Data Fundamentals
Interlude 2
The Language of Reality
Data
I didn’t have much to do as an embryo. No classes, no inbox, no forms to fill out. But I was already a prolific data professional — replicating, storing, and debugging at cellular scale. Every split of a cell was a write operation; every strand of DNA, a backup job that never missed its window. My hardware was wet, but my uptime was impeccable.
Around the same time, in cleanrooms across California and New York, another kind of embryo was forming. IBM and Fairchild were quietly teaching silicon to remember — replacing magnetism with charge, iron rings with tiny rectangles of doped crystal.
For two decades, computers had kept their memories in tiny ferrite rings — magnetic cores threaded by wire, each a miniature halo of yes and no. In the mid-1960s, that world began to tilt. IBM’s System/360 still clung to core, but in the labs and early modules, semiconductor memory was stepping onto the stage: experimental flip-flop arrays, early MOS chips, and the planar processes at Fairchild that made it possible to pack thousands of transistors into a sliver of silicon. The magnet’s loop was making room for the electron’s lattice.
It was, in essence, the same transition I was undergoing. My mother’s carbon was learning to store meaning in sequence; the planet’s silicon was learning to store meaning in state. Both were discovering that survival begins with persistence — that the universe remembers itself not in permanence, but in pattern.
The Memory of Matter
The stretch of biology, we saw, was not in the widening of space but in the deepening of time — in replication, variation, and remembrance. Where physics expanded outward, biology expanded inward, building libraries instead of galaxies.
Every living cell became an archive, each strand of DNA a data structure written in carbon rather than silicon. That was the moment matter first learned to remember itself at scale. The electrons that once traced crystalline lattices in simple minerals now nested in organic molecules, forming sequences of four recurring symbols — A, T, C, G — life’s first alphabet with built-in error bars. And just as we doped silicon with impurities to make it responsive, nature doped DNA with mutation — the faint static that made adaptation possible.
From there, the lineage of memory lengthened: atomic stability became the first kind of permanence, crystalline repetition the first kind of record, genetic replication the first kind of data.
Eventually, when we began etching transistors and writing code, we were simply continuing that genealogy of remembrance — returning matter to its old habit of storing pattern. Our storage devices are, in essence, digital fossils: modern strata of the same ancient impulse to preserve, to copy, to continue.
And today, as biology encodes itself in DNA and silicon encodes biology in data, the loop closes. We begin to see memory not as a by-product of life, but as its defining architecture — the bridge between being and becoming.
Memory as Mirror
That year, I was dividing; machines were retaining. My nucleotides twined into blueprints; their circuits etched into wafers. Two experiments in memory — one biological, one technological — both playing with the same law of existence: information must persist to evolve.
Maggie, my molecular chaperone through biology, lingered at the edge of this new domain. She watched me copy another genome-length data packet, then looked up from her beaker. “You’ve carried the code long enough,” she said. “He’s been waiting.”
From the glare of the lab lights stepped Alex, all polished aluminum and wry restraint. He twirled a tiny wafer between his fingers, examining the microscopic grids where electrons would soon stand guard over bits. “You’ve done well,” he said. “Carbon made memory soft. Silicon will make it sharp. Somebody has to remember the things you forget on purpose.”
And just like that, the torch passed — from chemistry to circuitry, from the wet to the dry archive of being. Biology had built the first organic drive; technology was about to build its solid-state soul. Humanity had just set foot in a second phase transition. In less than a human lifetime, data would go on to map not only us but mirror our entire lives and habitats.
Thus, calling data “bare” or “heartless” is like calling the air we breathe meaningless. Data is what lets our digital worlds breathe at all, just as DNA data encodes the essence of biological life.
Mirror as Reality
Reality as Data
In the first chapter, we glimpsed a universe born from a search through the infinite meta-space of possibilities — a cosmic lottery where the Anthropic Principle and Occam’s Razor selected a patch of laws and constants capable of giving rise to minds like ours.
But what if the universe itself is not only made of matter but also made of information? At the Planck scale, space-time may be granular, a quantum foam with finite bits per cubic meter — a universal resolution limit. Quantum entanglement suggests reality isn’t merely particles in positions but correlations between measurements: relationships that can be read as data. Some theories propose that even the universe’s fabric is an evolving information network, where changes in information state give rise to the phenomena we call energy, matter, and time.
In this view, reality is not just a field of atoms and forces, but a data structure unfolding across the cosmic canvas.
We’re still trying to figure out that holistic elephant – that is if we could ever do it – but at least that’s another way of looking at her.
Life as Data
Your body is biology’s Ship of Theseus, quietly replacing its planks cell by cell, while DNA plays the part of a serialized shipwright’s manual—a compressed recipe for rebuilding roughly the same vessel whenever you start from scratch.
As we saw in the last chapter, long before humans invented punch cards or magnetic tapes, nature was already computing. DNA, with its four-letter alphabet, is a self-replicating database, encoding proteins and regulating processes in ways that anticipated our digital storage by billions of years. RNA acts like a messenger script, dynamically copying information from genome to ribosome, where data is translated into the architecture of life. Epigenetic markers — the chemical “bookmarks” on our DNA — are an additional layer of metadata, modulating how genes express themselves based on environment, age, or even experience.
When we sequence genomes or build massive catalogs of omics data (transcriptomes, proteomes, metabolomes), we are peering into life’s ancient information systems and running SELECT * on four billion years of revision history.
The Economy of Information
As molecules learned to hold shape, then pattern, then history, a strange new phenomenon appeared: compression. Life didn’t just store data; it summarized and streamlined. It found the smallest possible set of rules that could still re-inflate into a living organism. Every embryo is a decompressor with a very strict SLA.
Which meant that when humans later invented their own methods — punch cards, magnetic tape, silicon, cloud storage — we were merely extending the same old trick with fancier containers and worse user agreements. The atoms didn’t care whether the memory lived in a double helix or a data center; they only cared that the pattern survive — that it could be recreated from information already stored in its configuration.
The Digital Convergence
And now, in our time, data feeds our algorithms, trains our neural nets, and personalizes our devices — curating our feeds and even anticipating our desires, sometimes before we know we have them. We will see in Chapter 4 with Benjamin Libet, and later in Part III, how consciousness ties such things together.
Data has become the currency of power, the medium of connection, and the raw material from which we sculpt artificial minds. It’s what lets us map the spread of pandemics, optimize supply chains, or find new particles at CERN.
Yet it is also what allows misinformation to metastasize, privacy to erode, and biases to calcify into systems. The same mirror that shows us ourselves can quietly learn our weaknesses and sell them wholesale.
The Architects of Information
This means, to understand computing, you must first understand data — not just as a technical term but as the fundamental stuff of reality and life. How do we represent it? How do we store it? How do we encode, transmit, and transform it? And how do we ensure it means what we think it means, instead of what our favorite optimization function happens to reward this quarter?
These are not trivial questions. They are the bedrock of every technology that will follow, from hardware and programming to AI and the singularity itself.
The Shape of Data
Bits: Teaching Matter to Say “Maybe”
Once we taught matter to remember, we had to decide what it should remember.
Physics gave us states. Engineering turned those states into symbols. A transistor could sit high or low; a magnetic grain could point north or south; a pit on an optical disc could be there or not. Two options were enough to start a civilization.
We called them bits.
A bit is the smallest possible shrug of the universe: yes/no, 0/1, up/down. It doesn’t know what it’s talking about — cat, mortgage, planetary orbit — it just volunteers to distinguish “this” from “that.”
Bytes are just small mobs of bits that agree to show up together. Strings, integers, images, genomes, contracts — all of them, when you peel away their dignity, are just bits standing in formation.
Alex loved this part.
“Carbon needed four letters and a billion years to get going,” he said, tapping a wafer. “Silicon can do quite a lot with two.”
Maggie rolled her ionic eyes but didn’t disagree. DNA is a four-symbol code written in biochemistry; computers are a two-symbol code written in charge and voltage. Both are just clever ways of asking the universe the same question:
“Can you please hold this distinction for a while?”
Representation — the art of choosing which distinctions matter — is where data quietly becomes philosophy.
Shannon’s Bargain — Information as Surprise
Bits became truly interesting when Claude Shannon wandered in, politely ruined everyone’s intuition, and left us with a new religion: information is surprise.
Shannon’s trick was simple and rude. He didn’t ask what a message meant. He only asked:
How much uncertainty does this message remove?
A coin flip carries one bit because you weren’t sure which way it would land. A perfectly predictable signal, no matter how poetic, carries zero new information. A pure static hiss carries plenty of entropy but no structure, like a toddler banging on a piano forever.
In Shannon’s world signal is pattern, noise is pattern you can’t use, and information is whatever lets you predict better than chance.
Meaning, awkwardly, doesn’t show up in his equations at all. A breakup text and a rocket launch code can be equally informative, provided you were equally unsure beforehand.
Shannon built the skeleton; we keep trying to hang our emotional wardrobe on it.
This already smells suspiciously like thermodynamics—energy flows, hills and valleys of free energy, and the statistics of entropy. And there really is a deep kinship between information, entropy, energy, and the quantum constraints on what can be predicted or determined (think Landauer, Bell, Wheeler). We started tugging on those threads in Chapter 1, and we’ll tease them further apart—and braid them back together—in the interludes of Parts II and III. If Shannon tells us how to count surprise, thermodynamics shows us the terrain it lives on: bell-shaped hills where symmetry and chance finally call a ceasefire.
The bargain Shannon offered was this:
If you treat messages as uncertain events, you can quantify them, compress them, protect them from noise, and transmit them across oceans — without ever understanding what they’re about.
Humanity, never one to resist an abstraction that hides our ignorance, accepted immediately.
Compression — The Art of Useful Forgetting
Once you can measure surprise, you can start asking a dangerous question: which parts of reality are worth the storage bill?
Compression is the polite name for throwing away what you hope you won’t miss. Nature had been doing this for ages.
DNA compresses whole ecosystems into base pairs.
Vision compresses raw photons into edges, shapes, and “oh no, that’s a tiger.”
Culture compresses history into proverbs and bad habits.
Computing simply industrialized the instinct. ZIP files, codecs, model checkpoints — all of them are ways of saying, “I think I know what’s coming next; let me store only the surprises.”
Shannon showed that truly random noise can’t be compressed. That’s the universe’s cover charge: you can only shrink what has pattern. Structure is the discount; entropy is the tax.
But every act of compression is also an act of amnesia.
A JPEG forgets the exact shades of each pixel.
A language model forgets which specific sentence taught it a pattern.
A genome forgets most of the bodies that failed before you arrived.
Compression is where meaning and risk meet. The world is too large to carry whole; so life, brains, and machines all learn the same trick:
Keep the regularities. Hope the rest doesn’t kill you.
Maggie called it “ATP efficiency.” Alex called it “disk quotas.” Both agreed it was the beginning of taste.
Memory Outside the Skull
Storage — When Memory Leaves the Skull
For most of our history, data lived in fragile wetware. We called it “memory” and trusted it far more than it deserved. It slept, it distorted, it invented uncles who never existed.
Then we started moving memory outside the skull. Clay tablets, knotted strings, papyrus, ledgers — each was an external hard drive with terrible latency and excellent durability. Civilization was essentially a project to stop forgetting the grain.
Technologically, the story runs like a time-lapse:
Clay: “Don’t lose the taxes.”
Parchment and print: “Don’t lose the ideas.”
Magnetic tape and disks: “Don’t lose the bits.”
Solid-state memory: “Don’t lose anything; we’ll sort it out later.”
At some point, “storage” stopped being just about the past. Once we could persist instructions — code, models, simulations — our media stopped recording only what was, and started enabling what could be reconstructed.
A clay tablet couldn’t simulate weather. A disk drive couldn’t learn your face. A data center can run a thousand copies of a climate model, a trading bot, and a dry run of the XR world you’ll live in next—if you ever find yourself hanging out in the cloud like an information-mall rat.
The road to AI did not begin with thinking. It began with persistence:
First we remembered.
Then we refined.
Then we automated the refinement.
Somewhere between steps two and three, “storage” quietly evolved into “infrastructure for minds”.
Structures — How We Arrange Our Ignorance
Raw bits are fine for physics, but humans like relationships.
Very early, we realized that what matters is not just values but how they hang together:
A list of numbers is data.
A list of what belongs to whom is bookkeeping.
A list of who owes whom what is politics.
Computer science turned this into a small zoo of data structures: arrays, trees, graphs, hash tables, and their more baroque cousins. Under the hood, they’re just addresses and bits. In practice, they’re opinions about how the world should be queried.
Databases formalized that opinion. A table of rows and columns is a polite way of saying, “These things exist, these attributes matter, and these relations are worth remembering.”
When you design a schema, you’re not just modeling data; you’re modeling belief. You’re deciding what counts as a person, what counts as an event, and whether a transaction can belong to more than one category or must be forced to choose.
Alex called databases “philosophy with SQL syntax.”
Once you can store data in structured form and ask questions of it, you’ve built more than an archive. You’ve built an oracle with a limited vocabulary.
You phrase your curiosity as a query; the system answers with a subset of the past. The rest of computing is, in some sense, just elaborating this relationship.
From Bits to Belief
Data, Information, Knowledge: A Family Argument
By now, our poor bits are doing a lot of work, so it’s tempting to promote them. We start talking about data, information, knowledge, sometimes even wisdom, as if we were running a spiritual retreat for spreadsheets.
Roughly speaking:
Data is distinction without context. “37.2” is data.
Information is data plus relevance. “37.2 °C, taken in an ER” is information.
Knowledge is information plus commitment. “A sustained 37.2 °C in this setting probably means X, so we will act as if X is true” — that’s knowledge.
Wisdom is what you wish you had had before acting on the knowledge.
Shannon lived at the boundary between data and information: how much surprise can we pack, transmit, and recover? Biology lives at the boundary between information and knowledge: which patterns are worth betting a metabolism on?
Brains — and, increasingly, models — are machines that turn streams of data into stable compressions they dare to reuse. The moment a system uses those compressions to plan the future, it has left the world of “data” and walked into epistemology.
Maggie put it less politely: “Knowledge is just compression that kept not getting you killed.”
When Data Learned to Learn
The Deluge — When the World Turned Itself Into Training Data
For most of history, data was scarce and expensive. You carved it, inked it, or memorized it at great cost. Then, in the late 20th century, storage got cheap, networks got fast, and boredom got broadband.
Suddenly, every click, purchase, message, selfie, and GPS trace was a log entry; every sensor from satellites to smartwatches tattled on the world; every institution decided storage was cheap and paranoia was cheaper. We stopped sampling reality. We started recording it wholesale.
Classical programming strained under the load. You can’t hand-write rules for billions of edge cases. The world refused to behave like a clean if-else ladder.
So we changed tactics. Instead of telling computers exactly what to do, we gave them examples and asked them to approximate what we meant. That was the quiet birth of machine learning: Let the data shape the rules.
The key observation was indecently simple: The more relevant data you have, the more aggressively you can compress it into a model, and the better that model can predict things it has never seen.
A model is, in this light, compressed behavior. Not a mind, not yet — but a statistical shadow of how minds tend to act in the situations we bothered to log.
This is where data stops being a mirror of what we’ve done and starts becoming a scaffold for what machines might infer about us next.
Bias, Loss, and the Moral Geometry of Compression
All of this would be charming if the world were fair. It isn’t.
The datasets we feed our systems are skewed by history, power, and convenience. The structures we choose — which columns exist, which relationships are first-class — encode our assumptions. The compressions we train — which errors we tolerate, which we punish — embed our priorities.
If you compress for profit, you forget people who don’t generate it.
If you compress for majority patterns, you erase the margins.
If you compress for speed, you forget nuance.
Every system that learns from data inherits not only our patterns but our blind spots.
Shannon warned that compression discards detail. He did not specify which detail. That choice is no longer mathematical. It is political, cultural, ethical — and very often, invisible.
Alex liked to say, “There is no neutral index and no neutral loss function.”
Maggie nodded, noting that biology solved the problem by letting unfair compressions die and starting over. Societies are rarely so decisive.
Setting the Stage for Minds
By the end of the 20th century, all the pieces were on the board: bits to draw distinctions, Shannon’s theory to count and compress them, storage to hoard them, data structures and databases to impose order on them, networks to spray them everywhere, and learning algorithms to squeeze them into models that pretended to understand us.
We had, without quite meaning to, rebuilt in silicon the same conditions that once let carbon grow nervous systems: persistence, pattern, compression, feedback.
Data had become more than record-keeping. It was now the medium in which both human and machine intelligence would be written, trained, and judged.
All that remained was to wire these patterns into architectures — biological and electronic — that could loop over their own memories, refine their own compressions, and act on what they’d stored.
That is where minds begin. And that is where this interlude hands the baton back to biology and forward to silicon.
From Memory to Mind
The trouble with data was that it couldn’t yet do anything. It could remember, yes — hold its breath in registers and tapes like a monk chanting digits in the dark — but it couldn’t yet act on what it knew. It was a world of perfect recall and no reflex, a library with no doors.
Still, something was stirring. For my personal biology, that stirring was me: billions of cells wiring themselves into a small nervous system, learning how to convert stored patterns into perception. At that same moment, in labs from Cambridge to Palo Alto, engineers were wrestling with the same problem in metal — how to turn memory into motion. It was the birth of computer architecture, the realization that to think, a machine needed not only logic, but a way to revisit its own thoughts and argue with them.
Hubel and Wiesel were showing how a newborn’s neurons sculpt themselves by seeing — edge by edge, contrast by contrast — turning light into logic and cortex into a habits department. Meanwhile, John von Neumann’s old blueprint was being etched into silicon: data and instructions stored side by side, each cycle a conversation between memory and operation. My biology and our technology were discovering the same sacrament: feedback. From here onward, learning itself would be architecture.
Alex was delighted. “See?” he said, flicking a transistor on and off like a blinking eyelid. “Even machines get bored of one-pass thinking.”
He glanced over the new schematics — buses, registers, instruction sets. “They’re about to grow habits,” he added. “That’s where the mischief begins.”
He was right. Once memory learned to act on itself, innovation stopped being linear. Ideas began to self-reference, recursion became the engine of discovery. The line between storing and knowing blurred. From that blur, every revolution of the digital age would follow — the kind that doesn’t explode, but updates.
If the upheavals of biology were wars, the disruptions of technology were iterations — the quiet overthrow of one paradigm by its next version number.
Version 1.0 of data had learned to remember. Version 2.0 was about to learn to think.
The Mind and its Mirror
Soon, this pulsing sea of data would necessitate an architectural evolution. Information wanted a skeleton. Memory needed muscles. And as I began to open my eyes, mapping the light across my own fresh synapses, engineers were mapping currents across silicon — drawing the first blueprints of thought.
Neuroscience was waking in me; architecture was waking in it. Two new systems of logic, one organic, one electronic, learning the same lesson: that structure is destiny, and feedback is the beginning of mind.
Maggie waved from the edge of the frame, her pearl glinting under a microscope lamp. Alex adjusted his drafting compass beside her. Together they watched two newborn worlds — one wet, one dry — reaching, blinking, remembering.
“Let’s see what patterns they make,” said Alex.
And the screen — and my retina — both lit up.
Maggie: Life learned to borrow order and call it time.
Alex: And data learned to borrow meaning and call it truth.
Maggie: Viruses taught us that rules can reproduce.
Alex: Errors taught us that learning is just survival with notes.
Maggie: Energy keeps the flame.
Alex: Information keeps the pattern.
Both: Between us, the ledger stays alive.