Tomorrow I’m going to visit Las Meninas at the Prado in Madrid, and I hope to learn something about how we are designing AI machines. How can a painting from 1656 and a technology from the twenty-first century have anything in common? Well, in a sense, both address the problem of subjective and objective reality, perspectives on the world and on memory. Diego Velazquez would have been an outstanding AI ethicist!

Artificial Intelligence (AI) machines construct models of the real world within which to consider and address problems. They observe, primarily through image and sound recognition, and use those observations to build models of the world. Yet their models are devoid of a subject, of a unique value model within which to contextualise their observations (as we humans possess) save that of the AI designers, though even that will itself be aged and imperfect at the point of decision. Is the way in which AI views the world somehow objective? Is it shorn of the poverty of subjective judgement? Or does that deprivation somehow render its observations meaningless?

Las Meninas is a painting about representation, art, memory, and who we are. At first glance it is merely a court scene with a painter and some people from the court in a posh room. But then you realise that the painter is looking straight at you, the viewer. Others in the painting similarly have their gaze fixed upon you. The mirror in the back of the room – with the King and Queen of Spain – is that reflecting the canvas, whose back is to you, or is it reflecting you? Are you in the place of the King and Queen? The primary question this painting seems to ask is what is painting for? It is also asking a question about the nature of reality, a questioning of subject-object relationships. The King and Queen are reflected in the mirror, painted on the unseen canvas, and perhaps most real of all they are in the gaze of most of the characters, even if unseen by the observer. This could be a painting of a painting, or it could be a representation of a representation. It could in a sense be a painting about what is not on the canvas – and in that, it could achieve a kind of supernatural quality, if we take the word natural to represent our scientific understanding of the world.

An old philosophy professor of mine was fond of a singular piece of wisdom to young students – ‘read old books’, he would say. There’s something about persistent memory, things that we as societies have managed to remember in spite of our profligacy with intellectual inheritance. Insights and inspirations, revelations and astonishments lose their lustre over time; gravity seems today like a boring physics equation notwithstanding innovations in quantum mechanics; but in Newton’s time it was genuinely brilliant. And so we should remember these ideas, and value them, and not take them for granted. Ideas in art, too, were at times outrageous, offensive, and blasphemous. In light of my recent research into political theology, that last pejorative holds particular merit: for blasphemy in centuries past was as much a slight against political order and scientific orthodoxy as it was offensive to God. New ways of seeing the world meant change, and if not quite disorder, then at least some novelty.

Our modernist understanding of the world in many ways persists as it would appear to have done for centuries. Our theologies may have transformed from religious to scientific to political to economic, but the core theology has persisted: a belief in the real, the observed, the ‘known’. Occasionally we contemplate zones beyond our existence, and give them some credibility, in order that we might explore the boundaries of humanity. Quantum mechanics allows a glimpse; art has consistently tried to do this since the late nineteenth century; most organised religions have been too politicised, but some eastern versions of Buddhism (and perhaps Thosophy) have tried to develop the idea. AI may actually be a vehicle for inadvertent advances, with such a reduced dependency on human discipline, and mortal memory. And who knows – perhaps an AI version of Las Meninas may not be too far away – and it could be better.