The Original Large Language Machine
Accept no substitutes. Catalyze those changes.
Let's talk about a new technology that generates written text at previously unimaginable scale, triggering demands that it be censored, lest it promulgate a flood of misinformation that sows dissent and changes the rightful order of the world...
…technology that we, however, in our sober wisdom, recognize as a profoundly transformative breakthrough. Yes, that’s right: I’m talking about
OK, fine, I’m half joking — but there really are some striking parallels to explore!
The Tale of Movable Type
The Gutenberg Press (henceforth “The O.G. LLM,” because you can’t stop me) was not the first instance of movable type. That was invented in Asia, centuries earlier, but didn’t take off. Gutenberg, though, had a much more efficient alphabet (Korea’s efficient Hangul alphabet postdates him), and as a former metalsmith, he made two engineering breakthroughs: a new alloy so suited to movable type it is still in use today, and a new method of molding those types.
His resulting press was 90x more productive than the previous state of the art, able to crank out 3600 pages/day vs. about 40. 90x! In an era not exactly accustomed to even 10x gains from technology, to put it mildly! The results were, accordingly, dramatic:
Gutenberg's first major work was the 42-line Bible in Latin, printed 1452-54 … In rapid succession, printing presses were set up in Central and Western Europe (Cologne 1466, Rome 1467, Venice 1469, Paris 1470, Buda 1473, Kraków 1473, London 1477) … by one estimate, "by 1500, 1000 printing presses were in operation throughout Western Europe and had produced 8 million books"
The printing press ushered in a new era of human history, in which we still live; the era of mass communication. This was amazing, epochal, transformative! It also, as Douglas Adams might put it, “made a lot of people very angry and has been widely regarded as a bad move.”
The world was already changing, of course; but the press was a hell of an accelerant. Constantinople fell in 1453, erasing the last relict of the Roman Empire just as Gutenberg was printing his seminal work. Meanwhile, in Italy, the creative flywheel called the Renaissance spun up to its world-changing torque, with the printing press as part of its fuel.
This did not go without a backlash. In the Papal Conclave of 14921, the infamous Rodrigo Borgia, played centuries later by Jeremy Irons—
—was elected Pope Alexander VI. In 1501, having apparently had enough of the O.G. LLM, Borgia/Alexander published the censorious decree entitled Inter Multiplices, which stated that henceforth
ecclesiastical [authorization] was necessary before print publication would be allowed … Archbishops were to prohibit the printing of books in their provinces without this … the censorial powers could be delegated to local authorities … The jurisdiction extended over corporations, universities and colleges … to motivate the local authorities, they were to receive half of the monetary penalties collected.
The decree’s preambular quote may sound familiar :
The art of printing can be of great service in so far as it furthers the circulation of useful and tested books; but it can bring about serious evils if permitted to widen the influence of pernicious works. It will, therefore, be necessary to maintain full control over the printers so that they may be prevented from bringing into print writings which are antagonistic to the Catholic faith, or which are likely to cause trouble to believers.
A ritual nod to how the new technology could be good, if used in acceptable ways, then a stern order that it be carefully controlled lest it spread bad ideas and overturn the order of things. The narrative hasn’t changed much in five hundred years!
I’m not saying that Borgia/Alexander was wrong. He wasn’t. The rise of printing was, despite his attempts, catastrophic for the Church:
The printing press became the single most important factor in the success of the Protestant Reformation … Luther's 95 Theses, which would have circulated only among the literate scholars of Wittenberg, became a bestselling pamphlet within a year of its initial posting in 1517. Between that date and c. 1525, Luther would publish over half a million works, establishing him as the first bestselling author
The Reformation, of course, was “the 16th-century religious, political, intellectual and cultural upheaval that splintered Catholic Europe.” What’s more:
The clash between Luther and the Catholic Church in Rome was also history's first "media event." Johannes Gutenberg's development of the moveable-type printing press about seventy-five years earlier had a profound impact on the spread of Luther's thought. Thanks to the new technology, his theses soon reached a broad circulation that surprised even Luther.
This despite the fact that the industry in question was a complete mess. “Luther complained … of the reckless and scandalous character of many reprints of his books, so full of blunders he could hardly recognize them.” (One might say ‘hallucinations’…)
The chaotic, eruptive growth of the printing industry may, however, also be why the Church’s reaffirmed attempts at pre-emptive censorship ultimately failed: “Its decree seems to have been poorly implemented until fear of civil unrest coming from the spread of the Reformation … led Catholic rulers to enforce it.” By then it was too late. Mass communication met corrupt hierarchy, and mass communication (basically) won.
To be clear, the printing press did not cause these changes; but it catalyzed them. Similarly, the rise of daily newspapers did not cause the French Revolution(s), but they catalyzed it/them. (“The” Revolution was really several revolutions over several years.) It’s no coincidence that many of the major figures — Marat, Robespierre, Desmoulins, Hébert — published their own newspapers, and in doing so became political titans.
From GPT to GPT
You may well be thinking: “OK, this is an amusing/interesting historical aside, but your two categories of LLMs have nothing meaningful in common!” But I think they do. I think that just as the O.G. LLM ushered in a previously all-but-unimaginable era of mass communication … so too will today’s LLMs, in an even weirder way.
Throughout history, when we’ve said “communication,” we have meant, with very few exceptions, “communication with other humans.” But when a kid logs onto chat.openai.com and has a conversation — no, GPT-4 is not intelligent, but it is an independent interlocutor. Its conversation is restricted by RLHF, but it is not programmed, predictable, or deterministic. (And even RLHF is ultimately another model, not a human-built constraint!)
As such, we may be — I think we are — entering a new era of mass communication. We may not see the effects for a generation, until the kids who grew up talking to LLMs achieve adulthood; or, we may see AI-generated comms infest the entire Internet like kudzu within only a few years. But I would be surprised if the effects were ultimately anything less than very significant.
Is this a good thing? That’s kind of the wrong question; all change has both costs and benefits, and with rare exceptions, the more significant the change, the more of both. Is it a good thing for the status quo, the way of the world, the modern order of things? Almost certainly not. As we’ve seen, new eras of mass communication tend to catalyze pretty dramatic upheavals. Is it a net good thing in the long run, for humanity? …I think yes, but I also think the road there is going to be a sometimes bumpy one.
A strange and wonderful highlight of my 2022 was the opportunity to participate in a semi-professional re-enactment of the 1492 Papal Conclave. If there’s a better way to teach history than to literally immerse people in its creation, I sure don’t know what.