Magic and divination in the age of AI

New Statesman​

3 days ago

Illustration by Ben Jennings
In a bookshop off London's Strand, fitted out in expensive utilitarian style – exposed ducting, unpainted plaster, birch plywood – a room of mostly young, artsy people are gathering for a talk about the future. There are cocktails and a DJ who is wearing a T-shirt that says 'Brutalism'. Facing the crowd is a sofa, above which is a projector screen displaying an AI-generated video of white, wafting things that look like fungal growths or jellyfish. Within their strands and webs are the faces of young women, their mouths moving without forming anything recognisable as words. The video was made, we are told, by someone who is a 'digital shaman'.
The talk has been convened by a group of people – a 'platform' – who are interested, in a fairly literal sense, in 'technology with spirit'. They are here to talk about the links between artificial intelligence and 'spiritual intelligence', forecasting and divination, the old gods and the new.
On the stage are an artist called Beccy, a tarot reader called Bel, and Sasha, a computer scientist who works for Google DeepMind but is also a meditator and practitioner of the Vajrayana Buddhist tradition. Beccy introduces herself as a 'climate storyteller' who creates 'imaginative acts of resistance'. This has involved, she tells us, training an AI model 'with love and respect… to welcome migrants and refugees', and also to 'generate a sense of inter-species solidarity'. People in the crowd nod enthusiastically.
Bel, the tarot reader, is not restricted to cartomancy. She also helps people – including corporate clients – to 'create oracles… systems for divination', she tells us. 'Many of them work.' Bel also works at the College of Psychic Studies, where other courses include alchemy, remote viewing and connecting with angelic beings.
Sasha, the computer scientist, works on reinforcement learning in multi-agent systems for Google DeepMind, which is owned by Alphabet, which has a market capitalisation of £1.64trn (at time of writing). But he sees a link between the old cultural practices of divination – scrying for answers in playing cards, tea leaves and the movements of birds – and the modern practice of asking ChatGPT or Gemini for the truth about something. Both types of system, he says, are 'cultural tools for generating behaviour'.
Joining the talk via the screen, in a brief and welcome break from the digital shaman's undulating forms, is Kevin, who runs an AI research group at Coventry University and a practice called Philosophy Machines. Kevin's work in AI is informed, he tells us, by his metaphysics, his willingness to ask questions about the underlying nature of reality: 'Is reality somehow constructed in a model that lives in our heads, or lives in a computer somewhere? Do we believe in things that we can't see… Do we believe that objects – rocks, clouds – have a soul, a spirit? Do we believe in angels, deities?' Companies and governments, he notes, 'exist only on paper, they only exist to the extent that everyone else believes in them'. Perhaps this is one to try in bankruptcy proceedings: does a business stop owing money if we stop believing in it? (No.)
Give any first-year philosophy student a bag of weed and you'll have to listen to similar questions being pondered for as long as you can stand it. But these seem like sensible people, who work for major companies and institutions. And they are not the only people having such conversations; beneath the AI boom is a philosophical debate, and a struggle for power.
Subscribe to The New Statesman today from only £8.99 per month Subscribe
For 45 years, people have debated a thought experiment created by the American philosopher John Searle. The experiment goes like this: there is a room into which messages are posted. The messages are written in Chinese. Inside the room is a person who doesn't speak Chinese, but who has a book containing very detailed instructions on how to arrange Chinese text into convincing responses. So, the person matches symbols they don't understand to instructions, assembles new strings of characters – which are, to them, still meaningless – and posts them out again. To the people outside the room, it appears as if the occupant can speak Chinese. But they can't.
This thought experiment, published in 1980, has become much more relevant today, as it is fundamental to the debate over whether 'AI' is, in fact, artificial intelligence. A large language model (LLM) is 'trained' by breaking up text and images into chunks (not words, but strings of characters that appear together as parts of words) and then calculating the sequence of pixels and characters most likely to be accepted as a response. The companies that sell the LLMs say this amounts to reasoning; scientists and philosophers of language disagree.
One response to this is to argue, as Sam Altman (CEO of ChatGPT creator OpenAI) does, that in this case you aren't really thinking or feeling anything either. Like his computer, you are just regurgitating input to create responses. Or, like Elon Musk, you can argue that reality is probably just a simulation anyway. In both cases, the convenient conclusion is that you can do what you like to the world and other people, because they're only as sentient as your computer – if they exist at all.
And as for magic, the amount of money being thrown around in the AI boom is so vast that anything short of the supernatural would be a disappointment to investors. A week before the talk in the bookshop, Nvidia – the chip manufacturer that is selling the shovels in this particular gold rush – achieved the highest market value in history, at $4trn. Fundamental to this valuation is the idea that machines can see the future.
In the middle of the 16th century, John Dee, the court astronomer and magician to Elizabeth I, carried in his pocket a case made from shark skin. Within the case was a circular mirror made from obsidian – black volcanic glass – that had originally been created for an Aztec priest. To a modern eye it shows nothing but a dull reflection, but to Dr Dee it was said to show angelic communications and predictions. This gave Dee a significant political power.
Most of us carry a similar artefact today, of course. We seek answers and communication from our black rectangles. We choose the news we want to read and the people we want to hear from. We see what we want to see in the black mirror, but Kevin reminded us: 'Whoever controls what is seen in the mirror, in some way, controls the future.' The mirror of ChatGPT is darker and more obscure than obsidian. Its priorities are hidden in code that is not public, and which only a handful of people could understand.
What a tarot reader really does, Bel explains, is to deal out a random sequence of cards and give them the meaning the client is looking for. The cartomancer, the haruspex and the seer are not only predicting the future, they are answering the question of how a person should meet that future. 'The underlying question,' said Bel, 'is always: tell me how to live.' As anyone from the Pope to Mark Zuckerberg will attest, this can be a very profitable question to answer.
Throughout history, people have found that the trust they put in priests and oracles was misplaced. Today, the most popular use for AI chatbots, according to a recent survey, is for therapy; millions of people are entrusting this technology with access to their most personal feelings, allowing it to intervene in their thinking, their ethics and their sense of self. They are imbuing it with what esoterics call egregore – a form of being invoked when lots of people put their faith in something. A new power, no more accountable and no less dangerous than the clerics of the past. I think I'd rather put my faith in a pack of cards.
[See also: No 10 is scrambling for a Silicon Valley payday]
Related