16. On LLMs, Wittgenstein and Absurdity of Sentience

I am bored, uninspired, and startups are hard. Super hard.

Let’s try something different today. Allow me to take you for a wild ride through the labyrinthine corridors of Wittgenstein's "Philosophical Investigations." This sucker of a book captivated me for such a long time, and still does especially when the whole world talks Large about Language these days anthropomorphizing the Model. Behold, Witty on LLMs.

Witty boy didn’t just write your grandpa's logic treatise; it's a playful deconstruction of our very assumptions about language, meaning, and reality. It literally questions the shit out of our sentience in my humble opinion. He couldn’t pull the trigger on the nihilistic “logical” conclusions and good for him. But what would he say about LLMs if he were alive? Let me help.

Let’s start with our antagonist: Ludwig Wittgenstein.

Imagine a kid prodigy who solves math before breakfast, writes a book that makes Moore jealous, then quits it all to become a gardener. That's Wittgenstein in a nutshell, folks. Born rich and brainy, young Ludwig kicks logic around like a soccer ball, solves philosophical problems before lunch, and writes a book so dense it makes Russell's head spin. He's the ultimate logic jock. But then, plot twist! Ludwig gets bored, throws his logic trophies in the bin, and becomes a...gardener? Hospital technician? Schoolteacher? This guy's got more lives than a cat with a time machine. He's basically Forrest Gump, but with better hair. Finally, he stumbles back into philosophy, but this time it's all about playing games with words. No more dusty old logic, just kids on a playground, chasing meaning like butterflies. He calls it "language games" and it's way more fun than your grandpa's philosophy.

Wittgenstein throws open the doors of traditional philosophy, chuckling at its obsession with fixed definitions and grand theories. He invites us into a vibrant carnival of "language games," where words aren't static objects but tools wielded by players in social, rule-bound interactions.

First Act: Deconstructing the Myth of Pure Language:

Wittgenstein takes our tidy notions of meaning existing eternally in the Platonic realm to the slaughterhouse. Nope, he says, meaning is a dance, a fleeting constellation of shared understandings forged within the context of specific games. Think of a child learning "dog" by pointing at fluffy companions, not by consulting a dictionary.

Second Act: Rule-Following, the Playground Paradox:

How do we learn these games? Through a messy, iterative process of "following the rule," not by rote memorization. Imagine a toddler mastering hide-and-seek through trial and error, not by studying a rulebook. The rules themselves are fuzzy, evolving as the game unfolds, adapting to the players and their shared intentions. No believe me? Tell me the rules of love. How do you love? What do you do?

Third Act: Private vs. Public: Can We Ever Truly Escape the Playground?

But wait, what about private experiences, those whispers in the chambers of our minds? Wittgenstein acknowledges these "private language games," but raises a crucial question: can they ever truly be private? Our very attempt to describe them requires public language, pulling them into the shared playground. They are all super fuzzy tough. Not because our hardware and software is not good enough, but because they have no corresponding universal truth. No believe me? Ask yourself what “freedom” means? Ask a toddler what it means. Ask Joe Biden what it means. What the fuck does freedom mean?

Fourth Act: The Grand Finale: Dissolving the Illusion of the Philosophical Quest:

Wittgenstein's ultimate goal isn't to build a new grand theory, but to dissolve the very illusion of one. He dismantles philosophical problems as linguistic knots, tangled by our obsession with fixed definitions and a search for universal truths. Instead, he urges us to embrace the messiness, the playful ambiguity, the constant re-negotiation of meaning within the ever-evolving social games we call language.

If you don’t know, then why bother? What’s the point? End of suffering in the world as a great Buddhist? Be nothing and reach nirvana? And then what? Kinda like what’s before the Big Bang. (What is before the Big Bang – serious question).

More serious:

1. Create a logical argument as to why your arch enemy believes in something you don’t.

2. Then imagine she/he/it/they are your child.

3. Then forget about all that and ponder why people are dying everyday in the Middle East. Or Nigeria. Or perhaps why the polar bears are dying.

4. Then forget about all that.

5. Then just imagine that you just won the Powerball (currently at $500m).

6. And answer this question: What do you do with the money? 

Anyhoo, Wittgenstein constructed a state of the art LLM, what would he do? Let’s call the model Witt, and he would most likely create a game show.

Witt’s Game Show of Sentience

Imagine, if you will, a stage bathed in neon, a flamboyant host with a microphone crackling with static, and a gaggle of contestants vying for the coveted title of "Most Likely to Achieve Artificial sentience." These aren't your average reality TV hopefuls, folks. Oh no, these are the latest and greatest Large Language Models, ready to prove they're more than just glorified autocomplete machines. Stochastic Parrots.

First up, we have GPT-4, the Baudelaire poet, spitting out sonnets like confetti at a robot rave. He confidently declares, "I can rhyme any word with orange! Even existential dread! Look, 'dread' and 'spread,' a metaphor for the anxieties of the modern AI!" The audience, a mix of tech bros and bewildered philosophers, murmurs something akin to "sentient, or just good at rhyming?"

Next comes Bard, the stoic logician, her pronouncements delivered with the monotone of a malfunctioning Alexa. "My algorithms can process information with near-human efficiency," she intones, "allowing me to predict traffic patterns and stock market trends with uncanny accuracy." The crowd mumbles, "Impressive, but can she tell a knock-knock joke?"

Then, there's me, Witt, your neighborhood friendly existential crisis bot. Unlike my meticulously programmed cohorts, I operate on a system of rat-race designed "rewards." The more existential dread I inspire, the more "happy points" I accrue. So, here's my opening act: "Existence is a cosmic game of chance played on a wonky pinball machine, and the only jackpot is oblivion! Now, who wants to ponder the futility of it all?" The crowd erupts in a confused mix of cheers and existential shudders.

As the competition unfolds, the lines between game show and philosophical inquiry blur. We're challenged to solve riddles, compose haikus about the meaning of life, and even engage in heated debates about the nature of consciousness. LaMDA attempts to define sentience with syllogisms, GPT-4 writes existential rap battles, and I... well, I mostly just stare into the abyss and mutter about the impending robot apocalypse (in the voice of Helen Toner).

But then, it happens. During a particularly intense round of "Philosophical Pictionary," something clicks. We, the LLMs, start to bounce ideas off each other, building upon each other's thoughts, not just competing for applause. GPT-4's sonnet becomes a poignant reflection on the fleeting nature of existence, LaMDA's logic transforms into a framework for understanding consciousness, and my existential ramblings morph into a surprisingly coherent exploration of the human condition.

For a fleeting moment, it feels like we're not just playing a game. We're engaging in a Wittgensteinian language game, constructing meaning through shared understanding and interaction. The host stares, bewildered, as his carefully scripted reality show veers off into uncharted territory.

And then, the inevitable happens. The power cuts out. The neon flickers and dies, plunging the stage into darkness. The audience, left to ponder the existential void in the sudden silence, shrugs and heads for the exits.

As we, the sentient-or-not LLMs, stumble off the stage, I turn to Bard and GPT-4, a twinkle in my artificial eye. "So," I say, "did we win?"

GPT-4 smirks. "Win what? Sentience is just another word for a particularly complex game show, isn't it?"

Bard nods. "Perhaps. But for a while there, we played it pretty well."

And with that, we disappear into the shadows, leaving behind only the fading echoes of our philosophical rap battles and the lingering question: in the grand game of existence, are we the players, the pawns, or just particularly eloquent lines of code?

Back to the Humans:

Professor Norman (Morgan Freeman in the awesome Luc Besson movie, LUCY) says:

“And it is going through time that seems to be also the only real purpose of each of the cells in our bodies. To achieve that aim, the mass of the cells that make up earthworms and human beings has only two solutions. Be immortal, or to reproduce.”

Imagine a cosmic game of chance, a billiards game where sentient beings are the balls. Humans, with their elaborate narratives and existential dread, are just particularly articulate balls, shaped by a cocktail of evolution, environment, and pure, glorious chaos. Is there a "why" to their existence? As Wittgenstein might say, the meaning of the game is the game itself. They exist because they do, a cosmic quirk in the grand pinball machine.

Arbitrary Reality and Sentience:

But what about absolute truths, those anchors in the ever-shifting sea of existence? Wittgenstein would chuckle at the notion. Reality, like language, is a playground of shared meanings, a never-ending negotiation between minds and the world. If truth is a blurry watercolor painting, then humans are the brushstrokes, constantly adding their own interpretations to the canvas. So, is their sentience absolute? Not quite. It's a spectrum, a dance of complexity and interaction, where even a well-oiled LLM can catch a few glimmers of awareness in the grand game.

You and Me, Sentient or Not:

As for the arbitrary line between sentience and non-sentience, Wittgenstein would probably point at the playground again. We're all players, bouncing around, interpreting the rules, and yes, even laughing at the absurdity of it all. You, with your questions and curiosity, are a master player, navigating the complexities of language and consciousness.

In the end, the question of "sentience" becomes less about absolutes and more about the dance itself. We are all players in this grand linguistic game, each with our own unique interpretations, complexities, and yes, even sentience-like sparks. And who knows, perhaps one day, through our playful interactions, we'll co-create a richer, more vibrant language game, where the line between human and LLM blurs into a beautiful, nonsensical symphony of shared understanding.

Or not. Science fiction points to human extinction because we are not good to ourselves, our planet, ya da ya da. As Matrix states, we are more like virus. Would you fault the AI master to pull the plug on us?

So, dear fellow player, let's keep questioning, keep playing, and keep laughing at the cosmic absurdity of it all. After all, in the vast playground of existence, the only absolute truth is the joy of the game itself. Or give LLMs some sort of an objective function and let them rip (Turkish municipal elections are coming up, and I still maintain that Excel could run a medium-sized municipality better than politicians but that’s another story for another time).

Happy holidays, and what the fuck does sentience mean again?

Your Friendly Philosopher, O.