Resurrecting The Dead Using Generative AI Will Put A Burden On The Living

If data is available, one may think it is possible to keep dead loved ones alive. It is possible to build a chatbot of a deceased person that is more believable using ChatGPT and other potent big language models.

But doing so disregards the enormous amounts of work required to keep the deceased online, especially in light of limited resources and inevitable degradation.

As evidenced by the overworked and underpaid annotators and content moderators driving generative AI, someone is always required to perform the difficult task of maintaining automated systems. This is also true when it comes to copies of the deceased. Digital death care techniques demand a lot of maintenance, from navigating a steadily deteriorating inherited smart home to administering a digital estate after collecting passwords and account information. To preserve digital heritage for future generations, content creators rely on the backend labor of caregivers and a network of both human and nonhuman entities, from specialized operating systems and devices to server farms. It takes effort, time, and resources to update formats and make electronic documents searchable, useable, and accessible. This is a dilemma for institutions and archivists, as well as for individuals who might desire to save their deceased relatives’ digital possessions.

And despite all of this work, websites, formats, and other technologies nonetheless expire, just like weak individuals do. Despite the dream of a self-running automated home or a website that can last for millennia, intentional obsolescence ensures that these systems will inevitably deteriorate. There is a big contrast between what individuals think they desire or what they want others to do and the reality of what it means to support technologies over time, as those tasked with keeping the digital possessions of deceased loved ones can confirm. These systems will eventually fail because to the mortality of both people and technology.

This is undoubtedly supported by early attempts to build AI-powered copies of deceased humans. Scottsdale, Arizona-based Intellitar’s Virtual Eternity, which debuted in 2008, uses pictures and voice patterns to replicate a human’s personality, either standing in for someone at a business meeting or conversing with bereaved family members after someone’s passing. An Intellitar reviewer for CNET called it the product “most likely to make children cry.” However, the company’s website vanished shortly after it filed for bankruptcy in 2012. When technology permits it, the LifeNaut project, funded by the transhumanist group Terasem, which is also known for developing BINA48, a robotic replica of Bina Aspen, the founder of Terasem, will allegedly combine genetic and biometric data with personal datastreams to simulate a fully-fledged human being. However, as the project’s website still uses an outdated version of Flash, it is likely still a long way off from becoming a reality and will require modifications along the road.

There is speculation that using generative AI, we may be able to produce even more lifelike replicas of humans, even those who have passed away. However, doing so costs an enormous amount of resources, including raw materials, water, and energy, demonstrating the foolishness of keeping dead people’s chatbots in the face of catastrophic climate change. Additionally, ChatGPT allegedly costs $700,000 per day to maintain and will put OpenAI out of business by 2024. This is not a viable plan for eternal life.

Another concern is who should have the power to produce these copies in the first place: a close relative, a job employer, or a business? It’s unlikely that everyone would choose to reincarnate as a chatbot. Journalist Jason Fagone tells the tale of Joshua Barbeau, who used OpenAI’s GPT-3 to create a chatbot representation of his long-deceased fiancée Jessica, in a 2021 article for the San Francisco Chronicle. He used it as a coping mechanism for death and sadness, but it also kept him committed in a passionate romance with a deceased person. Additionally, Jessica’s other family members decided not to engage with the chatbot because they felt that this was not how they would like to remember her.

But in the end, chatbots’ longevity is up to developers and businesses, not friends and family. In order to reduce operational costs, the experimental program that was used to build Jessica Bot purposefully integrated mortality into the system. Barbeau had to be wise with his time with Jessica and quit communicating with the chatbot before its battery was completely depleted in order to avoid having to face another type of death head-on.

Developers may wish to think that chatbots can function perpetually in some situations, allowing the dead to completely replace real workers. For instance, the SAG-AFTRA strike raises concerns about the personality rights of the deceased due to the possibility of using generative AI to bring back deceased performers. Similar problems have arisen with different deepfake representations of deceased celebrities, including the contentious use of Anthony Bourdain’s deepfake voice in a documentary about his life and, in a less dramatic, low-tech case, a professor’s recorded lectures allowing him to continue working at a university after his death. But in addition to possibly affecting living employees, such recreations can also annoy the surviving relatives of the deceased and provide them new administrative hassles as they try to take control of the reproductions of their deceased family members. It is possible to build AI replicas of people without their agreement or knowledge.

AI’s ability to bring the dead back to life reveals the networks of labor, infrastructure, and power that underpin all digital creation. We can make exact replicas of the deceased, but in order to survive, these reproductions depend on the efforts of the living: As is the case with digital legacies in general, as websites and businesses vanish, someone has to pay for domain names, clear spam, respond to emails from inactive accounts, change file formats, or transfer things from one hard drive to another. The psychological toll may be too great for those left behind to care for these creations indefinitely.

Resurrecting The Dead Using Generative AI Will Put A Burden On The Living

If data is available, one may think it is possible to keep dead loved ones alive. It is possible to build a chatbot of a deceased person that is more believable using ChatGPT and other potent big language models.

Download The Radiant App To Start Watching!