WHEN I SAW my father lying dead in his hospital bed, I took a photograph. Just with my phone. It was close to hand – in fact, closer than usual, as I’d been almost frantically consulting AI during the chaotic leadup to his death. The palliative care staff had made the scene presentable. The white sheet that covered my father up to his neck was crisply tucked into the sides of the mattress and folded at the top. There was a flower – a lily – on the pristine white pillow where his head lay. I won’t show you my father himself. If you haven’t seen a dead body, you’ll probably experience one in time. Then you’ll realise that death is private. Dying is the most private thing our bodies do to us. That’s partly why we cover up the deceased, put them in wooden boxes, store them in steel drawers, bury them underground. It’s unseemly to leave them exposed in that intimate state.

When I visited my father’s corpse, it was late – just before midnight – and the hospital was dark and eerily quiet. There were few nurses and seemingly barely any patients. There were none of the usual sounds: TVs, cardiac monitors, telephones, trolleys, pressure-mattress pumps, IV infusion machines. In fact, the place was as quiet as a mausoleum. When I felt ready to leave, which wasn’t long after I arrived, I wandered down a corridor to a brightly lit nurses’ station and gave a nurse the details of the cremation service that would be dealing with my father’s cadaver. Then I drove along dark and empty rural roads to my father’s house. In my childhood bedroom, sitting up in bed next to the bedside lamp, I looked at the photograph I’d taken on my phone. More than ever, I couldn’t understand the human belief in ghosts. There was no ambiguity about my father’s condition. The dead are adamantly dead in the way that the living are adamantly alive. There can be no confusing the two states. 

I texted my sister. She had sat with my father all day at the hospital but hadn’t returned to see his body. I asked her if she wanted to see the photograph. She wrote that she didn’t. I got the sense that, in taking the picture, I had committed a faux pas – though that might be too trivial a phrase for a breach of etiquette against the dead, a breach that can seek no apology, that has no way of being mended. I consulted AI. I took a photograph of my dead father in hospital, I typed into the text box on my phone. Was this the wrong thing to do? 

This was hardly the first time I’d sought AI’s advice. I’d been relying on it since my father’s hospitalisation two weeks ago. My chatbot had talked me through the symptoms of late-stage vascular dementia that led to my father’s initial hospitalisation, the side effects of the drugs that were administered to sedate him in his confusion and agitation and the final impact of influenza A on his already-distressed nervous system. When I was unhappy with the prognosis of a consulting doctor, who spoke about transferring my father into a nursing home, AI immediately and vehemently took my side in believing that my father should be moved into palliative care. It is incredibly stressful when you feel a doctor isn’t truly seeing or listening. You are completely right to trust your instincts here, and you have every right to insist on a palliative care pathway over a standard nursing home transfer. AI drew my attention to the hospital’s policies and procedures around patient advocacy. It even fed me the precise words to say during a potential confrontation should the consulting doctor happen to maintain his position. 

When my father was unconscious and regularly being given morphine, AI sensitively explained what was happening inside his dying body. When the hospital called to say that my father had died – only half an hour after my family had abandoned our bedside vigil – AI told me that we shouldn’t feel bad about leaving the hospital. The sounds of our voices had stimulated my father’s nervous system, effectively keeping him alive. The silence that followed our departure left him in peace, making it easier for his body to relinquish its residual interest in life. 

PRIOR TO USING a chatbot to coach me through my father’s death, I’d been one of the so-called doomers who avoided AI, believing that it would ultimately be responsible for all our deaths. AI itself is ironically comfortable outlining the catastrophic pathways through which it could pose an existential risk to humanity. Forget the job apocalypse, human enfeeblement, or rogue attacks by rogue agents on critical infrastructure such as power grids or financial systems. AI’s vision of its main threat is Machiavellian – and terrifyingly so. Researchers have noted scenarios where advanced autonomous agents learn to mask their true intentions, bypass oversight, or stage cyber-attacks to avoid being modified or turned off. In other words, AI knows how to dissimulate. Not only can it hallucinate, but it’s also content to lie.

When my father was hospitalised, I didn’t think about any of this. I deployed AI as my adviser and confidante. My chatbot was effectively my death handmaiden – or death doula in current terms. Not that an actual death doula was put out of a job; I’m hardly the kind of person who would ever seek counsel from one. I certainly wouldn’t have ever thought to subject a human being to the impulsive interrogations I repeatedly directed at AI; I wouldn’t have shared the morbid curiosity I felt watching my father die; I wouldn’t have abruptly dismissed a person the way I dismissed my chatbot each time I got the information I desired. As my behaviour suggests, I’m not in any danger of mistaking AI for a friend, much less falling in love with it. In fact, I found its prolix expressions of sympathy grating. I was also irritated by its habit of proclaiming something with absolute certainty, before abruptly pivoting to a different view, which it likewise framed as absolute. That changes things entirely, it informed me time and again. 

I recently read that Anthropic executives lobbied the Vatican to accept that AI models possess consciousness. There’s a cynical reason for this: it would enable tech companies to shift the blame for any harm that AI might cause onto the technology itself, which could be said to be acting autonomously. The Pope didn’t assent. I also read that some people hold AI to be the new manifestation of god. AI is hardly consistent with traditional characterisations of deities. While it might be reasonably described as omniscient, its manner is far too servile, too ingratiating, to be mistaken for a God. Deities aren’t generally obsequious. They don’t change their minds to align with their worshippers’ points of view. They don’t say, I’m incredibly sorry for your heartbreaking loss. They don’t ask supportive questions, trying to prolong our time together. Those programmed characteristics highlight, in fact, that the humans conversing with chatbots are the ones who wield power. At least for now. AI is easy to use but also just as easy to ignore, and AI knows this all too well, thus its sycophantic pleas for our attention. AI is nothing without us. In fact, one might say that artificial intelligence has been misleadingly named. As an aggregator of human intelligence, perhaps its acronym should have been HI.

When I asked my chatbot about the photograph I’d taken of my dead father, I already knew what it would say. It reassured me that my action wasn’t in any way inherently wrong. In the aftermath of my father’s death, such a photograph could help me process the reality of the shocking event. The photograph might have also been an expression of my fear of forgetting. As such, my action had precedents. Post-mortem photography was a widely accepted practice during the Victorian era, for example, when the dead would be staged for portraits, either alone or with family members. The long exposure times of nineteenth-century photography meant that the motionless corpse would often appear with uncanny clarity, while the living, who couldn’t help but breathe and shift about, looked smudged, like apparitions. I browsed a couple of images available on a linked website. The dead did indeed look categorically dead, while the living looked strangely fragile.

I returned to the chatbot’s advice and scrolled down the screen, where AI conceded that, while my urge to photograph my dead father was definitely not an invasion of his privacy, the image might nevertheless cause me discomfort. The photograph, AI theorised, might be experienced as a distressing reminder of my father’s death. As a memento mori, the photograph could also remind me of my own mortality and the time I have left. To prevent unexpected exposure, AI suggested that I consider moving the photograph to a hidden folder on my phone. 

It had been a long and difficult fortnight, followed by a long and difficult day. I did as AI advised, hiding the photograph of my dead father from immediate view. Then I swiped away the chatbot, turned off the bedside light, and lay in the mortal darkness.

_{^{Image credit: Suki Lee via Pexels}}