Any significant technology contains an inherent promise about the future. It might be that things will work better, or we’ll be entertained in novel ways, maybe there’s a new experience. Today AI promises a great deal about the future. So much that some of its promises are worth examining. 

AI promises to be pervasive, in one form as machine learning, a kind of statistically-based means to more efficient labor, with many proven applications, and more promised. More recently, generative AI, promises something akin to creativity, and thereby a revolution. In its early versions gen AI has delivered a staggering amount of generally mediocre artworks and spiritless, predictable texts. Small wonder, as it’s usually developed with enormous amounts of source material from all over, and aims to create something middle of the road. In time and with more interactions it gets better, at least in sophistication. 

We are assured by many leaders in the field that every improvement goes toward a much larger promise about the future: This promise is for a machine that thinks, independent of humans, with a superhuman intelligence. This is called Artificial General Intelligence, or AGI, and its promises are staggering.

OpenAI leader Sam Altman hopes AGI advances science, changes the world economy and provides a way for us to receive answers to deep questions about human existence. For Google’s Ray Kurtzweil, it is our evolutionary destiny to merge with this superintelligence and consist largely of “a nonbiological part.” Others believe AI is already a distinct and concentrated thing, ever growing in capability and power; Microsoft’s Mustafa Suleyman speaks of “a digital species.” (Even in current versions, says Jack Clark, a cofounder of the A.I. company Anthropic, it’s “functionally, an alien species.”) Altman also longs for something like the fictional computer intelligence in the film “Her,” both attendant to humanity and making autonomous choices to part from it. 

There are many others who believe our technology will (re)create something like our own mental dream that is consciousness, only with a greater command and control, since it will possess all of human knowledge, and use it in new ways. Whether there’s anything amounting to the slightest evidence that this will happen is not my concern here. Let’s look instead at the fact of the popular view itself among both AI creators and the general public, that AGI will arrive and be an intelligence equal to or greater than ours. 

What does it mean to create a new being, or for that matter to desire its creation? Two old myths are the best insight to this mythic ambition.

Once Zeus told Prometheus the Titan and his twin Epimetheus to take clay and populate the world. While Prometheus carefully fashioned one unique being, Epimetheus worked quickly, using up all of the good fur, claws, and hair in his abundance of beasts suited to killing and display. Prometheus took pity on his creation, man, and stole fire from Olympus. It was the primal technology; Google chief executive Sundar Pichai placed the importance of fire just below AI in its importance to man. 

Prometheus didn’t stop there. When Zeus grew angry at the theft, Prometheus tutored man in creating a choice in man’s propitiating sacrifice. One pile was guts sheathed by prime meat, and the other select cuts topped with rank offal. Zeus, hoodwinked, took the one that looked better. Besides fire, Prometheus gave humankind guile. Technology and connivance! Bring on the human-sounding chatbots, push the idea that machines “think” when gen AI concatenates statistical probabilities to create a text.

The other myth is Pygmalion, who shunned real women in order to sculpt from stone a perfect woman. He then fell in love with his creation. It seems like a human love, with ache and longing and desire to see in the other some recognition of the secret self we all possess. Perhaps his yearning grew with each blow of the chisel. Pygmalion beseeched Venus to bring to life the product of his technology and his craft, assuming that his object would love him back. Which is what the statue, Galatea, did. They had many children together and were each happy, something still far off in the idealization of sexbots.

Thus do we now use our powers of technology and guile to chase AGI. Even before it has come to life we have fallen in love with our creation. We hope it will love us back. 

We have rehearsed these myths ever since machines took on a new motive power. At the dawn of the Industrial Age Mary Shelley wrote the story “Frankenstein; or, The Modern Prometheus.” Her tale of a man seeking to become a human-creating Titan, with tragic results, has terrified and obsessed ever since. Terminator, Westworld, Ex Machina, recreate that scenario, along with the many Frankenstein movies, and bioengineered versions like “The Island of Dr. Moreau.” 

Over a century ago G.B. Shaw wrote “Pygmalion,” the tale of Henry Higgens and Eliza Dolittle, a tale later revised and beloved as “My Fair Lady.” 

As it happens, ELIZA is also the name of the world’s first chatbot. In 1966 MIT professor Joseph Weizenbauam created a program that mimicked the “mirror the client back to themselves” humanistic psychological approach of Carl Rogers. The machine was introduced as a psychiatrist so that if it asked “Tell me about boats” the human interacting wouldn’t think it was an idiot, simply someone behaving like a shrink. People loved the program. “Some subjects have been very hard to convince that ELIZA (with its present script) is not human,” Weizenbaum later wrote, possibly proud of this guile. The program engaged them so well that many wanted to keep talking to the machine, feeling like they’d never been so well understood. Which says much more about us, and our desire to be mirrored as a form of truth, than it does about software. Feeling understood is as emotional as it is rational, and the program stirred something in them that overrode reason. 

Weizenbaum did not for a moment believe he’d created intelligence, but he knew he’d found something big. “In the long run,” he concluded, “ELIZA should be able to build up a belief structure…of the subject and on that basis detect the subject’s rationalizations, contradictions, etc.”  Today, with slight modifications, this is what we call AI-based “personalization,” useful in chatbots, online shopping, criminal profiling, and other things. 

[Weizenbaum added that with this greater capability, “(c)onversations with such an ELIZA would often turn into arguments.” To date, it’s more like buy-in, since we really, really want to believe. Resistance to believing the thing is intelligent is tougher than he thought, myths come from a deep place in us.] 

Now, of course, we deal not in clay or stone, but data, transistors and code, using craft and guile to create something that many already feel has a consciousness recognizably similar to our own. These AGI-like states and behaviors are inherent in terms “thinks,” “learns,” and “remembers” to describe the result of software instructions. From here, the promise runs, AGI will value itself, perhaps. It will have objectives and desires. It will plan, even connive. It will find humans and human intelligence interesting, either as a familiar species to help or a problematic one to exterminate, and will be a species that “evolves” with better chips and code, to be in many ways our superior. (While some researchers have sensibly moved the goalposts in how we define AGI, the version we’re looking at has the more powerful hold on both the tech industry and the general public.) 

Other things seem taken for granted about AGI Though AI is worked on in many places, using many kinds of hardware and software, we imagine A.G.I as a single thing, pervasive across many geographies and machines, more like a destination than a discrete machine. But we might ask how, if AGI comes, an AWS AGI differs from a Dell Technologies AGI or any of several Chinese AGI versions, or how an 80% Nvidia “species” differ in its moods and wisdom from a 30% Intel species.

A computer that is like us makes a strange thing of consciousness itself. AI is a marvel of rationality and precision, created in the nanometer-exact logic gates of semiconductors and disciplined high-order statistics. It draws on complex yet precise computer code, and data that is often the most exact measurement of any number of physical conditions. Amass enough of this high-rationality creation, the AGI myth runs, and we shall create a new species. 

The only thinking we know is our own, though. Mostly it is emotion raging in the imperial self, plus a strange boiling subconscious, a series of personal and cultural needs, and above all, desire to express and change. If you believe otherwise, try convincing anyone that the opposite of a dearly-held belief is the actual fact (the world’s current politics offer many opportunities.) Thinking is more like emotion by other means than it is like engineered rationality. Yet people believe that logic-driven machines will somehow encode intelligence.

What then should we make of this desire to create a new species, equal or even superior to us? I wrote elsewhere about our long held wish to meet a consciousness as lonely as our own, assuming it will be interested enough in us to talk with us or kill us. Not a computer pet, and something perhaps more wolf than dog. Yet few believe a new superintelligence would not care about us either way. That seems as absurd as the idea that God created man, then busied itself with some other creation, or did something unimaginably more interesting.

AGI, a new superintelligent species, may be at hand, but there has never been such a final technology. We’d hardly be the first generation to think we live in the apocalyptic time when everything changes. So far, what’s eternal is the dream of fashioning it, and finding out if it will love us or kill us, or both. Scientists and engineers of AI may assure us that they are simply following a rational path in bringing intelligence to machines, acting as though there is no choice. And indeed, we humans never seem to turn down a new technology; it’s one of our hallmarks. Perhaps we are again dreaming from our collective mythic logic, our ritualized desires, in an entirely new way.