What can robots teach us about love? How would we teach a robot what love is? Reading Jeanette Winterson’s book, 12 Bytes, last week awakened a new way of thinking about the world for me. ‘I love, therefore I am’ is the title of the final essay in the book in which she says all of the technology and tools humanity has acquired is now missing only one vital ingredient to enable us to build a fair and kind new world: love. It is the highest human value, revered in every religion, widely believed to be the centre of what makes humans— humans. Thinking, as Winterson explains, will no longer be the preserve of humans – AI will be so, so much quicker than we are at thinking. But love is something we, as yet, can’t teach or programme.

I read an article recently in Wired magazine which talked about AI nannies. In a not-too-distant world where children are cared for by machines, the writer says that children will, naturally, love their caregivers. What would happen if the AI are more consistent and patient than humans are, if they give children the feeling of safety and nonjudgment they require, without ever needing to consider their own needs like human carers have to? Could they ultimately love their faithful machines more than they love humans? Could they crave machine bonds in adulthood, to replicate those early connections? In this way, love could mean something different when between humans and machines – perhaps we need a new word for it.

What does ‘love’ mean? Is the truest form of love more about ‘wiping someone’s arse’ as Mrs. Elton says in Terence Rattigan’s ‘The Deep Blue Sea’ than that much-exalted feeling that is more eros than agape? Perhaps love, like character, is shown in action. Don’t we know best who loves us when we really need help? Is that when love counts most? By this reckoning, love isn’t declarations on balconies and poetry. Or at least, this isn’t, ultimately, what matters about love or, rather, what makes love matter. It’s in the thousand quotidian acts of devotion that love is tested and shown. The notion of love as a verb, as a duty, as a pledge that’s daily renewed through action is surely what AI could come to enact. This kind of love, this kind of care, is mostly the unending, often unnoticed work of women and it makes the world go round. In Katrine Marcal’s brilliant book, ‘Who Cooked Adam Smith’s Dinner?’ she shows how women hold up the capitalist system, through love. Adam Smith writes about the ‘invisible hand’ of capitalism, where participation in consumerism spreads wealth, but he fails to notice the unpaid labour of women (for example, his own mother who loved and cared for him, cooking and cleaning for him, enabling him to be the ‘father of economics’) that is its bedrock. It is this undervalued, unglamorous kind of love, the love that is hard work, that is women’s love, motherlove, that is the love we need the most. 

The lines of U A Fanthorpe’s much-loved poem ‘Atlas’ come to mind now: ‘there is a kind of love called maintenance, which stores the WD40 and knows when to use it.’ Could an AI learn those acts of love? Yes, perhaps they could. Perhaps this poem could be used to help programme AI to love, to teach them that these are the kinds of actions a loving partner might do. Love is work, a to-do list, a pile to get through. But there are other ways that love operates for humans. My favourite of the ‘Love Poems from Me to You’ that U A Fanthorpe and Rosie Bailey wrote to each to each other is ‘Idyll’, which strikes a different note. In this vision of love, love is a place with whisky and pets and poems to be read. It’s about reaching a state of complete comfort and joy with another, so that you never tire of their company. This is love as a heavenly state to exist within, with another; the poet’s way of articulating what being with someone you are in love with feels like, through the metaphor of a perfect place. Language, the tool we use to shape, influence (and, yes, programme) one another falls down when we try to atomise and articulate how love feels. Like in Carol Ann Duffy’s poem,’Words, Wide Night’, poetry can only attempt to express love, to endlessly say, ‘this is what it’s like, or this is what it’s like in words’. We can only ever say what love is like in words, which is still not the same thing as saying what the experience feels like. Yes, all words are signifiers for the signified, but with love the signifiers are harder to wrangle and find. Language misbehaves when love’s involved, meaning shifts, chaos ensues. Being in love is alchemistic, it ‘spins straw from gold’, to quote Duffy again, it can ‘make one little room an everywhere’ for John Donne. 

The magical power of love means that it has always already transformed anything, everything – a crisp packet, rain, an earring, a slice of beetroot, a word, the curve of a window ledge, a breath, a name (especially a name) into a totem, a talisman, a portal. So you don’t just see that scarf, smell that orange, hear the sounds that make her name as they were before you fell in love with her. All is imbued with new meaning. How could you teach an AI to feel a delicious, thrilling tenderness about anything in its field of sensory experience? Falling in love, like entering into any new religious faith, involves overriding all programming and grasping for this new plan, this new chance to be someone else through someone else, this glistening mirage in the distance. It’s often a chimera, often always the thing out of reach.

Could an AI ever be programmed to be undone in this way, to lose its own way, in the way humans so often do when we forget who we really are, ‘the one we have ignored for another’, as in Walcott’s poem ‘Love After Love’? What would be the point, anyway? It’s not the kind of love that combs for nits and remembers birthdays, that keeps the cogs of society turning. But it is the kind of love that teaches each of us a deep self-knowledge and, eventually, a chance to ‘love again the stranger who has loved you all along.’ The examined life, self-knowledge, self-acceptance- these are the hard won spoils of heartbreak.

If we could teach this feeling and we wanted to teach AI the true reality of love, grief, loss so that they could suffer and learn as we do, how would we ever be able to test what they were experiencing? As Winterson says in the opening of ‘Written on the Body’: ‘ why is the measure of love, loss?’ Perhaps it’s only in the losing of a love object that we’d be able to see if an AI ever loved someone; do they, would they, could they remember (really re-member that person through thought), reminisce, desire and dream of reconnection?

Beyond the visible acts of care and kindness, how can love be apparent? How can we know love exists? Just as consciousness is not measurable (Prof. Stuart Russell says in his recent Reith lectures that his cellphone could be considered animate), evidence of feelings would be very difficult to establish. But isn’t this also true for humans? I read a book about a year ago called Love and Other Thought Experiments by Sophie Ward. The first chapter was about two women in love. One of them, Eliza, wakes in the night and is convinced that she was an ant in her brain and she demands that her partner, Rachel, take her word for it. This covenant is needed as the foundation of their continued loving commitment. Rachel can never know if there really is an ant in her wife’s brain. A human can never know if another human really feels love in the same way she does. It’s always leap of faith, and it would be the same with love in robots.

**This essay originally appeared on Natalie Cotterill’s Substack