October 4, 2024, by Brigitte Nerlich

Superintelligence: From the divine to the digital and back again

The word ‘superintelligence’ has been bandied about a bit recently, most prominently by Sam Altman, the CEO of OpenAI, who said in a blog post on 23 September, 2024: “This may turn out to be the most consequential fact about all of history so far. It is possible that we will have superintelligence in a few thousand days (!); it may take longer, but I’m confident we’ll get there.” About ten days earlier Hannah Fry, the renowned science and maths communicator, had also used the word or words ‘super intelligence’ as a title for a podcast.

Hyped or hypothesised, superintelligence, or more precisely ‘artificial super intelligence’ or ASI, has become somewhat of a buzz word in the context of discussions about artificial intelligence and started to buzz around in my head nagging me to do a bit of conceptual archaeology. Where does the word come from and where is it going?

From the divine to the digital

The word ‘intelligence’ as the ‘faculty of understanding’ has a long history, as charted in the Oxford English Dictionary. It was, it appears, first used in 1390. There is a really interesting etymological blog post about it. ‘Super’, in the sense of ‘very good’ was first used quite a bit later in 1769. ‘Superintelligence’, in turn, is even younger. The word was, apparently, used for the first time in 1822 to refer to ‘divine superintelligence’.

At that time, the word referred to “A super-intelligent being, esp. a spiritual one, or one alien to humankind; spec. God or a god, considered as the creator of intelligent design.”

From 1839 onwards, the meaning shifted towards the modern one, although still linked to the supernatural rather than the superhuman: “The capacity to apprehend what is beyond the normal range of human intelligence or understanding; spiritual or paranormal insight or awareness”.

Around 1876, the modern meaning became established: ‘Greater than normal intelligence; great intellectual power or ability.” An interesting example of use that one should take to heart in current political debates can be found in 1916: ‘For all their belief in themselves, their super-intelligence and their Kultur, the German people are no Imperial race.” And the last attested use is from 2007: “Movie nerds, typically science or computer whizzes whose superintelligence makes up for social and physical limitations.” Hmmm, Hmmm.

Over time then, the meaning of ‘superintelligence’ has shifted from the divine to the spiritual and from the spiritual to the computational or digital, and from intelligent design to artificial, computational or digital design.

From scifi speculation to scientific debate

But 2007 is not the end of the story. The quote given in the OED is taken from Nexis, the news data base that I so often use. So I used it again and searched it for ‘superintelligence’.

The word seems to have first been used in English news in 1977 with reference to a scifi book by Algis Budrys entitled Michaelmas.

I hadn’t heard about this novel before, but it seems quite prescient, as it tells the story of a “newsman [who] controls world events just as much as he reports them. His means of influence is an immensely powerful self-aware artificial intelligence called Domino, which originated as a modest telephony appliance in Michaelmas’ youth. Over the years, Domino has evolved into a digital omnipresence that can penetrate and control any electronic or computerized equipment, most notably communication networks of all kinds.” That sounds quite familiar.

The 1977 newspaper article about the book says: “Budrys is a first-rate novelist, witty, humane and adroit, with the rare ability not only to posit superintelligence but to dramatize it convincingly.” I bet there are many more scifi stories like this out there.

A few years later, in 1981, we hear about the “Coming debate on ‘creating superintelligence’”, not in scifi but in reality: “Someday, either decades or centuries from now, people will have to decide about building extremely intelligent machines. Some will see this as a threat to our species’ survival, while others will see them as a natural stage of our own development–not as them versus us but as a natural step of our own evolution.” This, again sounds rather modern, as, in fact, most of AI history does, be it real or fictional.

Some trends and trendiness

I didn’t read the thousands of articles that came after this one, but looked instead at some overall trends. The OED frequency of use chart shows that the word, although rather rare, increased in use gradually between 1920 and 2010 (when the chart stops). But lots of things have happened since then. I therefore also consulted Google Ngram viewer. Here the word stagnates at the bottom of the graph until around 2014 and then rises sharply, while on Google Trends the biggest uptick happens in 2020 when the apparently rather mediocre film “Superintelligence” was released.

Interestingly, underneath the graph with that spike, is a list of related queries and topics and they almost all relate to Nick Bostrom’s book Superintelligence, published in 2014. And when you look at the timeline of coverage on Nexis, there is a first uptick in around 2014 and then a virtually vertical rise in around 2022 probably related to Open AI’s release of ChatGPT which opened the modern-day floodgates of debate about AI and superintelligence.

As many people will know, ten years ago, the Oxford philosopher Nick Bostrom published his book Superintelligence: Paths, Dangers, Strategies which explores the potential of superintelligence to surpass human cognitive performance, and the risks it poses to humanity. This book put the word squarely on the map within the AI discourse in rather controversial ways.

From the digital to the divine

Bostrom defines ‘superintelligence’ as “an intellect that is much smarter than the best human brains in practically every field, including scientific creativity, general wisdom and social skills”. One incarnation of such an intellect would be “superhuman artificial intelligence”. That is rather a grand claim.

Be this as may, this definition became quite famous. It is quoted unattributed in a recent news item on artificial super intelligence and followed by an even more hyperbolic quote apparently from an IBM blog: “In essence, an ASI would be an inexhaustible, hyper-intelligent super-being. A nearly perfect supercomputer available 24/7, with the ability to process and analyze any amount of data with speed and precision that we can’t yet comprehend”. (The Economic Times, June 26, 2024)

We are, it seems, back to 1822 when, according to a quote in the OED, ‘providence’ was defined as “the particular act of providing..[and] is applied to the Divine Superintelligence”; only now providence is the act of whatever is provided by an ‘inexhaustible, hyper-intelligent super-being’.

In my last blog post I said that one knows intelligence when one sees it. What about superintelligence? It’s possibly not something you recognise but something you believe in, and it is dangerous to do that. Projecting agency onto something beyond us humans is more dangerous perhaps than the machine we project it on.

I wondered at the beginning of this post where the word ‘superintelligence’ came from and where it was going…. It seems to be going back from the digital to the divine.

As an experiment, I have now turned this post into an AI/NotebookLM (Deep Dive) generated podcast!! From the super-human to the super-cringeworthy? 

Image: William Blake, The Ancient of Days (1794), depicting Urizen. Frontispiece to Europe a Prophecy, copy D. Colour relief etching and white-line etching in blue, black, red and yellow; with added hand colouring. British Museum, London (Wikimedia commons)

• Share on Facebook
• Share on Twitter
• Share on Google+
• Share on LinkedIn