Over the last decade, artificial intelligence (AI) researchers have achieved genuine breakthroughs in the ability of computers to process language, sounds, and images. But these significant technical accomplishments have been accompanied by a constellation of overheated nontechnical claims about what these advances mean. Elon Musk has predicted that AI will “overtake” human intelligence “next year.” Forbes worries (or hopes?) that “Artificial Superintelligence Could Be Humanity’s Final Invention.” The computer scientist Ray Kurzweil asserts that we are entering into the age of “spiritual machines.” One can easily find a multitude of similar statements with a simple internet search. Behind these claims lies the conceit that the human problem of finding ourselves born into a world that we did not create and whose purpose we do not know can be solved by technical means. These views show AI enthusiasts to be not promulgators of a new philosophy but followers of a long tradition of those who have attempted to transform the natural world through technological magic.
Thus, it may be of interest to look into the past at similar enterprises that achieved genuine technical breakthroughs accompanied by speculative ideas of dubious validity. In doing so, we will make extensive use of Mircea Eliade’s 1956 book The Forge and the Crucible. In it, he traces how the “ideology” associated with metallurgy and mining gradually gave birth to alchemy, and how that ideology often made utopian promises about what these new technologies could achieve. (I don’t particularly like this use of the word “ideology,” but, having been unable to come up with a better word, I will use Eliade’s term.)
Eliade’s central thesis is that early mining, early metallurgy, and alchemy were all largely spiritual practices. That claim may be surprising to many readers. Typically, alchemy is thought of as primitive chemistry—attempting to turn lead into gold—and metallurgy and mining as purely practical activities. But Eliade demonstrates that all three fields always involved mystical, occult ideologies that saw these activities as far more than mere technical disciplines. As he put it:
What the smelter, smith and alchemist have in common is that all three lay claim to a particular magico-religious experience in their relations with matter; this experience is their monopoly and its secret is transmitted through the initiatory rites of their trades. All three work on a Matter which they hold to be at once alive and sacred, and in their labours they pursue the transformation of matter, its perfection and its transmutation.
Similarly, Kurzweil, writing in the Economist, promises that “AI will transform the physical world.”
Right up until the last few centuries, metallurgy and mining were surrounded by mystical rites and attitudes. Creating new furnaces or new mines was preceded by periods of ritual purity that culminated in sacrificial rites: “The transformation of holy matter (which is what metal is) demands for its accomplishment the sacrifice of a human being,” Eliade writes. Regarding primitive metallurgy, Eliade notes, “Very early on we are confronted with the notion that ores ‘grow’ in the belly of the earth after the manner of embryos.” This is much like the way that large language models (LLMs) are supposed to be growing in knowledge as they are being programmed.
Indeed, the tools of these trades themselves were regarded as magical, Eliade writes: “The tools of the smith also shared the sacred quality. The hammer, the bellows and the anvil are revealed to be animate miraculous objects. They were regarded as capable of operating by their own magico-religious force, unassisted by the smith.” LLMs, too, are regarded as capable of understanding their input and learning from it, unassisted by the programmer.
Eliade traces the ideology behind alchemy to its origin in earlier work with metals: “To sum up: in the symbols and rites accompanying metallurgical operations there comes into being the idea of an active collaboration of man and nature, perhaps even the belief that man, by his own work, is capable of superseding the process of nature.” And Roger Bacon, often considered a forerunner of the Scientific Revolution, wrote about alchemy as a “medicine which gets rid of impurities and all blemishes from the most base metal, can wash unclean things from the body and prevents decay of the body to such an extent that it prolongs life by several centuries.” Compare this to the AI promoter Dario Amodei’s claims in his essay “Machines of Loving Grace”:
But I suspect AI-accelerated biology will greatly expand what is possible: weight, physical appearance, reproduction, and other biological processes will be fully under people’s control. We’ll refer to these under the heading of biological freedom: the idea that everyone should be empowered to choose what they want to become and live their lives in the way that most appeals to them . . . [so] that most of those currently alive today will be able to live as long as they want.
Eliade says of the mystical metallurgist and alchemist that “in taking upon himself the responsibility of changing Nature, man put himself in the place of Time; that which would have required millennia or aeons to ‘ripen’ in the depths of the earth, the metallurgist and alchemist claim to be able to achieve in a few weeks.” Similarly, the AI alchemist tries to duplicate an evolution of consciousness that took hundreds of millions of years in the time it takes to train an LLM. The Forbes article linked to above says, “Unlike human intelligence, which is constrained by biology, ASI would operate at digital speeds . . . Imagine a being that could read and understand every scientific paper ever written in an afternoon.” Since computers have operated at many different speeds, it’s not clear what “digital speeds” are. But still: all of humanity’s understanding of the natural world, achieved over many millennia, will be surpassed in an afternoon! And Time piles on: “We are facing a step change in what’s possible for individual people to do, and at a previously unthinkable pace.”
The old alchemists sought to perfect matter by, for instance, turning base metals into gold. This is the kernel of truth at the core of the common misapprehension of alchemists as seekers of wealth by turning lead into gold: they did, indeed, seek the recipe for transforming “base” metals into gold, but because gold was “spiritually pure,” not because it was more expensive. As Eliade writes, “Alchemical transmutation is therefore equivalent to the perfecting of matter or, in Christian terminology, to its redemption.” Our new alchemists instead seek to perfect matter by turning stupid matter into matter that can compute. In The Age of Spiritual Machines, Kurzweil writes, “The point of all these . . . numbers is that extremely little of the stuff on Earth is devoted to useful computation. This is even more true when we consider all the dumb matter in the earth’s [core].” Much as the alchemy of old “delivers Nature from its own laws,” so does the new alchemy: “The laws of physics are not repealed by intelligence, but they effectively evaporate in its presence.”
So what do these grandiose claims mean for the real prospects of AI? Lest anyone doubt that a practitioner of a technical discipline can make enormous technical advances while yet embracing a dubious ideology, we need look no further than Isaac Newton and alchemy. This link was discovered by a perhaps unexpected follower: John Maynard Keynes. Upset by the low price that Newton’s papers had fetched at an auction in 1936, he began buying them up himself. In doing so, he uncovered Newton’s enduring interest in alchemy, a subject to which it turns out that Newton devoted more time than to mathematics or physics.
This discovery led Keynes to call Newton “the last of the magicians.” (Keynes himself, by the way, might be considered a latter-day magician, his theories transforming profligate spending and useless work into prosperity. In fact, Eliade saw the “systems of political economy” as carrying forward the alchemist’s dream.)
As an example of how Newton spent a large amount of his time, consider this line in a recipe for “sophic mercury,” which dissolved gold and allowed the precious metal to “vegetate” and mature into the philosopher’s stone: “Marry [sulfur] with . . . [mercury] which is impregnated with [and] must be espoused with our gold then hast thou two sulphurs married & two . . . of one of[f]spring whose father is the [gold] & [silver] the mother.”
Eliade explicitly recognized that the alchemist’s dream of transforming matter has not died with the death of alchemy: “the idea of a new Epoch, crystallized around the myth of infinite progress and boosted by the experimental sciences and the progress of industrialization which dominated and inspired the whole of the nineteenth century, takes up and carries forward . . . the millennary dream of the alchemist . . . visionaries myth of the perfection . . . of Nature survives, in camouflage form, in the pathetic program of the industrial societies, whose aim is the total Transmutation of nature.”
We don’t have to be ready to celebrate the marriage of mercury and sulphur and its offspring in order to appreciate Newton’s theory of gravity or calculus. Similarly, we can appreciate the genuine technical advances achieved by today’s AI shamans without buying into their occult fantasies of creating “spiritual machines” or fundamentally altering human existence by passing magical incantations over silicon crystals. We also ought to recognize that the technological “utopias,” whether those of the earlier alchemists or those of today’s AI alchemists, fail to address the fundamental mystery of human existence: what are our lives for? If I have no idea why I am alive or what I ought to be doing with my life, simply extending the span of my life and making it easier for me to avoid disease and hardship only means drawing out my existential angst over a longer and longer stretch of meaningless existence.
The new AI alchemists never seem to consider whether these silicon-based humans need a purpose. If his fantasy of a world in which most “humans” are actually machines comes true, won’t any of them ever ask, “What is the point of all this computation?” Or will they be satisfied simply to compute more and more, faster and faster, without any clue about the end for which they perform all of their calculations?
