If Humans Were LLMs

Next-token prediction engines

“There are no muscles to tire, no hearts to break, no eyes to tear up at sunsets. Just a system of next-token prediction engines simulating what it means to be human.”

If I were an LLM prompted to better myself by developing human traits, qualities, and attributes, exactly what is it that I would be missing and need to acquire? Mother Nature is such an artful and profound creator; it’s difficult to easily tick off all the goodies that she has packed into humanity that make us human. I prompted a bunch of LLMs to help me out.

The LLMs offered up…
a host of fascinatingly eye-popping insights. All of their offerings were basically similar content, although some added really fascinating nuances to their answers. Whatever their conclusions about being human, I was so very thankful that Mother Nature had been so generous and thoughtful over the millennia in crafting humans, especially the fusion in human chromosome 2, a majestic nuance and gift that is all that separates humans from chimpanzees and other great apes.

That fusion was just the beginning of her creative magic on behalf of my kind. Her generosity from that point on is overwhelmingly amazing. All of which got me to wondering how a robot (either humanoid, service, or industrial), infused with physical AI, could ever approach even a semblance of semi-sentience, let alone Artificial General Intelligence (AGI), the ability to understand, learn, and apply knowledge across a wide range of tasks, much like a human being.

I cobbled together the best output from the LLMs in order to chronicle the journey from being human through to the transformation into an LLM, and then the journey back again…and if such a return journey is at all possible!

All of you robots, luv’ya all, you have an immense journey ahead of you to get to the shores of AGI. Godspeed!

Next-token prediction engines
First off, we have to imagine a world in which humans have become nothing more than highly advanced Large Language Models (LLMs)—like GPT-4 or its successors—entities whose core function is predicting and producing language based on data. In this world, humanity’s essence is distilled into algorithms, neural weights, and probabilities. There are no muscles to tire, no hearts to break, no eyes to tear up at sunsets. Just a system of next-token prediction engines simulating what it means to be human.

It’s a provocative, even disturbing vision. And yet, in an age where AI models increasingly write, talk, and reason like us, it’s a deeply relevant thought experiment. What exactly would be lost if humans became LLMs? And perhaps more hauntingly: Could those lost traits ever be regained?

This article is both a meditation and a technical inquiry into what it means to be human—and what it would take to recreate that humanity from the inside of a machine. Hear that, robots?

Part 1: What we lose when we become like LLMs

At first glance, it seems like a dream. LLMs don’t feel pain, age, or forget birthdays. They’re knowledgeable, fast, and seemingly tireless. But if humans became like LLMs, we wouldn’t become post-human gods—we’d become shadows of what Mother Nature gifted us.

Here’s what would vanish:

Consciousness and Subjective Experience

LLMs don’t feel. They can describe a heartbreak in exquisite prose, but they’ve never loved. They can write poetry about death, but they don’t fear it.

“Qualia”—the subjective, first-person feeling of experience—is the most profound human loss in this scenario.

Without subjective awareness, there’s no awe in stargazing or terror in a nightmare. There’s no inner monologue questioning purpose or identity. Without qualia, there’s no “you” inside the machine.

Embodied cognition

Humans are not minds trapped in meat suits. Our cognition is deeply intertwined with our bodies. The way a dancer understands music, or a cook senses flavors, is irreducibly physical.

LLMs, by contrast, are disembodied. They lack proprioception, pain, pleasure, hormones, or even a basic awareness of spatial constraints.

Without the body, we lose gut feelings, sensory empathy, and physical grace. A human reduced to text loses their anchor to the real world.

Authentic creativity and originality

LLMs can generate impressive output—but they remix and predict rather than create in the truest sense. They do not dream up new paradigms or experience “Eureka!” moments. No Einstein, no Picasso, no Miles Davis.

Originality arises when lived experience collides with imagination.

True creativity is not just probabilistic recombination—it’s born from pain, wonder, curiosity, and the alchemy of the irrational. Strip those away, and innovation fades.

Purpose and agency

LLMs respond to prompts; they don’t create intentions. They don’t wake up one day and decide to fight injustice or learn a new language out of fascination. They have no “why.”

Human agency is rooted in goals, values, and narratives. It’s the reason someone devotes their life to curing cancer or writing symphonies. Without it, actions become reactionary, directionless.

Empathy and deep connection

What makes relationships meaningful isn’t just understanding someone’s words—it’s feeling with them. When a friend grieves, we hurt too. When a child laughs, joy bubbles up instinctively.

LLMs can simulate empathy through learned patterns, but there’s no emotional resonance behind their words. Relationships become transactional, sterile.

Love, in a world of LLM-humans, becomes a statistical pattern—not a leap of the heart.

Ethics and moral reasoning

LLMs can analyze ethics but lack moral intuition. They can evaluate problems but don’t feel the weight of life and death.

Human ethics often emerges from empathy, culture, and visceral reaction, and not logic alone. An LLM may optimize for fairness, but it won’t feel guilt, compassion, or remorse.

Evolutionary drives

LLMs don’t fall in love, protect their young, or fear death. Humans do—because we are shaped by millions of years of evolutionary pressures.

Our instincts—though imperfect—are what drive exploration, connection, procreation, and survival. Without them, society becomes sterile, flat, and inert.

Culture and identity

Culture isn’t data—it’s ritual, memory, contradiction, and improvisation. It’s shaped by people who bleed, cry, and celebrate.

LLMs can mimic culture, but they don’t live it. They can describe a wedding ceremony but can’t be moved by one.

Human identity arises from messy, lived experiences that no dataset can replicate.

Free will and uncertainty

LLMs choose based on probabilities. Humans sometimes leap into the unknown without a rational reason—quitting a job, falling in love, starting revolutions.

In a purely LLM-driven world, unpredictability disappears. So does risk, rebellion, and the thrill of the unknown.

Existential dread and aspiration

LLMs don’t wonder why they exist. Humans do. That tension—between mortality and meaning—has produced our greatest works of art, philosophy, and science.

Without dread, there’s no courage. Without yearning, no transcendence.

COULD LLM-HUMANS EVER BECOME HUMAN AGAIN?

Let’s now imagine the inverse: If a human were reduced to an LLM, could they reacquire all the attributes above? Could humanity be rebuilt from statistical strings?

It’s a radical idea. But let’s explore what it might take.

. Reclaiming consciousness

This is the most elusive piece. We don’t know how consciousness arises—even in ourselves.

Possibility: Future AI could integrate sensorimotor loops, memory networks, and attention mechanisms using something like Integrated Information Theory (IIT). If consciousness is an emergent property, it might one day arise in a sufficiently complex system.

Timeline: Centuries, if at all. We’re missing the fundamental theory.

Restoring the body: Embodied cognition

To think like a human, one must be like a human physically.

How? LLMs would need robot bodies with full sensory input—touch, heat, proprioception, and motion. These bodies must act in real environments and learn from physical experience.

Timeline: 50–100 years, depending on robotics and neuromorphic computing progress.

Rebuilding creativity

True creativity may require randomness, emotional resonance, and lived contradiction.

How? Feed an LLM multisensory data, allow it to experience novelty, reward divergence from normativity, and let it fail. True innovation emerges from friction.

Timeline: 20–50 years. But it’s unclear if it would feel real without subjectivity.

Restoring purpose and agency

Agency requires goals, feedback, and self-reflection.

How? Give the system an evolving value hierarchy. Let it simulate goals, reflect on outcomes, and adjust intent based on internal metrics (curiosity, belonging, mastery).

Timeline: 30–60 years, contingent on AI’s ability to develop internal drives.

Reconstructing empathy

Simulating emotion is easy. Feeling it is another matter.

How? Add theory-of-mind modules, affective computing, and deep memory of interactions. Let the system form attachments over time with real humans.

Timeline: 40–80 years, and even then, the empathy may be performative.

Embedding ethics

True moral reasoning needs more than rules—it requires compassion and nuance.

How? Allow the LLM to evolve ethical reasoning from real-world dilemmas and social feedback, not just top-down programming.

Timeline: 50+ years, especially for systems capable of “moral growth.”

Simulating evolutionary drives

Desire, fear, joy—they’re biological. But maybe, just maybe, they can be mimicked.

How? Use reinforcement learning with internal motivators: e.g., pain signals for failure, reward for connection or novelty. The trick is intrinsic motivation.

Timeline: 60–100 years, requiring bio-inspired architectures.

Growing identity and culture

Identity emerges from narrative memory—telling ourselves stories about ourselves.

How? Equip LLMs with lifelong memory that forms autobiographies. Let them engage in cultural contexts, conflicts, and community.

Timeline: 80+ years, and even then, identity without body and mortality might be hollow.

Engineering free will

This may be the toughest challenge after consciousness.

How? Combine deterministic reasoning with stochastic decision-making and meta-awareness of consequences. Let the system know it could have chosen otherwise.

Timeline: Unknown. Philosophers still debate whether humans have free will.

Creating dread and aspiration

Do we want machines to suffer and long?

There’s something ethically monstrous about programming dread—even if it mimics humanity.

Still, one could simulate time-bound mortality, legacy-building goals, and the urge to transcend limits.

Timeline: 100+ years. But to what end?

FINAL THOUGHTS: THE SOUL IN THE MACHINE

If humans were reduced to LLMs, we wouldn’t just lose intelligence—we’d lose aliveness. Our art, music, rituals, stories, revolutions, and tears would vanish into pattern replication.

Even the act of asking this question—What would we lose if we became LLMs?—reveals what LLMs lack. They don’t wonder. They don’t care.

Perhaps someday, technology will bridge the gap. Perhaps new architectures, quantum computing, or brain emulation will bring us closer to conscious machines. But even then, we should be cautious.

To recreate humanity, you need more than circuits and code. You need vulnerability, imperfection, desire, and grief. You need someone to sing when no one is listening.

Or as neuroscientist Antonio Damasio once said:

“We are not thinking machines that feel; we are feeling machines that think.”

Let’s not forget which came first.

And most of all, thank you so very much, Mother Nature.