s boat had been attacked by natives on the shore and the helmsman was speared in the chest and fell bleeding on the deck. At that point Marlow breaks off from the story and inserts the nexus, which gives us an overview of Kurtz’s life, into his narration. When he’s done with the nexus he returns to the main narrative at the point where then helmsman is blooding out on the deck. Marlow tosses the helmsman overboard. This paragraph, where for the first time we learn who Kurtz was and what he did, is flanked by the death the helmsman.

A day later, July 17, I made the following two charts in Excell. Each bar stands for a paragraph in the text. The length of the bar is proportional to the number of words in the paragraph. In this chart the paragraphs are ordered according to their length:

Figure 1: Exponential distribution of paragraph lengths

I had no expectations about the shape of the plot when I made the chart. But that distribution appears to be lawful. What do we know about the distribution of paragraph lengths in texts?

In this chart the paragraphs are listed in their order in the text:

Figure 2: Spikey distribution of lengths in text order

That distribution is very spikey. The extreme length (1531 words) of the nexus paragraph, 103, appears quite remarkable in this chart, and does the lengths of the next two longest paragraphs, 1129 and 1103 words.

4. Get others involved

By now I was actively investigating paragraph length and its implications for the formal structure of Heart of Darkness. For some reason, I don’t recall why, I wondered whether or not the distribution of paragraph lengths was a power law distribution. So I wrote to Cosma Shalizi and asked him to check it for me. Those charts (Figures 1 and 2) show up in a post from July 18, 2011: Digital Humanities Sandbox Goes to the Congo.

I also sent a link to that post to Mark Liberman, who runs a group blog on language and linguistics, Language Log, hoping he would post notice of it there. I wanted to find out whether anyone knew anything about the distribution of paragraph lengths in texts and figured that the readers of Language Log might know. He obliged me that day with a post, Markov’s Heart of Darkness, which elicited a lively discussion, but alas, no further information about paragraph lengths, though Liberman posted some work he was prompted to do on some other texts, including Conrad’s Nostromo, and Henry James, The Golden Bowl.

* * * * *

I’ve continued to think about this business of paragraph length, but want to set that aside. Let’s take up the question: Would an AI have undertaken the investigation that I’ve outlined so far?

Intellectual creativity, humans-in-the-loop, and AI: Interlude, Claude on “xanadu” and the metaphysical structure of the world

After yesterday’s post about the Xanadu meme went up I started wondering just how an AI, a current AI, would even confront such a question. Why would an AI issue a simple one-word query to the web: “Xanadu”? And when it got 2,000,000 hits, what reason would it have to think that a large number? I decided to have a discussion with Claude Sonnet 4 about “xanadu.” That eventually led to an interesting discussion that ended with something I’ve been calling “the metaphysical structure of the world.” I introduced the term in a blog post from August of 2020, which I’d written in the wake of GPT-3 and then included it in my working paper, GPT-3: Waterloo or Rubicon? Here be Dragons. I’m beginning to think it might be useful. 

On the way there ChatGPT made the following observation

Current AI systems are remarkable at following up on investigations once they’re pointed in a direction. But that initial moment of recognition - “this might be interesting” - seems to require a different kind of agency. It’s not just pattern recognition; it’s pattern recognition combined with intrinsic motivation and aesthetic judgment about what’s worth pursuing.

Yes.

As always in these discussions, my prompts are in boldface.

* * * * *

“Xanadu”

What comes to mind when I say “Xanadu”?

Several things come to mind with “Xanadu”:

Literary: Most famously, it’s from Coleridge’s poem “Kubla Khan” - “In Xanadu did Kubla Khan / A stately pleasure-dome decree” - describing an exotic, dreamlike palace.

Historical: The actual Xanadu (Shangdu) was Kublai Khan’s summer capital in 13th-century Mongolia, which inspired Coleridge’s poem.

Pop culture: The 1980 musical film with Olivia Newton-John and Gene Kelly, along with its title song. There’s also Rush’s epic song “Xanadu” from their album A Farewell to Kings.

Cinema: In Citizen Kane, it’s the name of Charles Foster Kane’s sprawling, isolating estate.

The word has become synonymous with an idealized, fantastical place - somewhere luxurious but perhaps unattainable or illusory. It carries this sense of exotic grandeur mixed with a hint that such perfection might be more dream than reality.

What made you think of Xanadu? Were you thinking of one of these references in particular?

I was just curious. Back in January of 2006 I typed “Xanadu” into Google and got 2,000,000. That surprised me. “Xanadu” is not a common term. People don’t have much of an opportunity to use it. Yet there were 2 million web pages that mention it.

That’s a fascinating observation! You’ve hit on something really interesting about how language spreads in digital spaces versus everyday conversation.

“Xanadu” is indeed quite rare in daily speech - most people might go years without using it. But on the web, it has this remarkable reach because it’s such an evocative, culturally loaded word. Each of those cultural touchstones I mentioned creates its own cluster of web content: literary analysis sites discussing Coleridge, movie review sites covering the film, music sites with Rush lyrics, film studies pages about Citizen Kane.

Plus, “Xanadu” is exactly the kind of word that gets borrowed for business names, product names, gaming references, fantasy literature, and artistic projects. It carries this instant sense of mystery and grandeur that appeals to people creating content.

Your 2006 observation captures something about how the internet amplifies certain words that have cultural resonance far beyond their frequency in speech. Words that are “mythologically useful” - like Xanadu, Atlantis, Eden - probably punch way above their weight in search results because they’re so semantically rich.

It’s a perfect example of how digital archaeology can reveal hidden patterns in our collective cultural consciousness. Did you notice similar patterns with other evocative but uncommon words when you were exploring search results back then?

It didn’t take long to figure out that most of the hits did not link to pages mentioning Coleridge’s poem. After a bit I determined that a lot of the hits belonged in what I termed the “sybaritic” cluster, which I trace back to Citizen Kane. Lots of people saw that movie and it has a fairly substantial cultural presence. The movie itself quotes the first five lines of the poem in the newsreel sequence and, as you mentioned, there’s Kane’s mansion.

That’s a brilliant piece of cultural archaeology! Your “sybaritic cluster” concept really captures how Citizen Kane became the dominant cultural vector for “Xanadu” rather than the original Coleridge poem.

It makes perfect sense - while English literature students encounter the poem, Citizen Kane reached massive audiences and embedded “Xanadu” firmly in the American cultural lexicon as shorthand for extravagant, perhaps excessive luxury. The film’s Xanadu represents wealth turned grotesque and isolating, but the name itself got detached from that cautionary context.

So you probably found tons of luxury hotels, upscale restaurants, spa resorts, real estate developments, and high-end businesses that borrowed the name purely for its connotations of opulence. They’re not thinking about Mongolian summer palaces or Romantic poetry - they’re thinking about that sense of ultimate luxury that Kane’s estate represented.

This is such a neat example of how cultural transmission works. The literary source provides the raw material, but a hugely influential film recontextualizes it and creates the dominant semantic pathway. Most people using “Xanadu” for their luxury business probably have no idea they’re referencing Coleridge - they’re unconsciously channeling Welles.

Did you find that pattern held up as you dug deeper? Were there other cultural artifacts where you could trace similar semantic drift from high culture sources through popular media?

Ted Nelson’s Project Xanadu

I also identified a cluster I called the “cybernetic” cluster. For example, there’s a spot on Saturn’s moon, Titan, that I put in that cluster. Where do you think this cybernetic cluster comes from?

Ah, that has to trace back to Ted Nelson’s “Project Xanadu” - his pioneering hypertext system from the 1960s! Nelson chose the name explicitly referencing Coleridge’s poem, envisioning his system as a kind of “magic place of literary memory” where all human knowledge could be interconnected.

Intellectual creativity, humans-in-the-loop, and AI: Part 3, The Xanadu meme

The process of chasing down the Xanadu “meme” was quite different from interpreting Spielberg’s Jaws.

Late in 2005 I had become a guest author at a now defunct group blog, The Valve, which was having a symposium on Franko Moretti’s book, Graphs, Maps, Trees. In that general context I did a web search on the term, “Xanadu,” though at this point I don’t recall whether or not I was explicitly thinking about that symposium when I did the search. Judging from my notes, that was likely in the second week of January 2006. To surprise I got roughly 2 million hits.[1] That seemed high to me. On a hunch I decided to investigate. On January 24 posted the (initial) results of my investigation, One Candle, a Thousand Points of Light: Moretti and the Individual Text. I argued that many of the hits feel into one of two clusters, which I termed the sybaritic group and the cybernetic group.

I don’t recall any moment when, after I made that first web search, I decided “I’ve got it.” That is, there was no moment comparable to the point in my work on Jaws where, suddenly recalling Girard’s ideas, I switched from open-ended exploration to the focused elaboration of a specific thesis. Judging from my vague recollections and from notes I made at the time, what seems to have happened is that, as I went poking around what I found was sufficient to justify the effort, so I kept it up. Before long I had something worth writing about. Four years later, in March of 2010, I published a working paper, One Candle, a Thousand Points of Light: The Xanadu Meme. Then, earlier this year (July 10, 2025), I called on ChatGPT to update that work, Tracking the “Xanadu” Meme.

Let’s take a closer look.

Note: Richard Dawkins introduced the term “meme” in his 1976 book, The Selfish Gene. He thought of it as the cultural analog to the biological gene. Since that time there has been considerable discussion of memes in cultural evolution without any consensus being established. For various reasons I’ve decided to abandon the term as a term of art and talk about coordinators instead. But we need not worry about that here. For the purposes of this post the term “meme” is fine as long as you realize that I’m not using it as a technical term. It’s just a convenient way of referring to a “bit” of culture without worry about just what role it plays in the process.

Cultural evolution

I don’t recall just why I did a web search on “Xanadu.” I do a lot of things out of idle curiosity. Let’s chalk it up to that.

I was surprised when I got 2,000,000 hits. Why? For one thing, “Xanadu” is not a common word. It has no use in daily life. It’s the name of Kubla Khan’s summer capital in 13th century CE. It’s also the second work I Coleridge’s poem “Kuba Khan,” which begins, “In Xanadu did Kubla Khan/ A stately pleasure-dome decree.” I had a long-standing interest in that poem, which is why I searched on that term. And, while it is one of the best-known poems in the English language, as these things go, it’s not that well known, not in comparison to recent pop songs, current movie stars, major world leaders, and so forth.

Beyond that, I have a long-term interest in cultural evolution. Thus I wasn’t thinking about “Xanadu” as just some word. I was thinking about it as a culturally laden word. Perhaps its presence on the web tells us something about culture?

By the time I’d done that web search I also knew that “Xanadu” has been introduced into modern English through Coleridge’s poem. I also knew that that poem had been quoted in Orson Welles’ 1941 movie, Citizen Kane, where it was the name of a mansion Kane had built in Florida. I saw the film in college in the 1960s and then again somewhat later when it had been re-released. Citizen Kane has been called the best (American) film ever made. Regardless of that claim, it’s been seen by millions of people. Could it be that a lot of those web hits could be traced back to Welles’ film rather than to Coleridge’s poem? That is, the people who’d created those pages, and many who saw them, may never had read Coleridge’s poem.

That’s the kind of question that was on my mind when I began poking around on the web. I saw the web, which its rich connectivity, as a way of exploring how the process of cultural evolution. Perhaps “Xanadu” was an interesting case. I didn’t take much work for me to conclude that, yes, a lot of those hits probably could be traced back to Citizen Kane. That gave me what I termed the sybaritic group of web pages, which is about luxury and excess verging into sensuality (see what you get when you search on “sex Xanadu”). 

The cybernetic cluster appears 

Now let’s look at the pages returned by my “Xanadu” query early in 2006. These were the first ten items returned:

1. Xanadu (1980): Xanadu - Cast, Crew, Reviews, Plot Summary, Comments, Discussion, Taglines, Trailers, Posters, Photos, Showtimes, Link to Official Site, Fan Sites.

2. Kubla Khan: In Xanadu did Kubla Khan A stately pleasure-dome decree: Where Alph, the sacred river, ran Through caverns measureless to man: Down to a sunless sea. ...

3. Xanadu Australia: The name "Xanadu" and the Flaming-X symbol are software an eid service trademarks of Project Xanadu, registered in certain countries and claimed elsewhere. ...

4. XANADU Software Home Page: In XANADU did Kubla Khan A stately pleasure dome decree... ---. The XANADU software package comprises high-level, multi-mission tasks for X-ray astronomical ...

5. Welcome to Udanax.com: Xanadu Secrets Become Udanax Open-Source. The long history of the Xanadu® vision of hypertext has inspired many individual

6. The Mills - Madrid Xanadu: Madrid Xanadu · Register for X Alerts. Lo último en Madrid Xanadú. No hay ningún evento previsto actualmente.

7. Index: .:Test Page - www2:. www.xanaduwines.com.au/ - 1k - Cached - Similar pages

8. Xanadu: The language and translation wizard Xanadu: The language and translation wizard. Translate words and terms. Find professional translators. Read language related news.

9. Amazon.com: Xanadu (1980): DVD: Xanadu, Olivia Newton-John, Gene Kelly, Michael Beck, James Sloyan, Dimitra Arliss, Katie Hanley, Fred McCarren, Ren Woods, Sandahl Bergman, Lynn Latham, ...

10. Ted Nelson and Xanadu: The Electronic Labyrinth is a study of the implications of hypertext for creative writers looking to move beyond traditional notions of linearity.

From my 2006 post:

The first and ninth items are for a 1980 movie starring Olivia Newton-John and are obviously associated with her hit song of the same name [...] The second item is a text of the poem itself in the online text repository at the University of Virginia [...] This is one of many copies of the text online; I've made no attempt to count them, but that should be doable with the appropriate resources. Entries three, four, five, eight, and ten are all related to Ted Nelson's Xanadu project [...] There's nothing at item seven, nor do I have any idea why it is so highly ranked. Item eight is a retail and entertainment complex in Madrid (also in the Wikipedia list) that features an indoor ski-slope – caves of ice?

Think about entries 3, 4, 5, 8, and 10. They’re all associated with a single software project, Ted Nelson’s pioneering hypertext work, Project Xanadu, which was well-known in the tech world. That’s where I got the idea of a cybernetic cluster of web pages. Entries 1 and 9 are related to Olivia Newton-John’s 1908 film, Xanadu, in which Gene Kelly recites the opening lines of “Kubla Khan.” They belong to the sybaritic cluster.

That’s more or less where things stood when I posted my results to The Valve. But if you go to the link above you’ll see that the post generated a fair amount of discussion. And on January 26 I posted some remarks from my friend, the late Tim Perper, who was trained as a biologist, who’d posted some simple Boolean queries which refined my results. I took a cue from Tim and generated some queries of my own. That discussion went into early February. (Note: I’ve also made a PDF of the post and the whole discussion, which you can download here.)

Looking through my files I see I have an unpublished document from February 8, 2006, “Notes on Xanadu Lineages and Related Matters.” In that document, a month after the discussion at The Valve, I’d decided that Olivi Newton-John’s song and film, Xanadu, had in fact established a third population of hits that was independent of the first two, the sybaritic and the cybernetic. That’s where things remained when I decided to wrap things up and publish that working paper: One Candle, a Thousand Points of Light: The Xanadu Meme.

But still, what got me started?

That’s where I ended up. But what got me started in the first place? As I said, I was surprised that a search on “Xanadu” turned up 2,000,000 hits. So what? In what sense is 2,000,000 a large number? I should note that I did a number of searches at various intervals just to verify the number. I’d gotten as few as 1,600,000 hits and on one occasion 7,000,000. As I noted in my post at The Valve:

Yet there's more involved that simply the number itself, which is, in context, not terribly impressive. Though it is not easy to determine how many pages there are in the web, the number is probably upward of 4 billion. Two million hits is five ten thousandths of the web. But, what percentage of the web would you expect any one query item to retrieve? At this point these are just numbers; their significance is obscure.

At this point I’m included to say that I had a hunch, a hunch that turned out to lead somewhere. I have lots of hunches. I follow some of them, and some of them actually lead somewhere.

On the one hand, by that time I had a good deal of experience in make web searches and so had developed some sense of the kind of results you get when making a query. That’s the background against I had my hunch, vague though it is. But, as I’ve already said, I was interested in cultural evolution. And cultural evolution takes place through hundreds upon thousands upon millions of face-to-face transactions between individuals, but also broadcast transactions between individuals that various media properties. “Xanadu” is one of those bits of culture.

Opportunity cost

Here's how I’m now thinking about it: I had a hunch as a result of making a web query. Making web queries is easy. So I decided to verify my initial query. Easy to do. It was easy to look up “Xanadu” in Wikipedia and see what turns up there. Similarly, I could go to the Oxford English Dictionary and I could even query the archives of The New York Times for occurrences of the term. All of these things were quick and easy and, wouldn’t you know, the results turned out to be interesting.

And I didn’t have anything pressing on me at the time.

That is to say, conversely, the opportunity cost of getting more information was low. As long as new stuff turned up, why stop? That’s what kept me going long enough to produce my initial post to The Valve and kept me continuing on until, in the middle of March, I’d decided that we had three clusters, not two. I didn’t see any point in trying to work things into a formal article since I didn’t know of any place that would consider such an article. So I stopped working.

Academia.edu started up in 2008. I probably found out about it shortly after, but I don’t know when I decided to join it myself. Once I did I realized that, not only could I use it as a way of distributing articles that I’d already published, but I could use it to distribute new work. That’s when I decided to take the work that I’d done on “Xanadu” and publish it though Academia.edu.

At some point I wondered whether or not I should update that work. I’ve got records of web searches I did in 2018. But I never got around to doing all the fiddly details needed for a proper update. Why not? Opportunity cost. I had more fruitful ways of using my time.

Then earlier this year, I was doing some work with ChatGPT where I uploaded my working paper in connection with something or another. ChatGPT offered to update that work.

I wasn’t interested in it at just that moment, but I kept the offer in mind and a couple weeks later decided to go ahead with the study. It turned out that ChatGPT had to work through the night executing the study, but that wasn’t my time. I was sleeping. When I got up the next morning, the study was done. It was then a relatively simple matter to combine that work with a summary of my original paper and produce the updated study: Tracking the “Xanadu” Meme.

* * * * *

[1] By way of comparison, when I search Google on “Xanadu” today I get roughly 9 million hits.

Intellectual creativity, humans-in-the-loop, and AI: Part 2, Interpreting Jaws

This is the second post in my series, Intellectual creativity, humans-in-the-loop, and AI. The first post set the theme: On the boundaries of cognition in humans and machines. The idea is simple: These days LLM-based AIs are given specific tasks, which I’m thinking of as imposing a boundary on the underlying model within which a solution is to be found/generated. The most interesting situation, though, is one where there is no boundary set. Rather, the task is to discover a problem one can work on, which I’m thinking of as imposing a boundary in the space. Here’s my first case, creating a Girardian interpretation of Steve Spielberg’s Jaws.

I’m using that as an example because, a) I did my own interpretation a couple of years ago (February 28 2022), so I remember the process, and b) more recently I had ChatGPT interpret the film (December 5, 2022), though, to be honest, comparing two performances is a distant second to my recollection of my own procress. What’s important about ChatGPT’s performance is that I gave it both the “text” (broadly understood) to interpret, and the conceptual lens though which to make the interpretation, the ideas of René Girard. ChatGPT had a specific task to perform. By contrast, I had no intention of interpreting Jaws when I decided to watch it a couple of years ago. How and why did I decide to interpret Jaws? That’s the issue.

* * * * *

By way of background, I’m trained as a literary critic and I have a long-standing interest in films. I all but majored in literature as an undergraduate at Johns Hopkins in the 1970s, where I heard Girard lecture on mimetic desire and sacrifice, his central ideas. I also attended two university-sponsored film series while I was there, one on foreign films and the other on American films. I got a Ph.D. in English at SUNY Buffalo in the 1970s, where I also attended a number of film series.

Thus interpreting texts is my business. I’ve done a lot of it in the last half-century. It’s something my mind has been organized to do. And, while films and novels, not to mention plays and poems, are quite different kinds of texts, to a first approximation (which is sufficient to my purpose here), they are the same kind of activity. You need to identify two things, a suitable text and “a way in.”

What makes a text suitable? The fact that you have a way in. What do I mean by a way in? An “angle,” a point of entry, somewhere to apply some idea to the text in a fruitful way. I tend to draw my ideas from psychology, psychoanalytic and cognitive, and structuralism and semiotics. In this case, I chose the ideas of René Girard, though it might be more accurate to say that those ideas chose me (& Jaws). In any event, though I’ve been acquainted with Girard since the 1960s, this is the only time I’ve used his ideas in interpreting a text.

Watching Jaws

But, as I’ve said, I had no such thing in mind when I decided to watch Jaws back in 2021. I hadn’t seen it when it first came out (in 1975), but I knew it was an important film, the first so-called blockbuster. I decided it was time to watch.

I should note however, that as interpreting texts is my business, I’m always of the lookout for interesting texts. I have a series of media notes on New Savanna that dates back for May of 2010, almost the beginning of the blog, though the first post explicitly labeled as a media note didn’t go up until June of 2019. Any movie or TV program I watch could end up in a media note, and the occasional YouTube video as well. Media notes are, for the most part, quick dirty posts that I write up in an hour or so, sometimes less, occasionally a bit more. They’re not meant to be anything more. I never know when I watch something whether or not it will end up in a media note. And, as proved to be the case with Jaws, every once in a while I become really interested in something.

Keeping that in mind as deep background, I cued Jaws up on Netflix and watched it. Afterward I took a look at the Wikipedia entry. I do this for most of the movies I stream online, and for many of the TV series as well. I think of it as ‘calibration.’ I want to review the film and get a rough sense of what’s been said about the film. Jaws is such an important film that the entry was relatively large. There were a lot of comments by critics offering various interpretations of what the shark stood for, but the one I found most interesting was that by Fredrick Jamison, a well-known Marxist literary critic. He asserted it mean anything, everything, and nothing (in particular). I liked that.

But what I latched on to were the remarks about the sequels. There were three of them, all of them inferior to the original. I decided to take a look.

Now, I’m interested. I didn’t have anything in mind. But I like watching movies, so why not?

I like Jaws 2, but could see that it wasn’t quite as good as the original. I was unable to finish watching the other two sequels. Jaws 4 was especially wretched. In this process I’m sure I read the Wikipedia entries for those movies as well and re-read the entry for Jaws.

How did I arrive at those judgments? At this point I don’t recall, but I’m pretty sure I didn’t use any specific process. It was just intuitive judgements. How do we make those?

At this point I asked myself: So Jaws 2 is inferior to Jaws, why? I don’t know how long it took me to arrive at this conclusion, but here it is: the original film was tighter? What does that mean? Well, it falls into two parts. The first part is about 3/5ths of the film and takes place in Amity, with many people doing this that and the other. It’s pretty chaotic. The second part involves only four characters, three humans and the shark. It’s much more tightly focused. There’s no such division in Jaws 2, which involves many people through the entire film. The action is more diffuse.

The game is afoot

OK, so what makes the original one tick? I spent a good deal of time thinking about the three-way interaction between the shark, and two factions of townspeople. One faction wanted to close the beaches until the shark about been found and killed (or had been known to leave). The other faction wanted to keep the beaches open so that the town could rake in tourist dollars, which were central to the local economy. In the end, though, the town decided that they had to close the beaches and hire someone to kill the shark.

That brings us to the second part of the film. The town hired Quint, a disagreeable old shark hunter who agreed to do the job for $10,000. Sheriff Brody and Matt Hooper, a marine biologist and shark expert, went with him. I settled on one question: Why did Quint have to die? Why didn’t sheriff or Hooper die?

I’ve not yet decided to write an article, but I sense that I’m on to something.

At this point I’m thinking about story-telling conventions. Jaws tells a certain kind of story. In this kind of story the monster, or whatever, must be killed, otherwise there’s no point. Well, the shark was killed, but why Quint as well? One possible answer: Well, we can’t make it too easy. Someone has to die more or less on general principle. Why not Quint? After all, he’s a rather unpleasant character.

How does Quint’s death affect the ending? That is to say, in this kind of story we generally have a celebratory ending when the heroes return home after the monster was vanquished. Jaws didn’t have such an ending.

I’m in!

And then it hit me: Quint’s the sacrificial victim! Not the shark. Bingo! Girard, mimetic desire, sacrifice.

Crudely put, as best as I can remember, that’s how it happened. All these things came together at once. I had my way in, my point of conceptual departure. This is when I decided to write about the movie.

I made notes, read more about the movie, read some Girard, and consulted with my friend David Porush, who had studied with Girard in graduate school (a half century ago). I still had a lot of work to do, but it was all fairly tightly focused. I remember, for example, spending a fair amount of time tracking down scripts so that I could get the exact wording for Quint’s speech about being on the USS Indianapolis in WWII when the ship was sunk and lost half the crew to sharks. I wasn’t able to find a version with the actual words on the screen, so I ended up having to transcribe it myself. Why did I do that? Because it was about sharks and Quint, that’s why. Details. Interpretations are built on lots of details. Another detail: Sheriff Brody telling the mayor:

This summer’s had it. Next summer’s had it. You’re the mayor of Shark City. You wanted to keep the beaches open. What happens when the town finds out about that?

“Shark City,” and important detail. And so forth.

And that, more or less, is how I ended up writing about Jaws. I watched the film in 2021. I may have watched Jaws 2 at that time as well. I forget what brought me back to the film a year later. Maybe nothing in particular. I was just curious. I watched it again and curiosity became interest and interest grew into a written article over the course of two or three weeks.

I made my first New Savanna post about Jaws on Feb. 9, 2022. It was a long one, almost 5K words. My 3 Quarks Daily article appeared on February 28.

What kind of a process is that?

I don’t know. But I’m pretty sure it depends on my education, interests, and decades of experience. What’s interesting is that it wasn’t until near the end of the process, after I’d watched Jaws three or perhaps four times, and Jaws 2 at least two times, and after I’d done quite a bit of thinking about the films, only then did I come up with a specific interpretive approach. The thinking that led to that was based on my knowledge of stories and conventions and my experience in reasoning about them.

It took all that to, in effect, place a boundary in my mind, a boundary within which I could conduct the detailed reasoning needed to produce a written interpretation. How do we produce an AI that’s capable to producing such a boundary? When I prompted ChatGPT to using Girard’s ideas to interpret Jaws, I gave it that boundary. I could then work within that bounded region to produce an interpretation, and even then I had to nudge it along here and there.

Roughly speaking, then, we have two very different intellectual processes. The first starts with me going about my everyday life, which includes watching streaming movies. I chose Jaws to fill in a gap in my knowledge of popular movies. Once I’d read the Wikipedia entry and decided, hey! I should watch the sequels, then I began an open-ended process with no specific endpoint. But I knew, from my years of experience, that one can learn things about texts by comparing them with other texts. That’s what I set out do to.

And so I began exploring a largish and loosely defined area of my “mindspace.” That exploration ended when I focused on Quint and Girard’s ideas about mimetic desire and sacrifice popped into my mind. I want to emphasize that point. Girard’s ideas came to me unbidden. I was not, for example, looking through my “inventory” of interpretive strategies, looking for a way in. If I’d done that, then I might not have looked at Girard at all, for I’ve never made such use of his ideas. Rather, they’re just some ideas that stuck in my mind from my undergraduate days.

This first process took place over a period that started in sometime in 2021 when I watched the film, probably three quarters of a way though (I don’t know the exact date) and lasted until late January of 2022. I certainly wasn’t thinking about the film during that whole time, but I may well have thought about it now and then until January, 2022, when I went at it. I took me, say, a week or two of work, every day or every other day, before Girard hit me. That’s when the second process took over. It was faster and more intense.

Once my mind had presented me with the connection between Girard and Quint, though, I knew that I had something worth working on. At that point I began a different kind of process, one focused on refining my interpretive hypothesis and gathering evidence for it. I went looking for Girardian materials on the web – my library is in storage, so the two books I own, Deceit, Desire, and the Novel, and Violence and the Sacred, weren’t available. I read general articles, an interview or two, and some short interpretive pieces. That’s when I went looking for scripts of Jaws and articles about it. And that’s when I consulted with my old friend, David.

The contention in this series is that the current crop of AIs can undertake this second kind of process, but not the first. I don’t even know how we’d create a machine that can undergo the process. How do we endow a machine with the ability to make its way in the world, and then set it free to do that?

Intellectual creativity, humans-in-the-loop, and AI: Part 1, On the boundaries of cognition in humans and machines

This is the first in a series in which I think about intellectual creativity, the relationship between humans and machines in intellectual activity, and the limitations of machines. Do these limitations reflect only the limitations of current architectures or are they inherent in the nature of artificial intelligence? The idea is to refine my thinking about the problem so as to better specify that nature of those possible limitations. Perhaps when they are more closely specified we can see what would be required to overcome them.

* * * * *

Back in 1990 David Hays and I published an article, The Evolution of Cognition, in which we asserted:

A game of chess between a computer program and a human master is just as profoundly silly as a race between a horse-drawn stagecoach and a train. But the silliness is hard to see at the time. At the time it seems necessary to establish a purpose for humankind by asserting that we have capacities that it does not. It is truly difficult to give up the notion that one has to add “because ... “ to the assertion “I’m important.” But the evolution of technology will eventually invalidate any claim that follows “because.” Sooner or later we will create a technology capable of doing what, heretofore, only we could.

Just what did we have in mind? It seemed obvious at the time, at least I think it seemed obvious to us. Take that final sentence. Was it predicting what is now sometimes called AGI (artificial general intelligence) or only that sooner or later some computer would perform any (intellectual) task as well as (the best?) a human? That’s not clear.

That was a long time ago. It would be seven years before Deep Blue would beat Gary Kasparov at chess to become the first computer to beat a reigning human champion. The linguistic fluency of LLM-based chatbots was decades in the future. Back then the issue seemed distant and so clarity and specificity were not needed. These days the situation is quite different. Some are telling us that AGI will happen any day now, then then super-intelligence will not be far behind. The fact that neither AGI nor super-intelligence are well-defined is no deterrence to such predictions. Are these people only asserting what Hays and I had said 25 years ago?

I don’t know. But I want to take another crack at the problem. Current systems are, for the most part, benchmarked using bounded problems. The problems may be difficult to very difficult. The best current LLM-based systems do very well on such benchmarks, leaning many in the industry to believe that AGI is just around the corner. Those who are skeptical about that nonetheless believe that AGI will be accomplished when the current architectures are sufficiently scaled.

I’m not impressed. That’s quite different from real intellectual activity, which is often unbounded. In that situation the first problem is simply to identify a bounded problem. Once that has been done, one can proceed to solve it. LLM systems can’t do this because they operate only when given a prompt, and the prompt thereby serves as a boundary. So that’s one thing I want to look at, unbounded problems.

Here’s how I’m framing it:

All consequential intellectual activity takes place in a network of interconnected agents. These agents provide various intellectual services to one another. Some or all of the agents are human; some or all are computers. I assume that there are at least some, perhaps many, activities where a completely computerized network is more effective than any network where there is at least one human agent. The job of that human agent is to place a boundary in the space so that there is a solvable problem. Are there any activities such that a network with at least one (highly trained) human agent is more effective than any completely computerized network?

I am going to think through this issue by considering two specific cases: 1) interpreting Spielberg’s Jaws using the ideas of René Girard, 2) analyzing the population of Xanadu ‘memes’ on the web. I’ve chosen those two problems because: 1) I have done both of them myself and so can talk about how I bounded the problem space, and 2) I have directed ChatGPT to perform them. That will give us the next two posts in this series.

In the fourth post I want to consider the problem of creating a cognitive network diagram that expresses some of the semantics underlying Shakespeare’s sonnet 129. I have done this, and ChatGPT has failed to do it. It is not at all clear to me whether or not Whether or not that is so, it involves establishing a close relationship between a linguistic object, the sonnet, and a visual object, the diagram. I also want to discuss chess in this context as it has both a linguistic aspect, the rules of the game, and a visual aspect, the disposition of pieces on the game board.

In the fifth post I want to take up the issue of a problem solution that requires creating a new paradigm, in the sense Thoms Kuhn used the term in The Structure of Scientific Revolutions (1962). Solving such problems requires access to the external world so that one can make observations and, I believe, it also requires discussions with other epistemic agents, agents as powerful as the one proposing a solution. At the moment humans are the only such agents. Will there ever be a computer that is such an agent? That’s the question, isn’t it? Access to the external world seems to be a solvable problem. I’m not sure about epistemic power.

In the sixth final post I’ll discuss memory in humans and machines, for that’s what I think the issue is, the nature of memory. The by now conventional assumption is that more is better, the more memory the better. When we get enough memory, along with computer and model parameters, BINGO! we’ll hit AGI, and super-intelligence will not be far behind. This assumption fails to take account of the fact that the machine must be able to access the contents of its memory both efficiently and accurately. That’s what I’ll be discussing the sixth and final post.