These automatic dolls can play music and talk, are they the predecessors of artificial intelligence?

Leviathan Press:

The word automaton in English means automatic machine or automatic puppet. The word comes from the ancient Greek word αὐτόματον (autómaton), which means "to act with one's own will". But in fact, strictly speaking, android is more suitable for the automatic machinery that appears in this article. In the application of automaton, there are many devices similar to music boxes, which are quite different from the machinery discussed in this article. Although in essence, they are both reproductions of human/humanoid mechanisms, the complexity of android is far higher than that of music boxes.

The word "android" is derived from the Greek word for "human-like" and was coined by French physician and librarian Gabriel Naudé, personal physician to Louis XIII and later designer of Bishop Jules Mazarin's 40,000-volume library. Naudé was a rationalist who opposed superstition. In 1625, he published a defense of the scholastic philosophers. It mentioned the 13th-century theologian and philosopher Albertus Magnus (Albert the Great), who was said to have made a robot out of bronze.

The story seems to have been created long after Albert the Great’s death by the prolific 15th-century critic Alfonso de Madrigal, also known as El Tostado, who adapted and embellished medieval tales of moving statues and talking bronze heads.

El Tostado says Big Albert spent 30 years creating an entire man out of metal. The automaton provided Big Albert with answers to all of his toughest questions, and in some versions of the story even cheerfully dictated much of his writings.

According to El Tostado, the machine met its final fate at the hands of Albert the Great’s student, Thomas Aquinas, who, unable to bear “its incessant chatter,” smashed it to pieces.

Noddy doesn't believe in Big Albert's so-called talking robot.

He dismissed it, along with other stories of talking heads, as "false, absurd, and erroneous." Nordy noted that these devices lacked "muscles, lungs, epiglottis, and all the necessary organs for a perfect production of speech." Nordy concluded that, based on all reports, Albert the Great probably did build a robot, but was never able to provide a clear and unambiguous answer to his question.

Albert’s machine might be more similar to the Egyptian Colossus of Memnon, a statue widely discussed by ancient writers: when sunlight hit it, it emitted a pleasant murmur, as the heat caused the air inside the statue to “rarefy” and be expelled through small pipes, making a whisper-like sound.

Illustration of the Colossi of Memnon from an 1800 account of Egypt. © wikipedia

Despite his disbelief in Great Albert’s talking head, Nordi gave it a powerful new name: “android”. In this way, he cleverly introduced a new term into the language, for according to the 1695 dictionary of the French philosopher and writer Pierre Bayle, “android” was “a completely unknown word, purely an invention of Nordi, who boldly uses it as if it were already established”. It was a favorable time for a newborn word: Nordi’s term quickly permeated new dictionaries and encyclopedias. Bayle repeated it in his dictionary entry for “Great Albert”.

Thus, "android" was immortalized as an entry in the first volume of the supplement to the Encyclopedia by Ephraim Chambers, the compiler of the British Encyclopedia. While denying the existence of Albert the Great's "android", Noddy gave "android" new life as a category of machine.

"Albert Magnus' Talking Head" from JH Pepper's Encyclopedia of Science Simplified (1885). © archive.org

The first actual robot in the historical record that informed the new experimental philosophy—an “android” in the etymological sense of Noddy, a working humanoid made up of “necessary parts”—was exhibited on February 3, 1738, at the annual Exposition Saint-Germain on the Left Bank in Paris.

This android differs from earlier musical automata, hydraulic organs, and figures on musical clocks because it actually performs the complex task it appears to perform, in this case playing the flute, rather than simply making suggestive movements.

The Shepherd with the Flute by Antoine Kosevoc. © wikipedia

The device was a novelty, but it must have been familiar to many fairgoers, as it imitated the look of a famous statue that stood at the entrance to the Tuileries Gardens and is now in the Louvre Museum: Shepherd Playing the Flute by Antoine Coysevox.

Like the statue, the android is presented as a faun - a mechanical faun holding a flute. The mechanical faun suddenly comes to life and begins to play its instrument, playing 12 pieces in a row. At first, the audience suspects that this must be a music box, with an automatic mechanism inside producing the sound, and the external figure just pretending to play.

Jacques Vaucanson's automata: the flute player, the automatic defecating duck, and the tambourine player. © Wikimedia Commons

But that's not the case; the android is actually playing a real flute, blowing air from its "lungs" (three sets of bellows) and using its lips, soft tongue, and leather-covered fingers. There are even reports of audience members bringing in their own flutes, and the machine can play them, too.

The flute-playing android is the work of a young engineer named Jacques Vaucanson. The youngest of ten children of a Grenoble glove maker, he was born in the cold winter of 1709, at the end of the long reign of Louis XIV, in the bloodiest year of a terrible famine and French defeat. Vaucanson emerged from this dark hour, his life synchronised with the Enlightenment, and his work would become an inspiration for literature.

As a child, he loved making clocks and repairing watches. During school, he began designing automata. During a brief monastic life in Lyon, a church official ordered the destruction of Vaucanson's workshop, and at the age of 19, he came to Paris to seek opportunities. At first, he wanted to become a doctor and took some courses in anatomy and medicine, but soon decided to apply these studies to a new field: reproducing life processes in machines.

The Flutist was the result of five years of work by Vaucanson. When it was completed, Vaucanson submitted a memorandum explaining its principles to the Paris Academy of Sciences. This memorandum contained the first known experimental and theoretical study of the acoustics of the flute.

"Vaucanson", from The Great Ancient and Modern Inventors, 1864. © gallica.bnf.fr

After an eight-day initial exhibition at the Saint-Germain Exposition, Vaucanson moved his robot to the Hôtel de Longueville, a 16th-century mansion in the city center. In the magnificent hall, about 75 people visited every day, each paying an expensive entrance fee of 3 livres (roughly equivalent to the average weekly wage of a Parisian worker). Members of the Paris Academy of Sciences also went to the Hôtel de Longueville to see the flute-playing android robot. Vaucanson received 10 to 15 people at a time, explained the principles of the flute-playing robot to the public, and then let it start playing music.

Critical reviews were positive. One critic wrote that "all Paris came to admire... perhaps the most singular and delightful mechanical phenomenon ever created", noting that the android "actually played the flute". Another critic agreed, calling it "the most marvelous mechanical work ever invented". Journalist and popular author Pierre Desfontaines, in his literary journal, described the flute-playing robot as containing "innumerable wires and steel chains... which, by simulating the expansion and contraction of the muscles, produce the movements of the fingers, just as in a living person. No doubt this knowledge of human anatomy... guided the author in his mechanical design".

Vaucanson's work became the model for the "android" entry in the Encylopédie, a popular work of knowledge edited by philosopher and writer Denis Diderot and mathematician and philosopher Jean d'Alembert. The entry, written by d'Alembert, defined an android as a humanoid that performs human functions.

Shortly after the members of the Paris Academy of Sciences visited the Hôtel de Langueville, Vaucanson read out a memorandum on the design and function of the flute-playing android. The android is powered by a weight driven by two sets of gears. The bottom gear set turns a shaft with a crank that pushes three sets of bellows connected to three air tubes to provide three different blowing pressures for the flute-playing android's lungs. The upper gear set turns a cylinder with a cam that triggers a lever frame that controls the flute-playing android's fingers, air tube, tongue, and lips.

To design a machine that could play the flute, Vaucanson studied human flute players in detail. He devised ways to translate aspects of their playing into the design of his android. For example, to mark the bars, he had one flutist play the tune while another person tapped the rotating cylinder with a sharp stylus.

That winter, Vaucanson added two new machines to the exhibit. One was a second musical automaton, a life-size Provencal shepherd who played 20 minuets on a flute held in his left hand while beating a drum hung from his shoulder with his right hand. The flute has only three holes, which means that notes are produced almost entirely through variations in blowing pressure and tongue. In striving to recreate these subtleties, Vaucanson discovered that human flute players used a much wider range of blowing pressures than they realized.

The flute player made another surprising discovery. Vaucanson had thought that each note was produced by a specific combination of finger position and blowing pressure, but he discovered that the blowing pressure required for each given note depended on the previous note. For example, a D after an E required more pressure than a D after a C, so he had to prepare twice as much blowing pressure as there were notes. The high overtones of the high notes resonated in the flute more strongly than the low overtones of the low notes; however, the flute player himself was not aware of compensating for this effect. The physics of this overtone was not explained until the 1860s by Hermann von Helmholtz.

These automata didn’t just make music—the music box had done that two centuries earlier—but the androids could play music with their flexible lips, tongues, fingers, and lungs. They simulated how humans play music, and as the centuries went by, designers of these automata moved on to the more complex task of making machines that could mimic human speech.

Vaucanson's Automatic Duck: This automatic duck can flap its wings and hop, but the most remarkable thing about it is that it can be made to swallow some grain and watch the digestion and excretion process. Scientific American, January 21, 1899. © Linda Hall Library

In 1739, the year after Vaucanson’s “automatic duck” debuted, a surgeon named Claude-Nicolas le Cat published a now-lost article describing an “automaton in which the principal functions of animals can be seen to be performed: circulation, respiration and ‘secretion.’” It’s unclear what happened to this early project, but le Cat returned to the idea in 1744, where he read a memorandum to the astonished audience, according to the minutes of a meeting of the Rouen Academy. “M. Le Cat has told us about his project for an automaton… His automaton will have respiration, circulation, digestion, secretion, chyle, heart, lungs, liver and bladder, with all the functions that go with it,” the audience said.

Leca’s automaton would be capable of “all the operations of a living being,” including not only the circulation of blood, the beating of the heart, the working of the lungs, the swallowing of food, digestion, excretion, the filling of the blood vessels and the exhaustion of blood, but also—clearly straddling the Cartesian boundary between mechanical body and rational soul—“even the pronunciation of speech.”

Don Quixote examining the Talking Head. From an engraving by Martin Engelbrecht, 1662. © The British Library

This idea, the possibility of simulating articulate speech, had sparked a flurry of philosophical discussion in the previous century.

While some still consider this a quixotic fantasy, when Don Quixote himself encounters a talking bronze head (actually connected to a hidden person), he is completely enchanted, while his less susceptible squire Sancho Panza is uninterested. Cervantes’ contemporary, Spanish writer Martín del Río, agreed that “it is not reasonable for an inanimate object to make human sounds and answer questions. This requires life, breathing, the concerted action of complete vital organs, and a certain logical ability on the part of the speaker.”

Decades later, del Rio’s artificial machine seemed achievable. Athanasius Kircher wrote in 1673 about the legends of Albert the Great’s talking head and the ancient Egyptian Colossi of Memnon. While some skeptics dismissed these devices as “either nonexistent, or fraudulent, or created with the help of the devil,” many believed it would be possible to create a statue with a throat, tongue, and other vocal organs that could produce articulate sounds when activated by wind.

Kircher included a sketch of a design for a talking puppet. His student Gaspar Schott, also a prolific natural philosopher and engineer, even mentioned a statue that Kircher built for the then Queen Christina of Sweden that answered questions. No doubt her former philosophy teacher, Descartes, had sparked her interest in the relationship between rational speech and mechanical bodies.

Although the idea of ​​simulating speech was not new, it saw a renewed interest in it among experimental philosophers and mechanical engineers in the mid-18th century. Speech, they argued, was a bodily function akin to breathing or digestion—they did not make a clear distinction between the mind and the physiological mechanisms of speech—and even skeptics expressed their doubts in physiological details rather than in principle. In his enthusiastic 1738 review of the Piper of Vaucanson, for example, de Fontaine predicted that artificial machines would never be able to produce intelligible speech because the physical mechanisms of speech would always remain incomprehensible: one could never know for sure “what goes on in the larynx … [and] the movements of the tongue, all the variations of the jaw and lips.” Speech, de Fontaine argued, was an essentially organic process that could only occur in a living larynx.

De Fontaine was not alone in his view. At the time, skeptics of the possibility of artificial speech generally believed that the human larynx, vocal tract, and mouth were too soft and flexible to be simulated mechanically. Around 1700, Denys Dodart, personal physician to Louis XIV, submitted several memoranda on the human voice to the Paris Academy of Sciences, arguing that the sound was caused by the contraction of the glottis, which was "incapable of being imitated by art." The writer and scholar Bernard le Bovier de Fontenelle, then permanent secretary of the Academy, commented that no wind instrument produced sound through such a mechanism (the variation of a single opening), which seemed "entirely beyond the realm of imitation... Nature has access to materials that are utterly unavailable to us, and she knows how to use them in ways that are utterly unknown to us."

The last to hold the “material difficulty theory” was the philosopher and writer Antoine Court de Gébelin, who observed that “the vibrations that spread to all the parts of the glottis, the tremors of the muscles, their impact with the hyoid bone, which moves up and down, the reverberations of the air on the sides of the mouth…these phenomena” could only occur in a living body.

There are many who disagree: for example, the polemical materialist Julien Offray de La Mettrie, after reading the Flute Player of Vaucanson, argued that a talking machine “can no longer be considered impossible.”

"Organs of sound", illustration from Antoine Cour de Gerbelin's Primitive World (c. 1773-1782). © gallica.bnf.fr

In the last 30 years of the 17th century, several people began working on artificial speech projects. They all believed that the sounds of spoken language required a structure that was as similar as possible to the throat and mouth. This assumption that a speaking machine would need to simulate the vocal organs did not always dominate thinking about artificial speech.

In 1648, John Wilkins, the first secretary of the Royal Society, described plans for a talking statue that would synthesize speech, rather than simulate it, by using “unintelligible sounds”: “We may observe that the waving of water is like the letter L, the quenching of hot bodies like the letter Z, the sound of strings like the letter Ng (sic), the lash of a whip like the letter Q, and so on.”

But in the 1770s and 1780s, builders of talking machines mostly believed that it was impossible to create artificial speech without building a talking head: to replicate the vocal organs and simulate the process of speaking.

The first to attempt to build such a machine was the English poet and naturalist Erasmus Darwin (grandfather of Charles Darwin), who reported in 1771 that he had "contrived a wooden mouth, with lips of soft leather, and a valve in the back for nostrils." Darwin's talking head used "a ribbon ... stretched between two slightly concave pieces of smooth wood" as a throat. It said "mama, papa, map and pam" in "a very plaintive tone."

Two talking heads of Father Mikael, by EA Tilly. © wikimedia

The next to simulate speech was the Frenchman Abbé Mical, who presented a pair of talking heads to the Paris Academy of Sciences in 1778. The heads contained "several artificial glottises, arranged in different forms on a taut membrane." Through these glottises, the two heads engaged in a conversation praising Louis XVI. One head said, "The King has brought peace to Europe," and the other replied, "Peace brings glory to the King," the first head added, "Peace brings happiness to the people," and the second head concluded, "Oh, King, you are the loving father of your people, and their happiness shows Europe the glory of your throne."

The Parisian gossip writer Louis Petit de Bachaumont noted that the heads were life-size but had a gaudy gilt look to them, that they mumbled some words, swallowed some words, and had a hoarse voice and spoke slowly.

Nevertheless, they undeniably possessed "the power of speaking." The scholars appointed to examine Micard's talking heads agreed that their articulation was "very imperfect," but still approved of the work because it imitated nature and contained "the same structures which we see in the dissection of ... the organs of voice." Bachaumont records that the scholars were so impressed with Father Micard that six representatives of the Academy of Sciences invited Micard to accompany them during the Montgolfier hot air balloon demonstration at Versailles on September 19, 1783, when a sheep, a rooster, and a duck became the world's first air passengers, and to present the maker of the famous talking heads to the king.

The following year, probably at the suggestion of mathematician Leonhard Euler, the St. Petersburg Academy of Sciences sponsored a prize competition to determine the nature of vowels and construct an instrument similar to the human voice organ pipes to express them. Academy member CG Kratzenstein won the prize. He used an artificial glottis (reed) and organ pipes that changed according to the position of the tongue, lips and mouth during pronunciation.

Wolfgang von Kempelen's design for the components of a talking machine, 1791. The bellows act as lungs, pumping air into a vocal organ equipped with vibrating reeds, whose sound is controlled by opening and closing valves. Not shown is the rubber "mouth" attachment, which is connected to the "o" via a rim with nostrils. © digital.slub-dresden.de

Several more people created talking heads before the turn of the century, including a Hungarian engineer named Wolfgang von Kempelen, who was hired by Empress Maria Theresa to work at the Holy Roman Court when he was 21.

Turkish chess puppet: This machine appears to be able to play a fierce chess game with a human opponent, but in fact it is just a complex simulation of mechanical automation: A human chess master hidden in a cabinet controls the Turkish robot from below via a series of levers. © digital.slub-dresden.de

Decades after he became famous in 1769 for creating a chess-playing Turkish puppet that concealed a skilled human chess player, Kempelen began exploring the secrets of clear speech.

Diagram depicting the components of artificial and natural speech from Wolfgang von Kempelen's The Mechanism of Speech (1791). © digital.slub-dresden.de

In 1791, he published "A Summary of a Talking Machine," reporting that he had connected bellows and resonators to instruments that resembled the human voice, such as oboes and clarinets; he also tried to transform the human voice into organ pipes, just as Kratzenstein had done. During his 20 years of experimentation, he remained convinced that "speech can be imitated."

The final device had bellows as lungs, an ivory glottis, a leather vocal tract with a hinged tongue, and a mouth with a rubber mouthpiece whose resonance could be changed by opening and closing valves, and two small tubes as nostrils. Two levers on the device connected the whistle, and a third lever connected to a string that could be placed on the reed. These made it possible to produce liquid and friction sounds: Ss, Zs and Rs.

The machine was reminiscent of Vaucanson’s findings (the pressure of the air blow for a given note depended on the previous note). Kempelen reported that he first tried to pronounce each sound in a given word or phrase independently, but failed because successive sounds needed to adjust to each other in order to be pronounced correctly: “The articulation of speech is revealed only in their proportional relations to one another and in their connection with whole words and phrases.” Listening to his machine’s fuzzy speech, Kempelen realized another limit to the mechanization of language: dependence on context.

Kempelen's machine was only moderately successful. It reportedly chirped vowels and consonants in a childish voice, uttered words such as "mama" and "dad," and slurred phrases such as "you are my friend—I love you with all my heart," "my wife is my friend," and "come with me to Paris." Today, the machine is housed in the Deutsches Museum in Munich, Germany.

Kempelen and his supporters stressed that the device was imperfect and was not a true talking machine in itself, but rather a device that demonstrated the possibility of creating a talking machine.

After this boom period in the 1770s, 1780s, and 1790s, interest in speech simulation declined.

A few in the 19th century, including the inventors Charles Wheatstone and Alexander Graham Bell, built their own versions of talking machines, but for the most part, designers of artificial speech once again turned their attention to speech synthesis rather than simulation: reproducing the sounds of human speech by other means rather than trying to reproduce the actual vocal organs and physiological mechanisms.

In 1828, Robert Willis, a professor of applied mechanics at Cambridge who had earlier dismissed the possibility of intelligent chess-playing robots, wrote disdainfully that most people who studied vowels “seem never to have gone beyond the vocal organs to inquire into their origin,” apparently assuming that vowels could not exist without them. In other words, they viewed vowels as “a physiological function of the human body” rather than “a branch of acoustics.”

Willis believed that vowel sounds could be produced by other means. Whether the vocal organs themselves could be artificially simulated became a separate question from whether speech sounds could be reproduced. Even as late as 1850, the French physiologist Claude Bernard wrote in his notebooks: "The larynx is the larynx, the lens is the lens, that is, their mechanical or physical conditions can only be realized in the living body."

Josef Faber's talking head Euphonia. © wikimedia

So deep was the disillusionment with speech simulation that when a German immigrant to the United States, Joseph Faber, designed a rather impressive talking head in the late 1840s, no one paid any attention. Faber's talking head was based on Kempelen and Mikael's model, but was much more complex. It had a human head and torso, again dressed up as a Turk, with bellows, glottis, tongue, variable resonance chambers, and a mouth with rubber jaws and cheeks. The machine could produce all the vowels and consonants, and was connected by levers to a 17-key keyboard that Faber could play like a piano.

Faber first demonstrated the machine in New York City in 1844, but it attracted little public interest. He then took it to Philadelphia, where it also received a lukewarm response. PT Barnum found Faber in Philadelphia, named the machine "Euphonia," and took it on tour to London, but even Barnum could not make it a success. Finally, the Euphonia was exhibited in Paris in the late 1870s, and soon all reports of it disappeared.

The age of talking heads is over. In the early 20th century, designers of artificial speech turned from mechanics to electronic speech synthesis. Simulating the vocal apparatus and speech process—the quivering glottis, the malleable vocal tract, the supple tongue and mouth—was a phenomenon specific to the last decades of the 18th century, when philosophers and mechanics briefly became obsessed with the idea that articulate speech was a bodily function, that Descartes’s gap between mind and body could be bridged in the vocal apparatus.

By Jessica Riskin

Tempura

Proofreading/Rabbit's Light Footsteps

Original article/publicdomainreview.org/essay/early-androids-and-artificial-speech/

This article is based on the Creative Commons License (BY-NC) and is published by Tempura on Leviathan

The article only reflects the author's views and does not necessarily represent the position of Leviathan