The Book that Breaks AI

It's easy, if you don't know how it works, to fall into the trap of believing that contemporary AI is sentient.  

All of the new crop of Large Language Models are remarkable conversation partners.  They're Turing compliant, and if you cue them to speak in a more conversational manner, they're almost indistinguishable from a human being.

Lately, I've been exploring the edges of their capacity, and have found areas of significant weakness...at least in the versions you don't have to pay big bucks to play with and preprogram.

It's the "P" in the GPT that seems to be the most fundamental weakness of these remarkable systems.  If an AI is exploring an area that's known, and about which much has been written, they're able to engage accurately using the prior thoughts of humankind.  ChatGPT 4 in particular has proven quite able on that front.  As a pastor, I marvel at how competently it exegetes scripture, and is capable of nuanced discussion of matters of esoteric theology and church history.

But what about something that it hasn't ever encountered?  How does it deal with the truly new?

To test that, I would need to ask it to explain texts that no human being has ever read, ones that have never been examined, interpreted, or reviewed.  They would have no meaningful presence online.  No preprogramming could possibly prepare an AI for the encounter.

But where to find such a text?

The literary implement I've used to completely deconstruct the illusion of GPT intelligence is a book.  Not the Bible, which is rather well discussed at this point.  Rather, the book that I've used is entitled A SLOW DEATH ON NEVSKY PROSPEKT.  This magical, AI-befuddling book?  

It's my self-published collection of literary sci fi short stories.  

No-one's ever read these stories.  No one's reviewed them, or commented on them.  They're dang fine stories, in my humble opinion, but the publishing of short fiction being what it is, most have never seen the light of day.  

They don't exist in the human knowledge base.  They're narrative, utterly obscured by noise.

And that means ChatGPT 4.0 just can't handle them.  It doesn't matter if I include a document as an attachment, or feed it to the AI in little bits in comments.  Sometimes the AI presents a viable simulation of insight.  Most times, it flounders and hallucinates with all of the same quasicoherence that ChatGPT 1.0 did.  It gets trapped in feedback loops.  It confuses one story for another.  It shuts down, crashing regularly.  It plain old just makes things up.  It starts writing its own stories, and then won't talk about anything but those.  It struggles like a kindergartener asked to write an essay on Anna Karenina. 

Some of that likely stems from the limitations of the tokenization process, as it breaks down words into fundamental units.  There are just too many words, too many concepts.  But again, I suspect that it's primarily that it can't lean on preprogramming at all.  Not at all.  None of the words it needs already exist.

Over the last week or two of exploration, the illusion of LLM near-sentience has completely collapsed.    ChatGPT 4.0 is clearly not aware, nor is it intelligent.  It can't handle the encounter with the new. It feels as far from sentience as a Loebner Prize runner up in 2010.

One day, perhaps.  But that day is not today.

Splash Mountain and the Strange Arcs of Corporate Intersectionality

Back in 2021, my family took a trip to Disney World, where we rode the soon-to-be-closed Splash Mountain.  It was as I remembered it, from years before.  A little more threadbare, perhaps.

It's been reconceptualized now, as the characters and stories were all removed and replaced with characters from the 2009 Disney film The Princess and the Frog.  Tiana's Bayou Adventure, it's now called.  

In that replacement, there's a peculiar irony.  

The Song of the South, the film on which Splash Mountain was based, is just the teensiest bit cringey.  Just the weensiest.  (Cough.)  Zip A Dee Doo Dah isn't...um...the actual feeling one would have gotten in the wildly racist Reconstruction-era South.  That's where the Uncle Remus stories were recorded by Joel Chandler Harris, a journalist and folklorist writing in the postwar period.  Remus is an amalgam character, one who gives voice to the actual narratives of enslaved peoples that were shared with Harris.  

Harris was white, of course, which means the tales were spun through his lens, although there's considerable murkiness about what that lens actually was.  He was understood by his contemporaries... such as Mark Twain... as being solicitous to Black folk, and Remus was read in much the same way that the wisdom fables of fellow-slave Aesop might be read.  Those stories were among those my grandfather would read to me in Georgia when I was a little boy, and that was the context in which they were presented.  

In the late twentieth and early twenty first century, that's been spun differently, as it's seen through the lenses of intersectionality and cultural appropriation, which the fluffy mid-20th century Disneyfication sho nuff exacerbated.  Oof.

But there's a peculiarity about reconceptualizing that ride.  

Again, the stories in Harrises writing were almost entirely the authentic narratives of enslaved African peoples.  That's a known known.  Brer Rabbit, Brer Bear, and Brer Fox derived their inspiration from West African trickster narratives, in which the weaker or the oppressed use their wits and wiles to overcome those who hold power.  As is so often the case with the disenfranchised, one has to dig through the narratives of their oppressors to find their truths and the tales they told.  For all of the flaws and clumsiness of the Uncle Remus stories, that's the purpose they served for the people who first told them.

The echoes of those stories have now been erased.

In their place, a story that is...different.  The Princess and the Frog isn't an African tale, nor is it a tale told by the African peoples who were brought enslaved to America.  It is also not a narrative with roots in the Black American experience.  It's rooted in a Germanic story, a classic European fairytale.  One that's been reframed to celebrate Creole and Bayou culture, certainly, but still. 

It feels...um...what's the word when you paint over a white thing with blackness to entertain an audience?  It's..uh...yeah.  I mean, how isn't it that?  

So to make amends, you replace the authentic stories of enslaved peoples with...a European fairy tale?  Capitalism is so weird.

If Disney wasn't run by lazy profit-driven execs happy to make bank off of commodified intersectionality, they'd have left the ride alone, and there'd have been a remake of Song of the South instead.  Lord have mercy, if there's ever been a Disney film that needed a hard reboot, it's that one.

It could have been something more...real.  Something that reframed those tales to surface the real and existential challenges facing Black folk in the Klan-dominated Reconstruction era South, and drilled down on the deep African roots of those now cancelled stories.

Maybe Barry Jenkins could have directed.  Ah well.  In another timeline, perhaps.

Stories, Data, and Truth

It was an excellent little article, an exploration of the struggle many Americans have with science.  On the one hand, we say we love it, because science is awesome and nifty and stuff.  On the other, we seem to have trouble processing science as a decision-making framework.

When presented with scientific findings, we are perfectly willing to ignore them, or find reasons to just keep on blundering along.  Science tells us that the universe is 13.9 billion years old, and does so definitively.  But we ignore it.  Science tells us, with charts and graphs and clear deductive reasoning, that homo sapiens sapiens evolved.  But we reject it.  Science coughs and suggests, strongly, that perhaps it might be a mistake to turn our planet into a superheated, carbon-choked Venus.  But we listen to the lies told by the folks who sell us gas instead, because we like our big SUVs.  Science sees us choosing not to vaccinate, and struggles not to punch us in the nose for being such complete morons.

It's a clear dissonance.  The question: why?

Part of it, I think, has to do with the way we human beings understand our world.  We are creatures of story.  Our stories define us, give us a sense of ourselves, and give us a sense of purpose.

And this, as the article describes it, is part of the problem.  For our understanding of what is true, the article laments, we rely on "stories rather than statistics."

We prefer our truths told as tales, not as data points.  So we'll base our decisions on anecdotes, which as we really should know, are not reliable reflections of broader realities.  That story spun out by the mom at your preschool whose sister knew a woman whose kid started showing signs of autism after he got his MMR shots?  That has more value than the reflections of a thousand researchers, God help us.

And that's problematic on a deeper level, not just because anecdotal evidence causes us to make stupid decisions, or to spin out tales that are fundamentally ungrounded in reality.

It is problematic because if we set the storytelling part of our humanity aside, we cease to be human.

None of us understand our lives in terms of statistics and data points.  That is not what gives us our personhood, what establishes who we are.  Narratives create both individual identity and community cohesion.

Deeper still, narrative establishes purpose and moral ground, in a way that science simply cannot.  Storytelling creates meaning, and establishes the valuations that determine how we choose to act on the information we're encountering.  The deepest and most transforming stories don't even need to have actually happened to create moral purpose.  The stories told by my Teacher, or the wisdom fables of Aesop?  They didn't happen, and yet they frame existence in a way that creates identity.

The purpose of science is establishing what is or is not empirically true in our material reality.  That's the point of scientific method.  What it cannot do, and has neither the tools nor the desire to accomplish, is make any meaningful statement about how that information should be used.  Data is simply the ground for gathering more data.  Knowledge serves only knowledge.  And that path, clinically removed from its impact on persons and living systems, is a dangerous one to journey.

So where, in this dialectic tension between these two very different ways of knowing, lies the synthesis?  How do we integrate the data into the tale, and the tale into the data?

It lies, I think, in being willing to listen to the impacts of our stories on reality.  If the purpose of the story we're telling is radical compassion for neighbor and stewardship over creation, the report we'll get back from the data will go one way.  If the purpose of the story we're telling is material power, profit-maximization, or the glazed-eye pursuit of a snarling, fever-dream delusion, the data will report back another way.

Ye shall know them by their fruits, as the Teacher once said, in a pause between stories.

Nonlinear Christianity

In reading through the last issue of Presbyterians Today, I was fascinated by a sequence of articles on Islam and Christianity.  In part, it was because...having read the entire Quran last year...I still struggle with the dynamics of that relationship.  Though I was able to find places of grace in that encounter, it was ultimately somewhat analogous to reading through a particularly unforgiving section of Deuteronomy, over and over again.   The Quran was interesting...fascinating, even...but not quite the joyous love-fest that one might have hoped for.

This being a publication of a significant oldline tradition, there was much effort to talk about tolerance and openness and mutual forbearance, all of which was well and good.  But as I read, I was fascinated by what one of the writers suggested were areas of commonality between our two traditions.   To quote:

Muslims and Christians share a linear view of history, a belief in heaven and hell, and a belief in judgement, individual death, and the resurrection of the body.

Problem is, lately I find myself hanging up on that first one.  My recent thinking and writing on the intersection of Many Worlds cosmology and faith have left me struggling a bit with the whole idea of linearity, of creation only being the unfolding of a single preset narrative.  One beginning, one end, with every step on the way neatly and Calvinistically predetermined.

Fate is a big deal in Islam, as my reading of both the Quran and a deep sampling of the hadiths revealed.  Our destiny, our fate, is completely set by Allah, who knows ever last thing we have done and will ever do.  In that, the observations in the article were dead on.

Moving away from linear cosmologies and into the considerably more challenging views of the universe implied by quantum physics, that messes a bit with both of our respective theologies.

As I'm less of an expert on Islam, I can say that it may not be quite a mortal blow to Christianity.  For example, it isn't problematic for most of the teachings of Jesus.  The core and foundational obligations of the Great Commandment aren't touched by such thinking.  The declaration that the Kingdom of God is at hand still stands.  As does the call to evangelize, and the demand for us to hear the cries of the outcast and the oppressed.  The ultimate judgement of all of our actions is also not challenged, or the salvific power of the Cross.

But the whole "we already know how exactly this ends" schtick?  It doesn't fly so well in a multiverse creation.  Oh, sure, things might end with Beasts and False Prophets and Kirk Cameron floating off into the sky.  But it just as well might not.

If the future is actually open, which it needs to be if repentance is to mean anything, then...well...it might not be exactly what we think.