Aurelien – Signs Are Not Wonders…And machines are not wiser than us.

Artificial Intelligence is just the newest link in an concatenation in the human search for truth

Cross-posted from Aurelien’s substack “Trying to understand the world”

This file is licensed under the Creative Commons Attribution 2.0 Generic license

https://www.flickr.com/photos/152824664@N07/30188200627/

I was born in a time of miracles and wonders, or so it seemed, at least.

Motor cars were still quite rare, especially in working-class districts, but televisions were slowly becoming more common, with their fascinating if disconcerting ability to project pictures of the outside world onto a small screen in a gigantic cabinet. Newspapers printed stories about aircraft with jet engines taking glamorous people to glamorous places, although the number you might actually see in the air was minuscule by current standards. And of course small boys of that era breathlessly followed the flights of cosmonauts and astronauts, culminating in the worldwide religious observance that was the flight of Apollo 11. For the more technically minded, there were radio telescopes, nuclear power, new antibiotics, DNA, and the Big Bang, among other things.

News about science was everywhere, including on the new medium of television, and the universal if unspoken assumption was that as technology developed, it would make everyday life easier and safer for ordinary people. In turn, this was because much science research was done by governments, and much of the rest by publicly-owned companies. The norm of the technology entrepreneur as a huckster just out to make money as quickly possible was still generations away.

But towering over these technologies, remote and inexplicable, the stuff of dreams, fantasies and nightmares, was the technology whose cultural significance I want to talk about today: the computer. Of all the exciting technologies of the postwar world, computers were the most mysterious, because most people did not have the remotest idea how they worked, or what they actually did. For a long time, computers were a solution looking for a problem, and in some cultural senses, as we shall see, they still are. Aircraft flew, after all; nuclear power was a way of producing heat, antibiotics made you better. But from the very beginning, computers were trapped in a conceptual fog, and what was claimed for them then (as it still is now) was always some time in the future, and always presented in extremely general terms. Those who foresaw great things for them were often not very technologically literate themselves, and it wasn’t always clear how the wonderful benefits they promised were to be realised.

Yet from the start, it was clear that the popular imagination both wanted and feared something entirely different from the massive, cumbersome number-crunching machines that actually existed in the 1960s and 1970s, and many quite responsible technologists and journalists believed that, whatever it might be exactly, its arrival was only a question of time. They were thinking of machines that were actually conscious, and had what we would now call, misleadingly, “artificial intelligence.” And these machines were going to do great and wonderful things, since they would be more intelligent than us, but also—if this was rarely said directly—wiser than us. And the urge to believe in machines that would solve all our problems for us, and the fantastic promises that were made as a result, haven’t really changed that much over the last fifty or sixty years.

I don’t know whether you ever saw a computer in those days. When I was a postgraduate student the College had one, taking up several rooms, where staff and students could “rent computer time,” if they were editing a text for example. I never used, it but I understand that there was an initiation process rather like that of the Rosicrucians. Meanwhile, people studying technical subjects walked around with stacks of green and white paper, and would show you how, after a lot of careful fiddling, the massive and noisy dot-matrix printers of the day could be made to disgorge something that vaguely resembled a picture of a human figure. And yet these machines, not tomorrow, perhaps, but the day after, or the day after that, were going to run our societies for us. Apparently.

Part of the disconnect was verbal. “Computer” is an English word, a noun from “compute” meaning to carry out calculations. This is what the first models did: the British used the world’s first true operational computer, the Colossus to break the German Enigma cypher during the Second World War. After the War, the pragmatic and commercially-minded British used the rapidly-developing technology not only for continued code-breaking and scientific research, but also for stock-keeping, logistics and accountancy. Still a long way from taking over the world, true, but a technology which was expanding into a void in the popular mind which had to be filled by something .

Oddly enough, the relationship is clearer in French, where the equivalent word is Ordinateur , first coined in the 1950s. The literal meaning is something like “organiser” or “bringer of order,” and the word has cognates in English, with “order” and, especially “ordain,” as with a Bishop. In French, the word preserves the rare sense of someone who exerts discipline within the Church. (Ah, perhaps we begin to see something, now.) It so happened, and with absolutely perfect timing, that the concept of the super-computer, rather than the reality, arrived just at the point where religious belief was entering a steep decline and where, in particular, the authority of the Church and of the interpreters of the Bible was being called into question. (This relationship was more obvious in disciplined Catholic France than in anarchic Protestant England.) In effect, the huge computers of the day were churches, the operating manuals were the Bible and sacred works, and the large numbers of staff, from cerebral programmers down to humble operators, represented the hierarchy of the clergy—an idea found in popular cultural clichés of the day. Needless to say, mortals could not address these machines in anything resembling everyday language, nor understand their output without someone to interpret it for them.

Thus, as well as the technological push effect of the advancing technology, there was the cultural pull effect of the demand for something which would fill a yawning intellectual, but also spiritual and ethical void. This void was partly related to the decline of religion, but it actually goes beyond just that, and reflects, following on from last week’s discussion of paganism, the need to have access to something wiser and more powerful than we are, in a disenchanted world. From the very beginning, therefore, there has inevitably been a gigantic gap between what is wanted, and what computers are actually able to do: a gap exploited since the beginning by technology enthusiasts, and more recently by con-artists as well. And as the real capabilities of computers have increased, so fantasies about what they might accomplish have increased even faster.

The two great influences on the way computers were popularly seen, and to some extent still are, were science fiction, and popular science publications. (The second was often inspired by the first, rather than the other way round.) Science fiction, as I’ve long argued, is the mythology of the technological era. In some cases the parallels are explicit (in 2001 , for example, or some of the work of CS Lewis, Roger Zelazny or Gene Wolfe) or indirect (the apocalypse stories of JG Ballard and others) or just pervasive, as with mythical props like magic powers, thunderbolts and the ability to fly. In the hands of maestros like Ballard the then-modern props like radio telescopes receiving mysterious messages from the stars took on a haunting, independent mythical life of their own, completely unrelated to the humdrum daily reality of scientific work.

And of all the traditional themes of mythology, and thus of science fiction, none is more powerful than the search for wisdom and for knowledge of the future, especially of one’s own destiny. This could mean oracles, divination (the I Ching has been compared to a simple binary computer), prophetic books or inspired prophets, but also, and more pertinently, talking statues, djinns, angels and demons, revenants and talking animals. Much of this, of course, has been assimilated into world literature, from Homer to the Faust legend, but it has also found its way into modern science fiction, which used “Artificial Intelligence” as a plot device long before todays would-be “AI” geniuses were born. Almost from the beginning, SF writers (many of them scientists) simply assumed that, as computers of all types became more complex, they would move automatically in the direction of consciousness: the basic idea has not changed since. And logically enough, as these machines became bigger and more complex, so it was thought they would rapidly become more intelligent than humans. But would they use this intelligence for good or ill? Would they be wiser than us?

The literature about computers as beings with godlike properties or effects would require an essay of its own, but it sometimes produces wry, almost Borgesian, reflections on our tendency to offload moral and ethical questions onto any convenient receptacle. The most famous example is probably Frederick Brown’s 1954 short story The Answer , where scientists link every computer in the Universe together to ask the question “is there a God?” “There is now,” the system replies curtly, fusing circuits so that it can never be turned off. (The story is parodied, of course, in Douglas Adams’s Hitchhikers Guide to the Galaxy , where scientists build a computer called Deep Thought which will tell them the answer to “Life, the Universe and Everything.” The answer turns out to be “forty-two,” which is unfortunate because they had forgotten to specify what the question was.)

But whether deadly serious or playful, all such stories incorporate one essential question, which has been present in monotheistic cultures since the beginning: is knowledge always a good idea? The story of the Garden of Eden and the eating of the forbidden fruit has been interpreted in very different ways, and I suppose that our modern, liberal, rights-based society would be on the side of Adam and Eve, and correspondingly against the injunction that they should not eat the fruit of the tree of life and so acquire knowledge of Good and Evil. Yet there has always been a conceptual problem here, since Good and Evil are ultimately value judgements, no matter how well founded we believe they may be in a given case, and not something that can be reduced to simple empirical “knowledge.” In effect, the story can be seen as an allegory of human attempts to arrogate to ourselves the right to make absolute moral judgements independent of a transcendent power. (Changing the argument just to the knowledge that there is a difference doesn’t really make the question much easier.) Experience suggests that, at best, human judgements about Good and Evil are tricky, but once belief in a divine creator disappears, there is no other option. In other words, we are all the descendants of Adam and Eve and the inheritors of their choice to make purely human judgements about Good and Evil. Unfortunately, history suggests we are not very good at it, and so we look around for God-substitutes to do it for us.

We have some difficulty today with the very idea of Forbidden Knowledge, and if anything we view the idea as an affront to our ego, and as a perverse encouragement to find out that which we are not supposed to know. From time to time, usually around the middle of August every year, someone will wonder aloud whether in fact it wouldn’t have been better if atomic physics had never been pursued as a field of study. And it’s common to bewail the negative effects of technological discoveries from plastics to silicon chips. But with very few exceptions (gain-of-function research on viruses for example) it is unusual to argue that the pursuit of knowledge should be controlled, or that certain types of research should actively be banned. (I’m not of course referring to research which might produce politically-unpalatable conclusions.)

It’s therefore significant that the only area where large numbers of experts and non-experts have suggested slowing, or even stopping, research is indeed that area known with tiresome predictability, as “Artificial Intelligence.” And here, once again, what people are afraid of is less the actual capabilities of computer systems today, than how they might conceivably develop at some stage in the future if the predictions that have been made for the last seventy-five years finally come to pass. In effect, and for the first time in a long time, people are starting to worry whether the search for infinite knowledge doesn’t have some small potential drawbacks attached. (You’ll remember perhaps that in Marlowe’s Dr Faustus , the protagonist’s last words as he is dragged off to Hell are “I’ll burn my books!”) Whether governments and private companies should be allowed to possess infinite knowledge about citizens is itself a live topic of debate.

Of course, some of the problems have already existed for a while. Many arise from the erroneous belief that “Artificial Intelligence” is already here, and the related tendency to anthropomorphise the behaviour of computer programs, which I’ll come back to in a moment. I first noticed this forty-odd years ago, working with the first generation of computers installed in government offices in Britain. To give you some context, I’ll just say that the operating system was more difficult to use than CP/M, and that the computer itself with its mighty 20Mb hard disk, and the printer which made a noise like a pneumatic drill, had to be kept in a separate room. The machine was a capricious god, punishing you arbitrarily for some tiny syntax error, and occasionally refusing to communicate for hours at a time, just because it felt like it. When the system stopped working, people would gather around the terminal where the problem had originated, scratching their chins and saying things like “have you tried… ?” There had been no training, because there were no trainers in those days, and nothing more than a pamphlet from the manufacturer, leaving out everything you might actually want to know. (The origin I suppose of the old joke: if all else fails, read the manual.)

But on one of those days when the machine actually felt like cooperating, and I was peacefully engaged in working on a spreadsheet, a Very Senior Person for whom we worked came into the office and looked over my shoulder. “Ah-ha,” he said jocularly, “is your machine going to tell me how to run the Budget then?” But I couldn’t blame him. The whole of popular culture up to that point— 2001 , Star Trek, Dr Who and many others—had encouraged that view, precisely because there was an enormous intellectual void to fill; so the idea of a computer that could, all by itself, manage government budgets, was sufficiently attractive that it eventually took on an independent symbolic and mythical life of its own. This was also at the very beginning of the period when computers were being widely deployed to individual desktops, often without much idea of what they would be used for, and anticipating very precisely the fatuous optimism and short-term thinking of today’s “AI” boosters. It was an age when you could still use phrases like “computer studies show….” with a straight face.

The British Treasury, of course (unofficial motto, “a penny saved is an end in itself”) immediately spotted the opportunity for saving candle-ends. Tens of thousands of skilled jobs could be “saved” (ie lost) they salivated, and everyone could now do their own typing. The results were predictable: instead of spending five minutes dictating a letter, quite senior people spent an hour painfully typing it, trying to format it, finding that the text overflowed onto the next page by a single line, and then forgetting to save it or saving it in the wrong place and not being able to find it again. Because I had a reputation for being slightly less computer-illiterate than average, people would approach me, ashen-faced, to say “I think I’ve lost the three-page paper I was working on. It’s disappeared off the screen.” Not very impressive for a technology that was supposed to take over the world, and indeed it’s not surprising that studies into evidence for increased productivity through information technology have generally failed to find any. This was a problem that was supposed to cure itself generationally, but effectively it hasn’t, not least because software and computer manufacturers have continued to add features and functions piled on top of each other to justify new versions and new prices. (For professional reasons, I was obliged a couple of years ago to download the latest version of Microsoft Office, which I’d managed to avoid using for some time, and spent several minutes looking at the interface for Word, overcome by a sort of numb horror.) Indications are that it’s the same with “AI.”

I’ve already been guilty of anthropomorphism, which is virtually impossible to avoid when you are dealing with a technology that goes through visible processes, and outputs words and numbers. But if you had to wrestle with the damn things in those days you did at least understand that there really wasn’t a little man in there doing everything. These days, in spite of most people being theoretically of the “computer generation,” the unconscious anthropomorphism is even stronger, because of attempts to make computers more “friendly,” with an actual interface you can theoretically understand, and even obtrusive “Help” menus, replacing the deadly “C:\>” prompt on a blank screen of thirty years ago. (I would occasionally wander into computer shops in those days and, for my own amusement, type “what is the meaning of life?” at the prompt. Unsurprisingly, the computer spat back “syntax error.”)

Now I’m all for intuitive interfaces and ease of use (I’ve been an enthusiastic Mac user almost since the beginning of OS X) but the problem has been that, as computers have been able to do more and more things, people have begun to believe more and more that they can do anything and everything, and better than we can. Thus, the present panic about “rogue AI” and the “AI apocalypse,” which are both marketing hype, and the lineal successors of the literature about the dangers of the excessive pursuit of knowledge that dates back thousands of years now. The difference is that, whereas djinns and talking heads were unabashedly magic and that was it, the modern oracles of “AI” are supposed to work according to scientific laws. But this doesn’t stop people investing them with non-existent human qualities, or treating their output as intelligent, so desperate are they for the truth about Life, the Universe and Everything.

Now, I am not a computer scientist, but Cal Newport is, and he has explained essentially what is going on. Large Language Models (which is what they are) including programs providing prompts for other programs to act, are restricted to describing their actions, and providing prompts, based on verbal formulas they have been trained on. If the training material includes lots of science fiction, including stories of computers taking over the world, then the results aren’t too hard to guess. If the training data was restricted to, say, the works of Plato and his followers, you’d get another kind of output.

None of which really resolves the actual question of whether we really want something or some machine to have infinite knowledge or not. Indeed, the question probably can’t be satisfactorily resolved, because our feelings about it will depend almost entirely on the context in which it arises. If a murder occurred in a given area, and all mobile phones in the area were recorded, and their owners cross-referenced against identifications of those passing by through facial recognition software and criminal intelligence records, and if this yielded a short list of five people, one of whom turned out to be the murderer, well, crime solved. Good idea. But let’s assume you were in the area, and your phone was one of those picked up, but you were automatically checked and dismissed without human involvement. Could you say that your rights were violated by a piece of computer code when neither you nor the person carrying out the search was even aware that you were on the list?

I don’t know, but I do worry about the Faustian temptations that “AI” will give rise to, especially when uninformed but powerful people become convinced that “AI” can do a better and more accurate job that humans. That is not generally true, of course, outside specialist areas, unless you define “better” as “faster and therefore more cost-effective.” But “AI” knows only what it has been told and is by definition only capable of producing probabilistic outcomes. This is why its apparent use for targeting in armed conflict is so worrying: human decision-making is not perfect either, but it does require actual thought and an overt decision-making process. After all, one of the problems with “AI” and its ancestors has always been the spurious and deceptive precision of its outputs, down to decimal points in some cases. You may recall the attempts to computer-model the Vietnam War, which produced highly detailed and convincing garbage outputs, resulting of course from garbage inputs.

Can you have too much information or knowledge? Are there questions that should not be asked? The age of “AI” has suddenly made these traditional questions deeply topical and of course we no longer have the intellectual framework to consider them properly, which is why the public “debate” ricochets around between enthusiasm, complacency and panic with no particular connecting logic. Faustus, you will recall, wanted to know everything , and it’s not fanciful to suppose that modern governments, battered and unpopular as they are, will stumble further and further into mass data collection and what they think of as “AI,” in an attempt to make up for the deep-seated sense of insecurity they must increasingly feel. Now historically, like a drug user increasing the dose, governments have tended to believe that they can compensate for their own ineptitude by acquiring more and more information, until they were drowning in it. They had forgotten, of course, that information by itself means nothing, that what is required for it to be useful is analysis and organisation. Enter “AI,” which makes the exciting but spurious promise of rapidly extracting signal from noise, but cannot actually do so with sufficient reliability, even in principle. But in government, a tool that addresses a problem, or at least makes it look as though a problem is being addressed, is too precious a resource to give up lightly.

Whether information is the same as knowledge or whether it will necessarily lead to it, whether more information is always better, whether there are indeed things we should not know and it would be dangerous for us to discover, are all questions for another day. Here, I would just recall that as a matter of practicality, and as much in our daily lives as in weightier issues, we do sometimes find that, after all, there are things that we wish we had never learnt. “After such knowledge,” asked TS Eliot a century ago in Gerontion “what forgiveness?” (The American science fiction writer James Blish wrote a tetralogy on this theme, under the title After Such Knowledge (1958-71), and CS Lewis retold the Garden of Eden Story in Perelandra (1943).) In that context, it’s amusing to find Eliot suggesting, in Gerontion that “Signs are taken for wonders.” What better epitaph could there be for “AI?”

To repeat, the fantasies we weave around “AI” are generated, not by the actual or reasonably anticipated virtues of the technology itself, but by existential holes in our culture that remain achingly unfilled. One of these, the desire for knowledge or at least information, I’ve already discussed at sufficient length. But there is also the related but separate idea of Wisdom. The two, though both very old, refer to subtly different things. “Knowledge” has always had the more technical sense of “awareness of things,” whereas “wisdom” has always had a normative element to it, of right judgement in a given situation, or on a specific subject. So we talk of “a piece of knowledge” as something essentially contingent and temporary, whereas “a piece of wisdom” is something to remember for the future. (It would be nice if one necessarily led to the other, but this is not necessarily the case.) We talk of the Wisdom of Solomon, not just the Knowledge of Solomon, though he clearly had both. Thus, in the famous Biblical story of the Judgement of Solomon in 1 Kings, Solomon awards the custody of a disputed child, each claimed as their own by competing mothers, to the woman who is prepared to give up the child, rather than see it cut in half, because she must be the one who loves the child. This story, which has its origin in mythology and folktales throughout the region, and appears in all monotheistic religions, is essentially about Solomon being a shrewd judge of character and adopting a wise course of action. There was no actual way of knowing which of the two women was the mother, after all.

This is wisdom, and it is something our society desperately lacks. It is not just knowledge, true or false, it is not just “facts,” right or wrong, or “expertise,” whatever that is exactly. It is something that is difficult to acquire without experience and age, which is awkward to acknowledge these days, and depends on hierarchic relationships and a willingness to learn, which are not part of our current culture. But that doesn’t mean that we dismiss wisdom: far from it. We may turn our backs on parents, older people, schools, institutions and the like, but we still instinctively seek out authority figures who know, or at least seem to know, how to get things done, and what to do. Sometimes this makes everyday sense. On YouTube for example (and leaving aside all the polemical and fringe political sites) there are any number of pragmatically useful sites run by people with lots of practical experience and a gift for communication, for times when the instructions that came with the flatpack furniture you ordered are incomprehensible in any language. That, you might say, is just trivial knowledge, but there are also sites that teach you, for example, what your grandfather taught your father about woodwork, or managing money, or dealing with authority, but which you never learned, because your parents separated when you were young and you didn’t get on with any of your step-fathers.

This, but on steroids and in all areas, is what we want from “AI,” and some people are foolish enough to believe that they can get it. In a world of discarded hierarchies and deconstructed authority, we want something to act as a reliable guide and interlocutor to help us resolve our personal and professional problems, without feeling we sacrifice anything of our Ego in the process. The technology of “AI” is such that any answer to any question on any subject has the same level of apparent verisimilitude, but in reality has entirely unpredictable and unknowable, levels of accuracy and usefulness. Among all the gurus on the Internet telling you how to live your life (and there are some good ones, let’s admit it) there will soon be many “AI” generated figures of more or less usefulness, programmed with anything from the complete works of the Buddha or Thomas Aquinas to, well, Mein Kampf. I don’t know whether there are interactive online “AI” psychologists and psychotherapists yet, (though Bots certainly exist) but if not there soon will be, and it will be impossible to control their use.

From the dawn of the computer era, humans have fantasised about machines which are vastly more intelligent than they are, and therefore wiser and able to guide and even direct human life. In popular culture terms, the classic example is the Culture (sic) series of novels by Iain M. Banks, who unfortunately died a few years ago. The novels concern a humanoid spacefaring species whose lives depend absolutely on benevolent, unbelievably complex and powerful Minds, which take responsibility for virtually everything of any importance, leaving the citizens of the Culture nothing much to do except enjoy themselves. (Indeed, finding something worthwhile for humans to do in the books was actually a continual challenge for Banks in developing his plots: thus, the creation of a number of unpleasant and dangerous alien species that had to be dealt with)

Much as I enjoyed the Culture books, the one thing that emerges from them (as I pointed out a few years ago now) is that the characters are all basically children, whose every want is instantly gratified. It’s no wonder then that boredom is one of the major problems of the Culture, and so its more thoughtful citizens devote themselves to Good Works on other planets, trying to guide them towards civilisation. But Banks began writing the Culture books in the days of MS-DOS (now wash your mouth out) when the idea of “AI” that could do everything for us was essentially only a fantasy. Now, if you believe the boosters, it’s just round the corner. There, we will have a world where everything difficult will be done for us by others. We will revert to the stage of pre-adolescent children, for whom their parents are all-powerful beings who do everything for them and solve all their problems. It’s often said that “AI” will “enable us to do” this or that. At best, that’s an oversimplification. In general it will enable things to be done for us by a machine, instead of us leaning how to do them ourselves.

In this sense, it’s absolutely unsurprising that the first uses of “AI” have been to cheat and deceive. Wouldn’t we all have liked to have someone do our homework for us? Wouldn’t we want someone to produce a wonderful job application for us? Wouldn’t some people, at least, like a Ph D without having to work for it? Wouldn’t others have liked a novel or a musical composition they could pass off as theirs, and earn money from? All without doing any actual work, according to the best neoliberal principles. The most obvious result of the widespread use of “AI,” even in its present miserable state, will therefore be a massive de-skilling of the population, in favour of technologies and processes which “do the work” for us, but are nonetheless not accurate enough of the time to make the productions of that de-skilled world really useful. I must have been ten years old when a teacher at my primary school said that there was less and less point in teaching mathematics, because soon we’d all be using pocket calculators. Well, this was true to a point, but it turns out that, with the large-scale use of such devices, mathematical skills at advanced levels have declined sharply in recent decades. That is probably an augury for the future.

The Minds in Banks’s novels are functionally indistinguishable from the kind of gods I wrote about last week, and of course the major, but largely undiscussed, risk is that we will eventually turn “AI” into a substitute Creator God, of the type that we have formally banished in the Western world. As Alasdair McIntyre pointed out, coincidentally at the start of the AI Fantasy age, human beings have shown themselves incapable of devising general ethical frameworks which might take God’s place. This is not so strange if you think about it: we can imagine the group being collectively more knowledgeable than the individual, but it’s much harder to imagine the group being somehow ethically superior. Yet as David Bentley Hart has eloquently described , a purely materialistic explanation of Life, the Universe and Everything necessarily disintegrates as soon as it is attempted. I suppose you might as well expect a group of fleas on Mozart’s waistcoat to discuss intelligently the tonal structure of the Magic Flute .

The idea that there are things about the universe and even why it exists that human beings can never know, sits uncomfortably with the egoistic ethos of Liberalism. The idea that somewhere, some Being may know things we can never know is psychologically unacceptable, and this probably does more to explain the decline of organised religions than any other single factor. And it is hard to see how humans can construct a system of thought, let alone a mere machine, that genuinely understands things they themselves are incapable of. Enter for the last time then, the magic of “AI,” which can provide the illusion of this understanding, and will, moreover, comfort us by giving us whatever answers we want, and without any intellectual or moral effort needed on our part. No doubt there are “AI” religious cults already; we can assume there will be many more, and all we will have to do is worship and obey.

BRAVE NEW EUROPE is one of the very few Resistance Media in Europe. We publish expert analyses and reports by some of the leading thinkers from across the world who you will not find in state and corporate mainstream media. Support us in our work.

To donate please go HERE .