The deep history of AI began 3,000 years ago

Big Think Home Open search Open main menu

• Search Popular SearchesCritical thinkingPhilosophyEmotional IntelligenceFree Will Latest Videos Latest Articles
• Topics

  Philosophy

  • Ethics
  • Religion
  • Flourishing
  • Knowledge
  • Philosophy of Science
  • Philosophy of Art
  • Language
  • Political Theory
  • Identity
  • Meaning & Purpose

  Science & Tech

  • Physics
  • Biology
  • Aerospace
  • Health
  • Geology
  • Computing
  • Engineering
  • Energy
  • Biotechnology

  Mind & Behavior

  • Psychology
  • Neuroscience
  • Decision-Making
  • Mental Health
  • Consciousness
  • Emotional Intelligence
  • Personality
  • Relationships
  • Parenting

  Business

  • Entrepreneurship
  • Leadership
  • Finance
  • Marketing
  • Innovation
  • Strategy
  • Management
  • Artificial Intelligence
  • Startups
  • Economics

  History & Society

  • History
  • Literature
  • Art
  • Music
  • Film
  • Progress
  • Culture
  • Sociology
  • Policy
  • Geopolitics

• Videos Latest Videos The brain-deep emotion that matters more than happiness with Kate Bowler The hard problem of consciousness, in 53 minutes with Annaka Harris Why the search for alien life is about patience, not belief with Jill Tarter Thinking too logically can actually hold you back with Dan Shipper Is there a Planet B? An astrophysicist answers.  with Sara Seager What nihilism acknowledges that other philosophies don’t with Alex O’Connor See All
• Columns Mini Philosophy A philosophy column for personal reflection. Starts With A Bang An astrophysics column on big questions and our universe. Books A literature column to feed your curiosity. The Long Game A business column on long-term thinking. Strange Maps A geography column on history and society. The Well A collection of essays and videos on life’s biggest questions. 13.8 A column at the intersection of science and culture.
• Classes Featured Classes Members 7 videos How to Lead With Integrity Steve Stoute Founder and CEO, UnitedMasters and Translation Members 6 videos Transform Your Organization with AI Daphne Koller Founder and CEO of insitro. Members 10 videos Unlocking Your Team’s Hidden Potential Adam Grant Organizational psychologist and author Members 6 videos The Secrets of Unreasonable Hospitality Will Guidara Restaurateur and Author, Unreasonable Hospitality Members 12 videos How to Afford Anything Paula Pant Host, Afford Anything Podcast, Afford Anything Members 7 videos True Ingredients of Successful Leadership Atul Gawande Professor and author Browse
• More
  • About Big Think
  • Work with Us
  • Newsletters
  • Monthly Issues
  • Events
  • Careers
  • Our Mission
  • Get Big Think+ for Business
  • Freethink Media
  • View our Twitter (X) feed View our Youtube channel View our Instagram feed View our Substack feed

Sign In Membership

• My account
  • My Classes
  • My Account
  • My List
  • BT+ for my Business
  • Early Releases
  • Sign Out
• Membership

• Topics Back

  Philosophy

  • Ethics
  • Religion
  • Flourishing
  • Knowledge
  • Philosophy of Science
  • Philosophy of Art
  • Language
  • Political Theory
  • Identity
  • Meaning & Purpose

  Science & Tech

  • Physics
  • Biology
  • Aerospace
  • Health
  • Geology
  • Computing
  • Engineering
  • Energy
  • Biotechnology

  Mind & Behavior

  • Psychology
  • Neuroscience
  • Decision-Making
  • Mental Health
  • Consciousness
  • Emotional Intelligence
  • Personality
  • Relationships
  • Parenting

  Business

  • Entrepreneurship
  • Leadership
  • Finance
  • Marketing
  • Innovation
  • Strategy
  • Management
  • Artificial Intelligence
  • Startups
  • Economics

  History & Society

  • History
  • Literature
  • Art
  • Music
  • Film
  • Progress
  • Culture
  • Sociology
  • Policy
  • Geopolitics
• Videos Back Latest Videos The brain-deep emotion that matters more than happiness Happiness collapses the moment hardship arrives. Joy doesn’t. Historian Kate Bowler explains why joy can coexist with pain — and... The hard problem of consciousness, in 53 minutes “Our conscious awareness is everything. And the fact that it's still so mysterious to scientists and to all of humanity,... Why the search for alien life is about patience, not belief Thinking too logically can actually hold you back "Rationalism is the idea that, in order to truly know something, you have to be able to describe it explicitly." Is there a Planet B? An astrophysicist answers.  30 years ago, we didn’t know other stars had planets orbiting them. Now, we may be on the verge of... What nihilism acknowledges that other philosophies don’t "The thing that the nihilist recognizes is that the values he or she holds are not grounded in anything other... The terrifying ways that social media is altering teenage brains Why social media is the perfect recipe for kids to become addicted to their smartphones. The computing revolution that secretly began in 1776 "In the process of mapping the heavens, it doesn't take long to realize the data problem they generated."
• Columns Back Columns Mini Philosophy A philosophy column for personal reflection. Starts With A Bang An astrophysics column on big questions and our universe. Books A literature column to feed your curiosity. The Long Game A business column on long-term thinking. Strange Maps A geography column on history and society. The Well A collection of essays and videos on life’s biggest questions. 13.8 A column at the intersection of science and culture.
• Classes Back Featured Classes Members 7 videos How to Lead With Integrity Steve Stoute Founder and CEO, UnitedMasters and Translation Members 6 videos Transform Your Organization with AI Daphne Koller Founder and CEO of insitro. Members 10 videos Unlocking Your Team’s Hidden Potential Adam Grant Organizational psychologist and author Members 6 videos The Secrets of Unreasonable Hospitality Will Guidara Restaurateur and Author, Unreasonable Hospitality Members 12 videos How to Afford Anything Paula Pant Host, Afford Anything Podcast, Afford Anything Members 7 videos True Ingredients of Successful Leadership Atul Gawande Professor and author
• My Account Back
  • My Classes
  • My Account
  • My List
  • BT+ for my Business
  • Early Releases
  • Sign Out
• Sign In
• Membership
• More Back
  • About Big Think
  • Work with Us
  • Newsletters
  • Monthly Issues
  • Events
  • Careers
  • Our Mission
  • Get Big Think+ for Business
  • Freethink Media
  • View our Twitter (X) feed View our Youtube channel View our Instagram feed View our Substack feed

View our Twitter (X) feed View our Youtube channel View our Instagram feed View our Substack feed Search Popular SearchesCritical thinkingPhilosophyEmotional IntelligenceFree Will Latest Videos Latest Articles The deep history of AI began 3,000 years ago

AI is not a rupture in history, but a continuation of intelligence emerging where information becomes systematically arranged.

by Joel J. Miller February 9, 2026

Credit: Big Think

Key Takeaways

• Artificial intelligence is the continuation of cognitive systems humans have been building for thousands of years.
• When information is organized — catalogs, indexes, metadata — and grows beyond human scale, it migrates from minds into systems.
• Modern AI continues this historical trend with the difference that today’s algorithms can infer procedures statistically from data based on their training.

In 58 BC, Cicero’s house was ransacked. Returning from exile, the Roman statesman found his property vandalized; his scrolls jumbled, torn, and scattered. A library assumes an order, a schema, something that renders it sensible and accessible. Cicero’s was chaos.

Enter Tyrannio, a Greek specialist in literature and libraries, owner of some 30,000 scrolls and famed expert on Aristotle — in fact, the same man responsible for restoring the philosopher’s tattered library after it was hauled to Rome. Tyrannio stepped in to sort through Cicero’s mess. He identified volumes, repaired damage, organized the scrolls, and created title tags. Cicero marveled at the transformation.

“You will be surprised at Tyrannio’s excellent arrangement in my library,” Cicero wrote to his friend Atticus. When the work was complete, his appreciation verged on the mystical. “Since Tyrannio has arranged my books,” he wrote, “the house seems to have acquired a soul.”

It’s a poetic but poor translation. Cicero used the Latin word mens — not “soul,” but “mind.” Another translation says the house “recovered its intelligence.” And he’s not talking about Tyrannio’s intelligence, but rather the intelligence Tyrannio imposed on the library, a structure that reflected Tyrannio but now existed independent of him.

Once organized, Cicero’s library possessed a discernible — even if artificial — intelligence. It now had a mind of its own. Properly arranged, the library could suggest connections, reveal patterns, answer questions, and synthesize disparate ideas. It facilitated and extended Cicero’s thinking beyond what his biological brain could manage on its own.

But the library’s intelligence didn’t depend on Tyrannio or Cicero’s ongoing presence. Anyone could now use it. The organizational structure itself projected coherence — for instance, title tags indicated location, and categories showed relationships. These are procedures that can be learned and followed. Just as a syllogism’s validity doesn’t depend on who states it, a well-organized library functions through impersonal operations. Once intellectual work becomes procedural, anyone can follow the process.

Accumulated information suggests systems and arrangements that make it comprehensible. “Our role as humans,” as Wired cofounder Kevin Kelly argues in What Technology Wants, “is to coax technology along the paths it naturally wants to go.” What we call artificial intelligence is the latest expression of a pressure that has been building since humans first began organizing libraries and archiving data — which means the birthplace of AI isn’t Silicon Valley; it’s more like the Ancient Near East.

To show you what I mean, I want to tour the last few millennia with a view toward how humans have organized and processed information and how ever-increasing amounts of data have suggested new and more robust ways of managing those feats. We’ll follow a track that runs through the ancient and medieval worlds, into the early modern and modern eras, and finally to our own day. Along the way, we’ll see how different solutions to the challenge of information access slowly nudged us toward machine intelligence and large language models, not as inevitabilities but as the result of humans solving problems following the suggestions inherent in managing data.

The original search engines

Thirty-three hundred years ago, Šuppiluliuma I ruled the Hittite Empire from his capital at Ḫattuša. His palace housed a library containing as many as 7,000 clay tablets, carefully organized and cataloged. He kept this archive to aid his decision-making. When, for instance, Pharaoh Tutankhamun died and his queen required a new royal spouse, Šuppiluliuma could call up an existing treaty with Egypt to determine whether the arrangement warranted serious consideration.

“Hittite kings were able to order specific tablets from their tablet rooms,” Ancient Near East scholar Theo van den Hout writes. How? The Ḫattuša archivists developed a catalog and corresponding metadata — titles taken from the first words of each text, brief descriptions of contents, genre classifications, and the like. Someone could scan the catalog, spot what he was after, and retrieve a single tablet or series of tablets for consultation. That might seem unremarkable in the age of Google and large language models, when knowledge of any kind awaits only a few keystrokes. But in the 2nd millennium BC, such recall was astounding.

As the amount of data that rulers collected and deployed expanded, solutions needed to keep pace. The job was considerably bigger when Ashurbanipal ruled the Neo-Assyrian Empire 2,700 years ago. A ruthless lord who humiliated his enemies and killed lions for sport, Ashurbanipal was also — incongruously? — a bibliomaniac. He kept an army of scribes and, when conquering neighboring kingdoms, prized books as plunder.

Cuneiform tablets from the Library of Ashurbanipal (circa 1500-539 BC) are housed at the British Museum in London. (Credit: Wikimedia Commons)

Archaeologists unearthed Ashurbanipal’s palace at Nineveh in the 19th century. They discovered several thousand works on some 30,000 clay tablets — texts covering omens, rituals, language, medicine, and more. The challenge for Ashurbanipal’s archivists was organizing all that information. Archivists created inventories and catalogs, attached labels identifying genre and content, and developed filing systems using baskets, shelves, and wall niches. They even kept incoming acquisitions listed on waxen writing tablets that could be erased and reused: ancient databases that could be updated in real time. Properly organized, libraries like Ashurbanipal’s could store thousands of documents for use — categorized, searchable, and available when needed. An internet of clay and sandal leather.

The game only intensified as holdings grew. Three and a half centuries after Ashurbanipal, the Great Library of Alexandria faced the challenge of organizing papyrus scrolls running in the hundreds of thousands. Alexandrian librarians employed different forms of metadata: attaching title tags to scrolls, establishing categories and genres, storing similar books together, and assembling a massive 120-volume alphabetical catalog that provided author names, professions, line counts, and more. In his cheekily titled book Index, a History of the, University College London professor Dennis Duncan referred to the whole operation as “Greek Big Data.”

Notice the emerging pattern: The solutions are remarkably similar across centuries and cultures because the problem is constant. Accumulated information demands an accessible organization and internal logic. Once that logic is established, knowledge becomes procedural. Classification, retrieval, and association become operations that can be learned and repeated independently of any particular person.

Gutenberg ups the ante

European scholars working with expanding collections in the 13th century naturally faced the same sorts of challenges. Robert Grosseteste of Oxford sought a means to retrieve his learnings from his vast range of reading. He devised an extensive tabular index of all the subjects he encountered in classical and patristic texts. Duncan described it as “a Google on parchment that takes its subjects and explodes them across the whole of known literature.” Around the same time in Paris, Hugh of Saint-Cher broke down the Bible and rearranged it into a massive alphabetical index of over 10,000 terms.

It wasn’t long before these sorts of indexes morphed into the back-of-the-book technology and reference works that have smoothed the way for scholars ever since — and just in the nick of time. By the mid-15th century, printing had arrived. It triggered an explosion of information and intensified the pressure toward organization and discoverability.

The advent of print created, as French historian Frédéric Barbier put it in Gutenberg’s Europe, “a phenomenon of mass mediatization.” Studies estimate that European scribes produced approximately 11 million books over the 900 years before the printing press. By comparison, in the 150 years following Gutenberg, presses churned out over 200 million. The archivists and librarians of Ḫattuša and Alexandria could find emergent order in their holdings, but those collections paled in comparison to the flood of print. What would scholars do now?

Old answers found new applications. The bastard son of Christopher Columbus, Hernando Colón, built a library of more than 15,000 books in Seville, Spain. More impressive, however, was the analog network of hypertexts and hyperlinks he devised for navigating his collection. 

The problem was keeping up; information tends to accumulate faster than we can manage it — a pressure that has only increased in the modern world.

With so many books to explore and limited hours to employ, the library’s readers needed a means of deciding which books warranted perusal. Colón’s Libro de los Epítomes (Book of Summaries) spanned multiple volumes and provided condensed versions of the works in his collection, along with metadata such as content details, author biographies, and writing styles.

These summaries were all cross-referenced with other catalogs Colón developed. His Libro de las Materias (Book of Subjects) cataloged not just the broad subject matter of books but all the individual topics covered within each. Working with the Materias, a researcher could follow the trail of a subject through history, philosophy, theology, poetry, the Bible, whatever — freely across categorical distinctions such as author and genre. Suddenly, the overwhelming ubiquity of books became not a burden but a benefit, as the index could help readers find choice bits while encouraging serendipity along the way.

The 16th-century Swiss scientist Conrad Gessner tackled the problem in a similar way. While still in his mid-20s, Gessner envisioned a Library of Universal Knowledge. He labored three years to build an elaborate, annotated, alphabetical bibliography — the Bibliotheca universalis — totaling 1,300 pages and featuring some 10,000 works by 3,000 authors. Then, three years later, he produced the Pandectarum, a massive index to his bibliography.

The pattern inherent to solutions developed by men such as Grosseteste, Hugh, Colón, and Gessner matches that of the ancient world, but is adapted to the compounded challenge of ever-increasing amounts of information. The problem was keeping up; information tends to accumulate faster than we can manage it — a pressure that has only increased in the modern world.

An electromechanical internet

By the late 19th century, the industrialization of knowledge had reshaped the information landscape. The growth of institutions, powered by steam-powered rotary printing presses, engulfed readers with more books and articles than they could handle. Just as enterprising librarians and archivists met the difficulties of their day with novel solutions, now a new generation looked for ways to adapt to the overflow.

Born in Belgium in 1868, Paul Otlet was appointed librarian of his Jesuit school at just 16. “It seemed a wonder,” he later said of the library’s cataloging system, “this instrument that allowed me to use all of these books.” 

But Otlet could already see that traditional catalogs wouldn’t suffice against the onrush of industrialized publishing. The solution: What if you atomize books by breaking them down into their smallest features? Researchers could then access the essential bits, per their particular needs, and synthesize the information in new and helpful ways.

“The ideal,” Otlet explained, “would be to strip each article or chapter of whatever is fine language or repetition or padding and to collect separately on cards whatever is new and adds to knowledge.” 

As he saw it, information could be reduced to four key components: facts, interpretations, statistics, and sources. Otlet first experimented with cutting lines out of books and pasting them onto index cards. With proof of concept, he later moved to typing the information on cards.

Paul Otlet (middle) at the Zurich Bibliographical Conference in 1930. (Credit: Mundaneum / Wikimedia Commons)

Essentially, Otlet unbound the book and extracted its contents for a different mode of engagement. But his real breakthrough went beyond atomization. Traditional library classification follows hierarchical structures. The Dewey Decimal System, for instance, follows a linear path, beginning with broad categories and narrowing to specifics. This tends to bento-box subjects in unnatural ways. By abandoning standard categorization, Otlet could tag a single fact with multiple topics, time periods, and other relevant identifiers. Suddenly, categorization could work both laterally and linearly.

The archivists of Ḫattuša and Alexandria proceduralized knowledge; Otlet created more robust procedures to meet the demands of rapidly expanding information. Working throughout the 1890s, Otlet built an industrial-scale version of Hernando Colón’s vision. When Otlet and his business partner debuted their creation at the 1900 Paris Exposition, they’d already produced more than 3 million index cards, a fifth of their eventual total of 15 million. Their goal was to assemble, as Otlet put it, “an inventory of all that has been written at all times, in all languages, and on all subjects” — neatly arranged in miniature filing cabinet drawers and accessible by telephone query.

Insufficient funds, wobbly state support, and two world wars shipwrecked the project, but Otlet had identified what accumulated information needed next: not just organization but instant retrieval across dimensions. His vision evolved into something remarkably prescient: a vast multimedia network combining his databank with telephones, television, radio, and film. His vision amounted to an electromechanical internet. And others were beginning to think along similar lines.

A machine for thinking

“We are becoming bogged down,” the American engineer Vannevar Bush said in 1939. “There is a growing mountain of research results; the investigator is bombarded with the findings and conclusions of thousands of parallel workers which he cannot find time to grasp as they appear, let alone remember.”

There was a cost to this bombardment. “Mendel’s concept of the laws of genetics was lost to the world for a generation because his publication did not reach the few who were capable of grasping and extending it,” Bush wrote in “As We May Think,” an article published in the July 1945 issue of Atlantic Monthly. “And this sort of catastrophe is undoubtedly being repeated all about us, as truly significant attainments become lost in the mass of the inconsequential.” The problem, Bush emphasized, went well beyond scientific research: “It involves the entire process by which man profits by his inheritance of acquired knowledge.”

Bush painted a vivid picture of the trouble. He asked readers of his 1933 essay to imagine a professor in a library: “Long banks of shelves contained tons of books, and yet it was supposed to be a working library and not a museum. He had to paw over cards, thumb pages, and delve by the hour. It was time-wasting and exasperating indeed.”

The solution? Bush envisioned an analog personal computer, dubbed the Memex. A desk would contain reels of microfilm, a keyboard, screens, and an array of buttons and levers. Users could input codes to quickly search the data, call up content on twin slanted screens, and manipulate documents with levers.

Foreshadowing the hyperlinks and hypertexts of the internet, Bush imagined users could create connections between documents called “associative trails.” Unlike linear connections between concepts, an associative trail could link items flexibly, moving between subject areas, even entire fields, joining text, images, and audio recordings into novel configurations. “It is exactly as though the physical items had been gathered together from widely separated sources and bound together to form a new book,” Bush explained.

A modern interpretation of Vannevar Bush’s Memex device is found at the Berlin Technical Museum. (Credit: Bunyk / Wikimedia Commons)

But more than a dynamic library of documents, he envisioned the Memex serving as a research librarian too. As Bush’s vision evolved between the 1930s and 1960s, he imagined the Memex shouldering some of its users’ cognitive burdens — observing preferences, updating memory, creating dossiers and research briefs autonomously. The machine would have a form of intelligence that users could harness to enhance their own.

“For mature thought there is no mechanical substitute,” Bush acknowledged. “But creative thought and essentially repetitive thought are very different things. For the latter there are, and may be, powerful mechanical aids.” As he saw it, man would begin to create “machines to do some of his thinking for him.”

Bush picked up the key contribution of the ancient archivists and librarians — proceduralizing knowledge — and took it a step further: automation. “We may some day click off arguments on a machine with the same assurance that we now enter sales on a cash register,” he said.

It’s not so big a leap as it might appear. Once you formalize intellectual operations into procedures, those procedures become inherently automatable. An argument is, after all, essentially an algorithm: a sequence of steps that produce a particular outcome. Depending on technical limitations, a machine can be programmed to go through the motions as easily as a human. Unfortunately for Bush, his vision outstripped what the technology of his day could manage.

Procognitive systems

J.C.R. Licklider, director of the Information Processing Techniques Office at ARPA from 1962–64, pushed Bush’s vision still further. In 1960, he published “Man-Computer Symbiosis,” an essay arguing people could “think in interaction with a computer in the same way that you think with a colleague whose competence supplements your own.” After studying his own practice, Licklider determined that as much as 85% of his intellectual work — such as searching for data, running calculations, testing conclusions, and synthesizing information — was fundamentally preparatory for deeper, more creative efforts.

What if machines handled some of that work? To accomplish this, Licklider proposed atomizing books into discrete blocks of information — exactly as Otlet had suggested more than half a century before, but now assisted by digital tools rather than index cards. Once computers digested the data, they could execute predefined procedures — searching, associating, transforming — and serve it back reconfigured, updated, elaborated. Licklider called these systems “procognitive” since they actively contributed to cognitive labor.

“A basic part of the overall aim for procognitive systems,” Licklider wrote in a 1965 report, Libraries of the Future, “is to get the user of the fund of knowledge into something more nearly like an executive’s or commander’s position. He will still read and think and, hopefully, have insights and make discoveries, but he will not have to do all the searching himself nor all the transforming, nor all the testing for matching or compatibility that is involved in creative use of knowledge.”

But that wasn’t all. Licklider then described a process that sounds remarkably like a person today interacting with a large language model, such as ChatGPT or Claude: “He will say what operations he wants performed upon what parts of the body of knowledge [essentially a prompt], he will see whether the result makes sense, and then he will decide what to have done next.”

As this trajectory shows, we didn’t invent artificial intelligence in the 21st century. We’ve been building it for 3,000 years.

Foreseeing more than information automatically processed by these “procognitive systems.” Licklider imagined users able to access data from distant locations. His 1963 memo — addressed, tongue in cheek, to “Members and Affiliates of the Intergalactic Computer Network” — described how information stored in one locale could instantly be “brought into the part of the system that I was using.” Information would go from bounded to ubiquitous.

There’s a direct line from Licklider’s work to ARPANET, a precursor to the internet and World Wide Web, through the work of Robert Taylor, Doug Engelbart, Ted Nelson, and Tim Berners-Lee.

When Berners-Lee proposed the Web in 1989, he cited the ancient problem: “Often, the information has been recorded; it just cannot be found.” His solution? “A web of notes with links“ that “could grow and evolve.” And he anticipated the dynamic we’ve seen so many times already: data begets data, and the solutions to manage it facilitate the creation of even more, outpacing those solutions and demanding new methods to make information accessible.

Once the Web grew large enough, Berners-Lee suggested it would need “automatic analysis” — that is, artificial intelligence — to remain navigable. “This is particularly useful,” he said, “when the database becomes very large, and groups of projects, for example, so interwoven as to make it difficult to see the wood for the trees.” 

The vast digital library would need a librarian, and because working with knowledge could be reduced to a set of procedures, that librarian could be a machine.

The 3,000-year deliverable

As this trajectory shows, we didn’t invent artificial intelligence in the 21st century. We’ve been building it for 3,000 years. Libraries, catalogs, metadata, and indexes did more than anticipate digital databases and search engines; they created the conditions that made such systems possible and even probable.

Humans excel at creating information. For millennia now, we’ve created more than we could easily manage with our native intelligence. As a result, we’ve created systems and procedures to shoulder part of the cognitive burden. As the floodwaters of data rose, those procedures suggested further developments to manage the deluge. 

Growth in data volume created the pressure. Faced repeatedly with the same challenge — making ever-larger bodies of information usable — humans kept responding the same way: discovering the logic implicit in the material, codifying it, and offloading it into a structure.

Each organizational choice created affordances and constraints that shaped subsequent innovations. Once you organize alphabetically, indexing becomes possible. Once you create cross-references, associative trails and hyperlinks become feasible. And once you formalize operations, automation becomes possible. Seen in this light, it’s a story of path dependence. We didn’t arrive here by accident; we followed a path shaped by the problem of scale and suggested by earlier solutions, coaxing the tech in the direction it pointed.

When Cicero said his library possessed a mind, he wasn’t claiming it was conscious. Nor was he saying that it merely reflected an intelligence imposed on it. Once imposed, the system no longer required the systematizer to facilitate thought. Same with Ḫattuša, same with Alexandria: By depersonalizing knowledge and rendering it into operable procedures, anyone could step in and get to work. And the more robust and predictable the system, the more likely it is that we could train the system itself to run some of those procedures.

The difference between ancient catalogs and modern AI isn’t the appearance of a new kind of sentience. It’s that operations long externalized from human minds — search, association, transformation, recombination — can now happen automatically at speeds that obscure their continuity with earlier systems. But the continuity is real, and it’s formative for the world we now inhabit.

When Otlet atomized and reassembled data on cards, and Licklider imagined procognitive systems that “transform” and “test for compatibility,” they were already pointing toward what LLMs now do at scale. The primary difference lies in implementation: Where Otlet and Licklider imagined explicit procedures, today’s systems infer procedures statistically from data based on their training; but the functional role — resynthesizing prior material into new, context-sensitive configurations — remains the same.

Both ancient libraries and large language models represent the same phenomenon: intelligence emerging from the systematic arrangement of information. Far from a rupture in the history of knowledge, artificial intelligence is the predictable outcome of a process in which thought was formalized, made transferable, and ultimately rendered executable without us. We just trained the procedures to run themselves.

Joel J. Miller

Author

Full Profile

Related Content

Books

The most successful information technology in history is the one we barely notice

Joel Miller, the author of “The Idea Machine,” joins us to explore why books are history’s most successful information technology.

by Kevin Dickinson The Past

The grim truth about the “good old days”

Preindustrial life wasn’t simple or serene — it was filthy, violent, and short. The Industrial Revolution was imperfect, but it was progress.

by Chelsea Follett The Past

The hidden legal engine of progress — from railroads to AI

Common law has long balanced innovation and accountability. Can it do the same for AI?

by Dean Ball The Past

“Mirror life” and the recurring nightmare of scientific apocalypse

The fear of unleashing forces beyond control has haunted science for centuries.

by Thomas Moynihan Learn from the world's biggest thinkers.

• Videos
  • Latest
  • The Big Think Interview
• Columns
  • Mini Philosophy
  • Starts with a Bang
  • Big Think Books
  • The Long Game
  • Strange Maps
  • 13.8
  • The Well
• Sections
  • Philosophy
  • Mind & Behavior
  • Science & Tech
  • Business
  • History & Society
• Classes
  • Class Library
• Subscribe
  • Membership
  • Free Newsletters
• About
  • Our Mission
  • Work with Us
  • Contact
  • Privacy Policy
  • Terms of Sale
  • Accessibility
  • Careers

View our Twitter (X) feed View our Youtube channel View our Instagram feed View our Substack feed © Copyright 2007-2026 & BIG THINK, BIG THINK PLUS, SMARTER FASTER trademarks owned by Freethink Media, Inc. All rights reserved.