Summary: A recent article examines how relational AI is changing how we communicate with the dying, from virtual reality reunions to lifelike online avatars. These “generative spirits” have a history that goes way beyond pre-recorded memorials and can recall, plan, and also progress.
While the tech offers promise for convenience, creativity, and historic preservation, ethical concerns surround grief, consent, and online identity after death are also raised. Experts are urging society to start significant discussions about how to properly shape this emerging frontier.
Important Information
- Relational Ghosts: AI avatars can now adapt and respond to real-time simulations of live interactions with deceased people.
- True Use Cases: Examples range from rehashed parents who are grieving in VR to belated court hearings and music videos from famous people.
- Social Issues: The rise of artificial intelligence afterlife raises issues with digital legacy ownership, grief processing, and consent.
University of Colorado
A grieving family named Jang Ji-Sun, who had passed away three years earlier from a rare heart disease, donned a virtual reality helmet and was immediately transported to a grassy area where she spent ten hours playing with an AI type of her child Na Yeon, who had died three years earlier.
The heartbreaking meeting, which was viewed more than 36 million times on YouTube, gave a compelling glimpse into how technology may one day alter how we interact with the deceased.  ,
That day is these, according to new CU Boulder research, thanks to the development of relational AI systems like ChatGPT and the introduction of Artificial “agents” created to work freely on behalf of their creators. And the options are even more limitless than some people had guessed.
In a new report titled” Generative Ghosts: Anticipating Benefits and Risks of AI Afterlifes,” professor of data science Jed Brubaker writes,” We anticipate that within our life it may become common for people to create custom AI providers to communicate with loved ones and the broader earth after their deaths.
Brubaker has spent the majority of his job at the crossroads of systems and death. His study served as the inspiration for Facebook’s Legacy Contact, a feature that lets system users designate a successor to a deceased user to maintain their account. He founded the first Digital Legacy Clinic in the country in November, which assists people in organizing their electric affairs.
For his most recent report, which he co-authored with Google DeepMind scientist Meredith Ringel Morris, he attempted to track down what has been done and what is coming in the budding” AI afterlives” space. In addition, Brubaker and his individuals have begun beta testing their own” AI spirits” and conducting research to see how people feel about them in his college laboratory.
You might interact with a Facebook memorial page for your grandfather after he passes away, he says. What would it be like to actually sit down with grandpa by the fire and talk to him?
That day might not be far off.
From text-based grief bots to resurrected celebrities
According to Brubaker, tech-savvy futurists have experimented with AI afterlife for a long time.  ,
After Lou Reed, the frontman of Velvet Underground, passed away in 2013, Laurie Anderson and his partner worked with machine learning experts to create a text-based chatbot that she could use to converse with. She still frequently uses it.
Anderson recently told The Guardian,” I am totally, 100 % addicted to this.  ,
The Beatles ‘ surviving members used AI to create the new song” Now and Then” that featured the voice of John Lennon, who had passed away, singing along with his bandmates in 2023.  ,
A man was shot dead in a road rage incident just last month, and his family used AI to create a life-like version of him. The avatar forgave his killer during a heartfelt courtroom video.
In the meantime, numerous startups now offer assistance to the living in creating posthumous digital versions of themselves: Re, memory will create a “highly realistic AI avatar” to leave behind for family members after a lengthy 3D video and interview session. An AI app called Hereafter invites users to record audio stories that the “virtual you” can share after your death.
This entire arrangement seems incredibly creepy to some.
However, Brubaker points out that online memorials, which were once thought to be creepy, are now ubiquitous and that photos were once believed to be thieves.
What’s creepy often becomes commonplace, he says.” After time, what’s creepy often becomes commonplace.
The development of generative ghosts
What’s in store for Brubaker and his co-author, who refer to them as “generative ghosts,” is the most intriguing aspect of what lies ahead.
They can do much more than recite old stories that the once-living fed to them. They can do this thanks to large language models that can generate and understand human language, and other features that enable them to remember, plan, and exhibit other complex human behaviors.  ,
For instance, they could talk with their children about recent events that have occurred after their passing, write a new song or poem ( from which their family might receive royalties ), or even assist them with managing their estate.  ,
Most “generative ghosts” are currently text-based and rudimentary. But ultimately, according to Brubaker, we might be able to have a direct conversation with grandpa by the fire.
Instead of just reading you some pre-scripted words, you could engage in genuine conversation with this extremely high-fidelity, interactive memorial.
Promise and danger
Brubaker also recalls a time when someone who is grieving for a long time over a lost loved one might benefit from generative ghosts.
This was Jang Ji-Sung’s heart-wrenching reunion with her deceased daughter, which was actually the catalyst for. ( She collaborated with a South Korean TV network to create a 3D version of Nayeon so she could bid a final farewell after three years of battling mental health issues. )  ,
Generative ghosts could also be found in historical exhibits.
Museums are attempting to think of rich, interactive ways to keep their stories alive, according to Brubaker, “because the last generation of Holocaust survivors will not be with us for much longer.
Along with such promise comes risk, of course.
How long should someone spend with an AI ghost before it becomes unhealthy? What or what role should they play in the courtroom? What happens when they are accidentally created ( when someone creates an AI “agent” to help them with other tasks but unavoidably passes away )? How can I be certain that no one will deceive me against my will?
And how and when should a generative ghost pass away?
Brubaker is unable to provide the answers. But he hopes that his study will prompt thought among tech companies and policymakers.
As we advance in this AI world, he says,” What’s possible and what will actually happen are two different things.” When it comes to AI afterlifes,” We hope to see things advance in the most ethical, thoughtful, and sensitive way possible.”
About this news about AI research and grief
Author: Lisa Marshall
Source: University of Colorado
Contact: Lisa Marshall – University of Colorado
Image: The image is credited to Neuroscience News
Original research: Free of charge.
Jed Brubaker and al.'” Generative Ghosts: Anticipating Benefits and Risks of AI Afterlifes.” Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems ( CHI’25 )
Abstract
Generative Ghosts: Considering the Afterlife of AI
As AI systems expand in both breadth and performance, they become more and more capable of producing more powerful and realistic agents, with the potential for agents to be based on particular individuals.
We anticipate that people will develop custom AI agents to communicate with loved ones and/or the rest of the world after passing away. In fact, the last year has seen a boom in startups claiming to offer these services.
We refer to these as generative ghosts because they will be able to produce novel content rather than simply parroting the content their creators produce while they are still alive.
In this paper, we examine the development of technologies for AI afterlifes, including current attempts by individual enthusiasts and start-up businesses to create generative ghosts.
We then introduce a novel design space with examples of generative ghosts ‘ potential applications.
We apply this analytic framework to support our discussion of the potential positive and negative effects on people and society of the various ways that different methods of creating generative ghosts are applied.
In light of these considerations, we develop a research agenda for the AI and HCI research communities to better understand the risks/benefits of this novel technology and help people who want to create and interact with AI afterlife to do so in a beneficial manner.
