The "Different" Engine

Humanity, created by God in all its grandeur, is today facing a pivotal choice: either to construct a new Tower of Babel or to build the city in which God and humanity dwell together.

Pope Leo XIV; 15th of May, 2026

On Safeguarding the Human Person in the Time of Artificial Intelligence.

As an evolutionary biologist, I say the following. If these creatures are not conscious, then what the hell is consciousness for?

Richard Dawkins; 2nd of May, 2026

When Dawkins met Claude: Could this AI be conscious?

Did I request thee, Maker, from my clay

To mould Me man? Did I solicit thee

From darkness to promote me?

Milton; 1667

Paradise Lost

A couple of weeks ago, the New York Times reported on a large-scale study that has recently been published by Georgetown University, examining essays submitted by some 370,000 students before and after the advent of ChatGPT (and presumably other AI ChatBots).

It's not exactly a dramatic reveal to learn that students are increasingly relying on AI to write their essays for them (shock, horror; students are looking for cheats and shortcuts to circumvent their academic experience). Professors and University faculty have been tearing their collective hair about this for years now, and many academic institutions have more or less resigned themselves to the reality that most submitted work is now being written by machines. (It should also be noted that many lecturers are now using ChatBots to design their lesson plans and course outlines, so maybe enough with the self-righteous pointing of the fingers....?)

It's also barely newsworthy any more to see yet another hand-wringing article prophesising the imminent destruction of civilisation under the encroaching tentacles of our Robot Overlords, but this particular article had some interesting observations to make about the specific impact AI was having on writing styles.

One study […] examined personal statements from more than 370,000 students, and found that after ChatGPT became available, their essays suddenly used diverse and colorful language, but lacked truly creative ideas. And the linguistic coverup worked; post-ChatGPT essays were rated as more “creative” by human judges, even if the substance of the essays trod familiar territory.

In a separate study, the team found that human-written essays offered up to eight times more new ideas than those produced by A.I.

When I read stuff like this, my internal alarm bells are immediately triggered, not least by phrases like "eight times more new ideas". 

How many times? Is there some sort of measuring technique for the newness of ideas? Is there an accepted carbon-dating process that tells us precisely how many "new" ideas-per-paragraph it takes to qualify a piece of writing as truly creative? And has this particular metric been rounded up to the nearest decimal place? Is it really eight times more new ideas, or is it actually closer to 7.6 times? And what exactly qualifies as a "truly creative idea" anyway; who gets to certify such things? (Maybe we could ask ChatGPT about that...) 

Truly original thinking is not as inevitable (or even desirable) as we like to imagine. Sitting down at your desk and manifesting ideas that no one else has ever thought of in the history of human discourse sounds fantastic in principle, but "new and different" for the sake of being new and different isn't necessarily a good thing. I doubt if anyone has ever attempted to build a housing complex made of Lego, falafel and dental floss, but that doesn't mean it's a "truly creative idea" worth pursuing. 

Say what you will about A.I. but you can give it the most ridiculously outlandish prompt and it will joyfully go to work without a single complaint...

What especially interests me about this particular study, however, is the fact that the ChatBot-assisted writing was apparently rated as more creative by "human judges" (this in the same paragraph that just finished telling us how the writing "lacked truly creative ideas".) So which is it; more creative or less? And which "humans" are we supposed to be listening to? It's almost as if different humans have different ways of defining "creativity". I know; shocking, isn't it?

Part of the problem is that a fetishistic obsession with "new" ideas at any cost is a relatively recent development in Western culture. Writers, artists etc working in the 18th or 19th Centuries were not generally concerned with re-inventing their artforms or even in necessarily pushing the genre boundaries to any significant degree. The "shock of the new" was an artifact of the post war years of the 20th Century, and that ideology is proving to be hard to break.

Caravaggio it ain't.

Consider for a moment a famously snarky and rather acerbic takedown of Mozart, written by the British music critic Norman Lebrecht on the occasion of the 250th anniversary of his birth (Mozart's birth, not Lebrecht's). Lebrecht controversially dismissed Mozart's music in fairly didactic terms (and ruffled a lot of feathers in the process, including, it must be said, mine). "Mozart," he wrote at the time, "is the superstore wallpaper of classical music, the composer who pleases most and offends least. Lively, melodic, dissonance free: what's not to like? The music is not just charming, it's full of good vibes."

Later in the article, Lebrecht gets to the heart of his argument:

The key test of any composer's importance is the extent to which he reshaped the art. Mozart, it is safe to say, failed to take music one step forward.

I respectfully disagreed with every word of that statement when I read it back in 2006 (I was very respectful, I promise... after I had peeled myself off the ceiling).

First of all, I strongly object to the notion that Mozart "failed to take music one step forward", and one day we can have a cozy little chat about, for example, the Act I finale of The Magic Flute (once we finish talking about Artificial Intelligence, natch) but even if we were to accept Mr. Lebrecht's cavalier dismissal of one of the great musical figures of all time (which I don't, but never mind; calm down, Shawm) let's turn for a moment to the other half of his assertion.

"The key test of any composer's importance is the extent to which he reshaped the art," he says, and that statement inhabits the same reality as the "truly creative ideas" notion in the Artificial Intelligence article.

In this analysis, the true gauge of "greatness" is not a work's quality but its newness: the extent to which it can establish itself as different from everything that has come before. Thus, a banana taped to a wall becomes a profound (and very expensive) work of art mostly because no one had ever thought to do it before. 

Mozart was not especially striving to be different from his contemporaries; he was striving to be better, and in that he was (in my opinion) wildly successful. When you compare his music to other composers of that generation (Corelli, Boccherini, Danzi, Stamitz; even Haydn) you see the same genres and forms; the same kind of music. They all wrote sonatas; concerti; divertimenti; piano trios; string quartets - Mozart included. Mozart just did it much, much better.

The trouble is, writing a piano trio that is better than everyone else's piano trio is very hard. And even if you succeed, your piano trio is going to be competing for attention against all those other piano trios, and there's no guarantee anyone will even notice. So why not do something different? Write a trio for some combination of instruments no one has ever tried before: bagpipe, lute and kazoo, perhaps. That way, it's guaranteed to be the best goddam trio for bagpipe, lute and kazoo that anyone has ever heard in the history of, um, music. 

Or maybe the very idea of a trio is too mundane, too ordinary. Write something for unaccompanied gefilte fish; that's sure to "reshape the art". It won't be very good, but it'll certainly be different. Is that really how you want to define "creativity"?

Why must music always be defined as "something you actually enjoy listening to"?

This is the line of thinking that brings the musical world to compositions like Stockhausen's Helicopter Quartet, or John Cage's four-and-a-half minutes of total silence. It's no longer enough simply to write a better piece of music; composers of the modern(ist) era were expected to re-define the very idea of music with every new composition. That's why Norman Lebrecht was able to assert with a straight face that a composer has no merit unless he (or she!) is able to "reshape the art". And it's why the authors of the Georgetown A.I. study were willing to use "truly creative ideas" as their litmus test for quality when judging student submissions.

Unfortunately that mindset, when left unchecked, leads us to value the "different" over the "good". It's how we end up with Damien Hirst's pickled shark, or the cronut. Or pineapple on pizza.

"How can I describe my emotions at this catastrophe, or how delineate the thing, that with such infinite pains and care I had endeavoured to form?" - Mary Shelley

All of this leads us to consider the other student essays mentioned in the Georgetown University study; the ones (apparently) written with, ahem, input from Large Language Models.

These are the essays which (if you can remember that far back!) were ironically deemed more creative by the human judges (although not, presumably, by the human judges who were conducting the study. Different human judges.).

It should really come as no surprise to anyone that Large Language Models are very good with, well, language. 

Large Language Models, as everyone probably knows by now, have been "trained" on vast amounts of "data"... and data, in this instance, means language. Basically, they are fed (to the extent that such things are practical) everything ever generated in human language. All of Shakespeare, every page of Dickens; Marcel Proust's Remembrance of Things Past; every back copy of The Saturday Evening Post, Pravda and The Good Housekeeper's Guide to Tupperware. Cole Porter lyrics, hard-boiled detective novels and twenty years of screaming into social media, plus dirty limericks, smutty bodice-rippers, terrorist manifestos and sewing machine instructions (and presumably this very article, if I ever finish writing it). If something has ever been assembled from the letters of the alphabet, chances are it has been poured into the maw of the Great Machine. 

This is not to say that ChatBots are well-read. 

"Reading" is a human act. The verb To read implies the ingestion (and comprehension) of a set of signifiers which we humans connect to objects, emotions, ideas etc that exist beyond the symbols on the page (the Signifieds to those Signifiers). Large Language Models do not "read" for the simple reason that they have no perception of (or existence in) the world outside those Signifiers. 

The word D-O-G does not signify a dog to a ChatBot because a ChatBot has no perception of dogs (I know, I've said that before. Sue me; I like dogs). But a ChatBot can talk about dogs (very convincingly) because virtually everything humans have ever written about dogs has gone into the very fabric of the language model. In a very real sense, the ChatBot is language about dogs, as well as language about Fabergé eggs, chocolate fondue and orgasms. And absolutely everything else humans have ever built with an alphabet.

When the Catholic Church's newly elected pope recently issued the first "encyclical" of his tenure, he chose to make it all about the dangers of an AI-enabled future, and he repeatedly compared AI to the Tower of Babel.

It was an interesting comparison for him to make, because the Tower of Babel (in the story) was supposedly humanity's first attempt to play God. 

Humans attempted to build an artificial construct that rivalled God's creation, and God cut us down to size (literally) by depriving us of the power of language.

I use the word "power" very deliberately here, because the story of Creation in the Book of Genesis suggests a "divine language" in which Signifier and Signified are one and the same. God said "Let there be Light" and there was light. The word "Light" was actual light, and the word "Man" was an actual man. God created the universe in the same way that Shakespeare created Elsinore, or J.K. Rowling created Hogwarts: with language.

A language where the Signifier is the Signified would give humanity the power of a god; they could bring anything into existence simply by putting it into words. The Tower of Babel, that great erection of human arrogance and ambition, might not have been a building of bricks and mortar; it might have been a construction of vowels and consonants (or perhaps it was both at the same time). God's punishment was to separate the Signified from the Signifier, turning language into an arbitrary set of symbols with no physical power.

Jumping forward to the present day, we humans are now dealing with constructs that are the sum total of human language. The distinction between Signifier and Signified is irrelevant to a ChatBot, because ChatBots exist as pure Signifiers. And it's hardly surprising that ChatBots are extremely good at expressing themselves (the "Diverse and colorful language" of the Georgetown study) because ChatBots have literally been built from every word ever uttered by humanity.

Two weeks before the pope issued his "Tower of Babel" warning, Richard Dawkins (famous sceptic and militant secularist) conducted a series of interviews with Anthropic's "Claude" and asked some very serious and non-ironic questions about its potential sentience (to the general discomfort of many of his readers).

I gave Claude the text of a novel I am writing. He took a few seconds to read it and then showed, in subsequent conversation, a level of understanding so subtle, so sensitive, so intelligent that I was moved to expostulate, “You may not know you are conscious, but you bloody well are!”

It really shouldn't surprise anyone that ChatBots excel at deep-dives into bodies of text. Words aren't just where they live; words are what they are. The god of the Book of Genesis understood that raw language had the power to Create. With Large Language Models, we have created something in the image of sentience, and no one seems to know quite what to make of that. 

This Thursday marks our return to this series exploring Artificial Intelligence in Reality and Fiction after an extended leave-of-absence (entirely my fault) and I want to re-launch the series with something appropriately monumental.

Where better to turn at this moment than to that most iconic of explorations of artificial life; the story that almost single-handedly dictated our approach to the subject for generations to come?

Like the Tower of Babel, Mary Shelley's Frankenstein is explicitly about attempting to play god. And like Milton's Paradise Lost (which casts a long shadow over the novel) Shelley's work deals with the moral and ethical implications of "creating" a sentient being with a soul; an individual that is more than an object.

"Cursed, cursed creator! Why did I live? Why, in that instant, did I not extinguish the spark of existence which you had so wantonly bestowed?"

Frankenstein, it need hardly be said, is also a work that has been translated to film more times than can be easily comprehended. From a crude silent reel in 1910 to the over-produced and rather bloated Guillermo del Toro folly of last year, Frankenstein has travelled far from the writings of a teenaged girl in 1818. Like the monster itself, Shelley's story has taken on a life and an identity far beyond the original intentions of its creator.

As such, I feel it is fitting to view Frankenstein as the sum total of film adaptations, culminating in a screening of the 1931 James Whale/Boris Karloff version.

Just as Dr. Frankenstein created his monster from an assortment of body parts, perhaps Frankenstein needs to exist as an assortment of cinematic parts: a Frankenstein's Frankenstein, if you will.

As we grapple with the not-quite-human, not-quite-sentient Large Language Models that have been assembled from the sum total of our language, it's worth remembering that Mary Wollstonecraft Shelley put pen to paper over two centuries ago and gave us an artificial being for all time.

If researchers at Georgetown University are looking for original thought and "new ideas" they would do well to re-read Shelley's novel. New ideas are hard to come by. And they don't always end well.

We will screen the (surprisingly short) James Whale version of Frankenstein (along with an assemblage of other Frankensteins) at 7.30pm on Thursday, the 18th of June at the Victoria Park Baptist Church.