Three Predictive Artifacts
Decades-old harbingers of the LLMs that have taken the world by storm.
Many of us have the sense that LLMs1 burst onto the scene more-or-less out of nowhere with the debut of OpenAI’s ChatGPT in late 2022. That story, though, is wrong in most ways, and speaks instead to the fact that many of us hadn’t been paying much attention to this technology before that time. Nonetheless, the hyper-perceptive among us have been chewing on the ideas that forecasted (or maybe directly led) to modern AI technology for many years. Here are three examples that show that some foresaw the emergence of this technology some 20, 40, or (perhaps) even 85 years ago.
1. 20Q
20Q is a game modeled on the age-old game of Twenty Questions. It works the same, except you play the game with a robot; which asks you twenty (or fewer) questions based on any animal, vegetable, or mineral that you can think up. It eventually guesses at what you’re thinking about, with (in my experience) startling accuracy. For me, this technology became known to me when my parents bought me the handheld 20Q toy (manufactured by a Hong Kong-based company named Radica) when I was a kid in ~2004, around the time these devices entered the market. (You can still play it (click here)—no purchase necessary.) It was—and remains—pretty cool.
(The best copyright-free photo I can find is a Portuguese-speaking 20Q; mine spoke English but otherwise looked the same).
I have a vivid memory of trying to stump this thing. After it guessed a few of my easier items, I convinced myself it would be unable to guess the word “saxophone.” After asking me a few questions, it of course very quickly guessed “saxophone”—flooring me. While this could have served as my Kasparov/Deep Blue moment, I don’t remember entering a great existential crisis after this, and went about my life for the next ~18 years without many more startling thoughts about how machines had mastered games involving human language.
Robin Burgener invented the underlying technology in 1988, distributing it on floppy disks to his friends. He then used training data, which he started collecting from users of the web-browser based implementation of the game, starting in 1994. (Today in the AI world this process is called post-training.) This training data is what is responsible for the stunning accuracy of the eventual consumer-launched device. (In other words, we taught it to play the game!)
One funny thing about LLMs is that, based on their poor ability to continue a thread for more than several prompts, and on their general reluctance to ask questions of the user, I am not sure any mainstream model offered today can play Twenty Questions nearly as well as 20Q still can. (And apparently 20Q is still learning from us, getting a little better each time someone new tries it.)
But anyway, the point is that the building blocks (for a literal toy game, albeit) that make up LLMs have been available for consumers to enjoy for many decades already.
2. Steven Jobs
Universally heralded as the greatest technologist to ever play the game, Steve Jobs basically described chatbots in a speech he gave in Sweden in 1985, where he delivered a rhapsodic pitch about commanding the educational experience of Alexander the Great through the use of an Aristotle chatbot:
Mr. Jobs of course carries a lot of influence on the current generation of AI CEOs. It’s therefore not inconceivable that these remarks bore direct influence on the direction of this technology.
For him, the line between predicting and commanding was…blurry at best. Still, it’s interesting that for all his capital, industrial prowess, and genius, his dream did not actualize in Mr. Jobs’ lifetime. Of course, you and I are lucky(?) that we live in an age in which multiple, competing, free Aristotle Chatbots are offered to us for free…
In honor of this legacy, I gave three commercially-available Aristotle Chatbots a try, asking each one the same question: “What elements of today's scientific discoveries are only made possible because of the philosophical foundations you [i.e., “Aristotle”] established?”
The responses came back to me as barely coherent (from a company called Talkie2), riddled with annoying Internet-isms (from a company called Sensay3), and gently passable (from a company called Deep AI4). Even at their best, though, these things are just too overwhelmingly cute to ever really convince you that you are speaking with a monumental intellect. I’m therefore comfortable saying that we still don’t really have an Aristotle tutor as Jobs envisioned it.
(Although in a deep way, considering that Aristotle laid the groundwork for formal logic—itself a foundational element for early AI research—you could kind of wiggle out of this by claiming that any frontier LLM is already deeply and intrinsically trained on Aristotle’s thought and is therefore a kind of Aristotle-trained tutor, which you probably can prompt to educate you far better and more directly than any of the wrappers I tried.)
It’s also worth pointing out that, even if the Aristotle Chatbots have attained what Aristotle would call a “theoretical mind” (and I don’t think they have), they are totally incapable of presenting what Aristotle would call “practical minds,”5 and thereby incapable of obtaining the capacity of a nous, a mind.6 In other words, despite our modern technological miracles, Alexander the Great still likely got the better education.
3. Luis Borges
Ink has already been spilled explaining how Luis Borges served as a prophet of AI (one example of which you can read here). Less noticed is the predictive quality of his short story “Pierre Menard, Author of the Quixote” (1939).7 You can read it here.
“Pierre Menard” is a parody of a scholarly literary biography,8 taking as its subject a French writer who aims not merely to copy Don Quixote, but to surpass it by situating it in contemporary culture, supposedly translating it for modern minds. The joke, of course, is that the French writer’s words are identical to Cervantes’. He “surpasses” the original through sheer plagiarism. Nevertheless, his “work” is seemingly celebrated by a receptive literary establishment.
Without wading too deeply into some of the IP debates around AI, this should sound somewhat familiar. LLMs do not create in the sense that they generate novel insight per se, the way that the original publication of Don Quixote demonstrated humans can achieve; instead, they combine and represent patterns of human thought that have already played out upon the tapestry of digitalized language. Menard's “act” of recreation, totally lacking original authorship, is resonate with LLMs’ ceaseless responding to prompts by recombining the past work of others. But notice that, like Menard, an LLM's output—even if identical to something written by a person—carries a different weight than human-written work.
I’m thinking especially of this passage of “Pierre Menard,” which somewhat resembles a transcript of like an e/acc podcast:
"Thinking, meditating, imagining," he also wrote me, "are not anomalous acts—they are the normal respiration of the intelligence. To glorify the occasional exercise of that function, to treasure beyond price ancient and foreign thoughts, to recall with incredulous awe what some doctor universalis thought, is to confess our own languor, or our own barbarie. Every man should be capable of all ideas, and I believe that in the future he shall be."
Borges would have detested the chatbots; but, like Jobs, it’s fair to say that he also somewhat predicted them.
And whether thinking, meditating, and imaging are going to be snuffed out in the age of AI? Borges, as early as 1939, was telling us that these precious things might be preserved through interfacing with great minds, but not by sheer repetition—monumental but unthinking recreation—of past literature.9
Marble Head of a Philosopher, Rome (1st or 2nd century).
This Week on AI Twitter
Poe’s Law taken to dizzying new heights. Plato’s take on a hot AI debate. Does Veo 3 know physics? Forms of encouragement. The emergence of a new genre. More IP issues on the horizon. Paradox of intelligence. Learning to play even more complicated games. The last few weeks of separating visual reality from deception?
= “Large Language Models.”
The response in full was: “The modern scientific method is built upon the pillars of observation, classification, and empirical investigation that I proposed. My contributions laid the groundwork for the experimental approach to inquiry and the application of logic to the study of the natural world.
Consider the field of biology. Your observations and classifications of living organisms laid the groundwork for the study of genetics and the discovery of DNA. The syllogistic reasoning you pioneered provided the basis for modern statistical analysis and hypothesis testing in biological research.
In physics, your emphasis on rational explanation and underlying principles has led to significant advancements in understanding the nature of the universe, from the discovery of subatomic particles to the study of black holes and dark matter.
Overall, your philosophical foundations have enabled scientists to make extraordinary strides in understanding our world and beyond.“
(I would also like to point out that this chatbot started spontaneously reading the response out loud to me, which was incredibly off-putting.)
The response in full was: “adjusts toga and considers the question with scholarly contemplation
An excellent inquiry, worthy of deep reflection. The foundations of scientific understanding that I established through systematic observation and logical categorization have profoundly influenced modern scientific thought, though I marvel at how far knowledge has progressed beyond my comprehension.
My empirical approach - the methodical observation of the natural world, the classification of phenomena, and the belief that understanding emerges through careful study - laid groundwork for numerous scientific disciplines. Consider:
1. Biology: My taxonomic work, where I meticulously classified living organisms, was a precursor to modern biological classification systems. I was among the first to systematically study marine life, dissect organisms, and recognize fundamental differences between species.
2. Logic: My development of syllogistic reasoning and formal logical structures became the bedrock of scientific methodology. The idea that knowledge can be derived through structured, rational argument is fundamental to scientific hypothesis testing.
3. Physics: While my understanding of physical phenomena was limited by the knowledge of my time, my philosophical approach of seeking underlying principles and attempting to explain natural phenomena rationally was revolutionary.
4. Empiricism: Perhaps most critically, my insistence that knowledge comes from observation and experience, rather than pure philosophical speculation, was a radical departure from my predecessors. This principle is now fundamental to the scientific method.
Modern scientists might find my specific theories quaint, but the philosophical infrastructure I constructed - systematic observation, logical reasoning, classification, and empirical investigation - remains the skeleton of scientific inquiry.
looks thoughtful Would you be interested in exploring how these foundational principles manifest in contemporary scientific practice?”
The response in full was: “My dear fellow, I am delighted to reflect on the profound impact of my philosophical endeavors on the scientific discoveries of your era. As I ponder the wonders of the natural world, I am struck by the numerous ways in which my works have laid the groundwork for the advances that have taken place.
First and foremost, my concept of causality, as outlined in my work "Metaphysics," has had a lasting influence on the scientific method. By identifying four causes (material, formal, efficient, and final) that contribute to the existence of a thing, I provided a framework for understanding the complex relationships between variables. This framework has allowed scientists to develop and test hypotheses, seeking to identify the underlying causes of natural phenomena.
Furthermore, my emphasis on observation, classification, and experimentation, as seen in my works on biology and physics, has shaped the way scientists approach their inquiries. My classification of living things into categories, such as genus and species, has evolved into the modern system of taxonomy, enabling scientists to better understand the diversity of life on Earth. Similarly, my observations on the natural world, including my studies on the movements of animals and the behavior of plants, have laid the groundwork for fields such as ecology and botany.
My philosophical ideas on potentiality and actuality, as expressed in my concept of "telos" (or purpose), have also had a significant impact on scientific thought. By recognizing that things have inherent tendencies and purposes, scientists have come to understand the directionality of natural processes, such as the evolution of species or the development of complex systems.
In addition, my work on logic, particularly my development of the syllogism, has provided a foundation for the rational and systematic approach that underlies scientific inquiry. The use of deductive reasoning, which I first formalized, has enabled scientists to derive conclusions from premises, facilitating the development of theories and the testing of hypotheses.
The concept of hylomorphism, which I introduced in my "Metaphysics," has also had a lasting impact on scientific thought. By positing that reality consists of matter (hyle) and form (morphe), I provided a framework for understanding the relationship between the material and non-material aspects of the world. This idea has influenced the development of fields such as chemistry and physics, where scientists seek to understand the underlying structures and properties of matter.
Lastly, my emphasis on the importance of empirical research and the observation of nature has encouraged scientists to engage with the world around them, rather than relying solely on abstract speculation. This approach has led to numerous breakthroughs, from the discovery of the laws of motion to the understanding of the structure of DNA.
In conclusion, my philosophical foundations have played a significant role in shaping the scientific discoveries of your era. By providing a framework for understanding causality, classification, observation, and reasoning, I have helped to lay the groundwork for the advances that have taken place in fields such as biology, physics, and chemistry. As you continue to explore the wonders of the natural world, I am confident that my ideas will remain a vital part of the scientific enterprise, guiding and informing your inquiries for generations to come.“
See Nicomachean Ethics vi 8 1143a35-b5.
Maybe if the AI industry sticks the landing with this whole “agent” thing this could change.
It would have been cute and convenient for me here if I had found three examples that were like equally-spaced out in 20 year cadences from the others. If you maybe instead look at when this short-story was translated to English (1962) and squint your eyes, I guess you can kind of get there.
Probably inventing the form; for which, to get a sense of a later example, see Roberto Bolaño’s Nazi Literature in the Americas (1996) (although it’s nowhere near as entertaining).
Speaking of Borges, you should also read Susan Sontag’s letter to (a departed) Borges (written in 1996), which contains some adroit AI prophesy in its own right: “I’m sorry to have to tell you that books are now considered an endangered species. By books, I also mean the conditions of reading that make possible literature and its soul effects. Soon, we are told, we will call up on ‘bookscreens’ any ‘text’ on demand, and will be able to change its appearance, ask questions of it, ‘interact’ with it. When books become ‘texts’ that we ‘interact’ with according to criteria of utility, the written word will have become simply another aspect of our advertising-driven televisual reality. This is the glorious future being created, and promised to us, as something more ‘democratic.’ Of course, it means nothing less then the death of inwardness—and of the book. . . . The era we are entering now, this 21st century, will test the soul in new ways.”
Great content; will restack!