We’ve been trying to chat with machines for centuries. See what came before AI
A new book traces the history of ‘smart’ devices, back to the 14th century: early divination circles, word looms, even robots that churned out film scripts.
We tend to think of AI in isolation, as this slightly menacing entity that has taken shape from the nuts and bolts of our algorithms and experiments in machine learning. But artificial intelligence is really part of a rather large family tree of “smart” machines, dating as far back as the 14th century.
All those years ago, an Arabic astrology wheel was built in an early attempt at automated chat. Then, in 17th-century Europe, a “mathematical cabinet” was built in an attempt to automate poetry, music and coded messages.
There was an American robot in the 1930s designed to spit out story ideas for Hollywood. Spell checks, text messaging and autocorrect would follow.
It’s an intriguing journey that Dennis Yi Tenen, a scholar and software engineer, traces in his new book, Literary Theory for Robots: How Computers Learned to Write.
His aim, he says, is to take some of the fear and confusion away from how we view this technology, and offer a reminder that humans have always woven threads of language and technology together, driven by a primal urge to protect and promote intelligence and knowledge.
“Language and technology didn’t just develop together; language is technology,” says Yi Tenen, an associate professor of English and comparative literature at Columbia University. “While many animals teach things to their young, only humans are able to do it remotely, for instance, preserving and transmitting our collective know-how across time and space.”
Can we go too far in this quest? “Of course we can. I find it difficult to use technology in a balanced way in my own life,” Yi Tenen says.
With AI, of course, the struggle isn’t quite the same. It is, potentially, a struggle over the nature of reality, veracity, truth.
Which is why it is even more vital to view this technology accurately and dispassionately as a product of human endeavour, Yi Tenen says, and not as something that is a being by itself. If we view it as an extension of our collective intelligence, we can hold technology makers responsible and accountable, he adds.
His book, accordingly, takes a step back, into the historical context of AI. How far back can we trace our efforts to create “intelligent” machines? Take a look.
The prophecy wheel, 14th century
This device, created in Arabia, was a maze of concentric circles and elaborate charts that bore zodiac signs, letters, symbols and numerical values. Cords zigzagged across the whole. A manual or rulebook of sorts helped soothsayers interpret the readings, particularly in matters of astrology.
What’s interesting, though, is that some of the circles contained verses from the Quran and other holy texts, “and in manipulating the circles, one could actually kind of have a conversation with the device,” Yi Tenen says. “The fact that we’re still in Q&A mode with ChatGPT… I thought that was such a cool parallel.”
Smart cabinets, 17th century
The earliest literary bots were shaped like giant cupboards, and operated like a sort of word loom.
One of the earliest examples is the Mathematical Organ created by 17th-century polymath Athanasius Kircher. Made of painted wood, this was a box-style chest made up of rows and columns of wooden slats or rods. Attached to each were paper booklets on subjects such as arithmetic, music, geometry, chronology and astrology.
In a spooky foreshadowing, Kircher called them “applications”.
Manipulating these rods and consulting the booklets, which were arranged in a matrix, could help the user compose poetry and music (of a sort), write encrypted messages, and work out certain kinds of mathematical and astronomical calculations.
Kircher sold a version of the device to the young Archduke Charles Joseph of Austria to help with his studies.
It sparked several debates, including a public one at a bar between Kircher and the German poet Quirinus Kuhlmann. The latter argued that the path to knowledge should be “torturous, accessible only to those willing to walk it properly”.
It wasn’t the organ that was intelligent, the poet added; the intelligence lay with Kircher. “Without the box, the young duke remains an idiotic parrot.”
Essentially, an early instance of the argument that computers are making us lazy and stupid.
The Analytical Engine, 1830s
It was Kircher’s Mathematical Organ that inspired the English mathematician Charles Babbage to create what would go down in history as the world’s first computer, in the 1830s.
Babbage called it the Analytical Engine, and developed its functions in collaboration with the countess and mathematician Ada Lovelace.
It eventually looked like a weaver’s loom. The machine — only partly built by Babbage in his lifetime — had a mill (like the central processing unit), store (like memory storage), reader (input device) and printer (output device). The mill could process calculations with the help of punch cards inserted into the reader.
The Thirty-Six Dramatic Situations, 1895
By the 19th century, amid the early years of the industrial revolution, templates were emerging: for furniture, machinery, clothing, equipment, and for the arts (literature, film, music), journalism, even philosophy.
In the arts, templates have always been looked down upon as a poor substitute for missing genius.
But, Yi Tenen argues, they have always lain at the heart of how we learn. Could a machine, then, help a not-so-great writer write a great tale? Long before ChatGPT, an outline generator was created by the French writer Georges Polti, in 1895.
Titled The Thirty-Six Dramatic Situations, it listed three dozen predicaments designed to help a struggling playwright. These included supplication, deliverance, vengeance, pursuit, disaster, revolt, rivalry and adultery. Each was explained briefly with examples.
For instance, pursuit came with the instructions that the plot should be “held by the fugitive alone; sometimes innocent, always excusable, for the fault — if there was one — appears to be inevitable, ordained; we do not inquire into it or blame it, which would be idle, but sympathetically suffer the consequences with our hero, who whatever he may once have been, is now but a fellow-man in danger”.
Combining elements from his set of dramatic situations, hundreds of thousands of stories could take shape. Though rarely acknowledged, such templates were frequently used by writers well into the age of cinema, Yi Tenen says.
The Plot Robot, 1931
This machine, built by Los Angeles screenwriter Wycliffe Hill, could produce a complete outline of a story in 20 minutes, its whirring gears drawing background, characters and dramatic situations from a series of tapes housed in its chassis.
The robot found no takers (it is unclear why, but expense may have been a factor; it is also possible that the stories weren’t very good). But it prompted Hill to write a book titled Plot Genie (1935), about a wheel that could be spun to point to seemingly random numbers, which could be referenced against a set of charts, to generate plot requisites (locale, character, the beloved, the problem, and so on).
As it turns out, Hill wasn’t a polymath or early technocrat. He was a scriptwriter who became fixated on decoding the intricacies of what makes a good plot, when his own screenplay was rejected. His aim was to take the effort and uncertainty out of writing.
There hasn’t yet been a magic wand created that can do that.
Recommended Topics 
