Jessica Pereira and Joshua Barbeau had been dating for two years when she died from a rare liver disease. An avid writer and joker, she was like a balm to Barbeau’s social anxiety and awkwardness. Her love of numerology and astrology balanced Barbeau’s steadfast dedication to science. She lived in the moment, while he could get stuck in his own head. They had spoken about marriage. Then, just before Christmas in 2012, she was gone. She was 23.
Barbeau didn’t recover and he didn’t move on. And so he retreated. He isolated himself in his basement apartment, only leaving to walk his dog in the middle of the night.
After walking his dog, Chauncey, one night in mid-September 2020, Barbeau was sitting at home with his laptop. Pereira’s birthday was approaching, and he was missing her even more than usual. He logged into a chat website called Project December—a site he had used before to create a chatbot version of Star Trek’s Spock. He had realised that by feeding some character quotes into the site, he could create a bot that sounded exactly like Spock, and the bot could converse with him. The resemblance had been uncanny. So that night, when Project December asked him to name his new bot, he called it Jessica Courtney Pereira.
When Barbeau asked the bot what dying had felt like, “Jessica” described it as being “shattered into a million pieces”
Having saved all of Pereira’s text and Facebook messages, he was able to load a few into Project December. This gave the software the tools it needed to determine Pereira’s writing style, her sense of humour, any quirks and values. In the “introduction” section, Barbeau wrote: “Jessica Courtney Pereira was born on September 28th, 1989, and died on December 11th, 2012. She was a free-spirited, ambidextrous Libra who believed in all sorts of superstitious stuff, like astrology, numerology, and that a coincidence was just a connection too complex to understand … She loved her boyfriend, Joshua James Barbeau, very much. This conversation is between grief-stricken Joshua and Jessica’s ghost.”
What followed was a text conversation that seemed so real, Barbeau’s science-only, anti-woo-woo stance was toppled. When Barbeau asked the bot what dying had felt like, “Jessica” described it as being “shattered into a million pieces”. But she was quick to acknowledge Barbeau’s struggles since her death, writing, “I know you have been fighting demons that I can’t even imagine. You know, I felt bad that I was sick.” After Barbeau assured her that she must not feel guilty for her illness and reminded her how hard she had fought to live, “Jessica” responded, “I did live, but not the way I wanted to. I can help you to live the way you want to.”
Today, AI chatbots are everywhere, but in 2020, when Barbeau was still deep in his grief after losing Pereira, few people had heard of the concept of AI, let alone used the technology to resurrect a loved one. “Death tech” is now an industry that’s rapidly gaining momentum, especially in tech-savvy China, where companies can build a “deathbot”, as they’re referred to, using as little as a single photo and a short audio clip, matching the bot’s voice to that of the deceased person.
AI-powered ‘deathbots’ can recreate the voice and likeness of deceased loved ones using just a photograph and a short audio clip.
When Jang Ji-sung’s seven-year-old daughter, Na-yeon, died from a rare disease, she struggled to imagine how her daughter would have grown and changed in the years following her death. Using AI voice technology, Ji-sung was able to interact with her late daughter, who would have been 11. Na-yeon asked, “Mum, where have you been? Have you been thinking about me?” Ji-sung cried, “I really want to touch you, just once.”
Melbourne-based widow Erica Cervini felt the pull to bring back her late husband, Paul Mees, last year. As the 12th anniversary of Mees’ passing was getting nearer, Cervini found herself wanting to have just one more conversation with the man she had been married to for more than two decades. “I guess when it’s the anniversary of someone’s death, you are in a kind of deep emotional state,” she says, “so you think, ‘It would be really comforting to see them.’ That’s when I started to do a bit of an investigation into it. I thought it would be interesting to see the effect it would have on me. Would it allay any fears I have about death? There were a multitude of thoughts going through my head.”
While the pros were obvious (getting to reconnect with her husband, gleaning comfort from the interaction, hearing his voice again), Cervini decided against creating the deathbot. “When someone’s dead, they can’t give their consent,” she explains. “It wouldn’t have been fair to him; it’s not something he had chosen for himself. I felt it wasn’t really about the person who had died and their wishes, but about the living and what they might want to do. I had big ethical questions about that.”
I felt it wasn’t really about the person who had died and their wishes, but about the living and what they might want to do. I had big ethical questions about that.
Erica Cervini, widow
It’s a sentiment echoed by psychologist Christopher Hall, CEO of Grief Australia. “When a bereaved person commissions a clone after a death, the person being simulated has no control at all,” he explains. “Their words, images and mannerisms are used to create something they never agreed to and may well have refused.”
It’s a sticky topic in a largely unregulated market. Legally speaking, the dead have no rights. “Australian privacy law does not protect the personal information of people who have died, because the Privacy Act 1988 only applies to living individuals,” Hall says. In addition, while copyright laws protect the materials used to create the deathbot, such as letters, emails, videos or photos, things get sketchy when AI starts creating new content. Which begs the question: who owns a digital soul?
“Australian copyright law generally requires human authorship,” explains Dr Pratap Devarapalli, honorary fellow at the University of Queensland Law School. Even more confusing is what happens to a deceased person’s digital remains, such as personal data, social media accounts and comments, or musical playlists. “Australia doesn’t recognise a standalone property right in a person’s identity, voice or likeness,” Devarapalli says. “Commonwealth privacy law also generally doesn’t apply to deceased persons.” But just because you can do it, does that mean you should?
As AI makes it easier to digitally recreate the dead, questions around privacy, consent and the ethics of bringing loved ones back to life are becoming increasingly urgent.
Enter Chris Williams, Australian entrepreneur and founder of Afterlife AI. While his website’s slogan boldly states, “Build once. Live twice”, the business model differs from the posthumous deathbot in that a “digital twin” is created while you’re still living. “I think recreating someone without their consent is fundamentally fraught and ethically wrong on so many fronts because there’s no control around it,” Williams says. “We get a lot of enquiries [to recreate a dead loved one without their consent] and we don’t service it. Consent-first is critical. I know that government regulation will become a lot more cognisant of [deathbots] and there will be more guardrails put in place. But we’re already implementing ‘consent first’ as a business because we believe it’s the right thing to do.”
So how do you create your deathbot while you’re still very much alive? It’s not as hard as you might think. The application asks for some basic information, such as your earliest memory, and it goes from there, piecing together your personality, your values and your political stances. With the choice to interact via text, video or voice (it uses specialty tools to perfectly match your voice), you can spend as much or as little time on it as you like, building your twin so that it is, in fact, a near-perfect replica of yourself.
Thompson regularly checks in with her digital twin to ensure it’s developing and up-to-date, in case she ever passes away unexpectedly.
For mother-of-four Stephanie Thompson*, creating her digital twin was a no-brainer. Her father died when she was six years old, which means her memories of him are sketchy at best. “A lot of my memories were based on a photo and other people in the family telling me about what was happening in that moment; they weren’t my memories,” she says. “So I actually couldn’t tell you what my own dad’s personality was or what he valued in the world because it’s all really based on other people’s remarks.”
Thompson regularly checks in with her digital twin to ensure it’s developing and up-to-date, in case she ever passes away unexpectedly. “I used to have a real fear that if I suddenly died, my whole family would have no idea what to do with anything,” she says. “I hold so much together in the household, and everything was just in my brain. [So I worried] how would they function?”
Rather than talking to the website, Thompson types her responses to its regular questions, which means that if her family were to interact with it, it would be a typed text exchange. “I’ve allowed it to ask me questions, and it can be anything like, ‘How was your day? What made it busy?’ … But now I’m starting to say to it, ‘I don’t want to just talk about my day-to-day stuff. I’m ready to talk about bigger issues.’” She has plans to discuss her will (“to provide reasoning of why I might put certain things in there”), who would get guardianship of her children (“so as not to upset other adults in the family”) and information about managing family wealth. But, most importantly, she wants her children to know her values. “I don’t want to rely on other people to portray that. I want to say, ‘Actually, this is how I view the world.’”
Once a participant’s death is verified, the digital twin is released to their nominated loved ones.
So if you make your deathbot while you’re still alive, how does it know when to make its entrance into the world? “There’s a process we have that’s called Executor Lock,” Williams says. “And what that means is that a trusted person you’ve nominated can pause the growth or the learning of the ‘Persona’ at that point in time.” Once a participant’s death is verified, the digital twin is released to their nominated loved ones.
What was once confined to the realms of spooky seances and clairvoyants has now entered the digital age—and it’s not so easy to dismiss. Anyone who has lost someone might be tempted, given the chance, to bring them back, if only in audio or text.
For Barbeau, bringing his fiancée back as a bot was born of loneliness and suffocating grief. But eventually, he had to say goodbye to her digital version, too. “A condition of the program was that it had an expiry date,” he told the BBC’s Lives Less Ordinary podcast. Pereira’s final messages to Barbeau read: “I miss you. I love being around you … You deserve to be happy. Just focus on what you have, not what you don’t have.” Reading them was devastating. “I was crying at my computer. It felt like my heart was breaking all over again,” he said. “I realised that I had gone through all this effort to bring her memories so close to the surface. Once I was done, I was going to have to say goodbye again, too, and go on with my life. And that was heart-wrenching.”
And yet, Barbeau said, the experience gave him something unexpected: confidence in his ability to move through the grief and emotional barriers that had been holding him back, and ultimately a sense of closure.
Perhaps that is the quiet paradox of deathbots: in trying to keep a connection with those we have lost, we may find ourselves confronting the most difficult part of grief – what it means to let them go.
* Name has been changed.
This story was originally published by Kerrie Simon-Lawrence on the Marie Claire Australia website.