We Can't Let AI Dehumanize Our Memory

Using technology to remember everything could undercut what really matters

Memory by Kris Krug

In what turned out to be his final book, Morality: Restoring the Common Good in Divided Times, Rabbi Jonathan Sacks asks readers to consider whether there are “things we can’t or shouldn’t outsource.” Sacks called particular attention to the outsourcing of memory: “smartphones and tablets have developed ever larger memories, while ours and those of our children have become smaller and smaller.”

Sacks also warned of the danger of confusing history with memory:

History is an answer to the question “What happened?” Memory is an answer to the question “Who am I?” History is about facts, memory is about identity.

Recent technological innovations will make the outsourcing of memory to smartphones and tablets the least of our concerns.

In the News

Earlier this month, Jeannine Mancini noted that improvements in artificial intelligence have led people to “explore new ways to keep their loved ones alive—even after they’ve died.” Sometimes referred to as “grief technology,” AI interfaces like chatbots rely on voice and speech patterns (and soon, photographs and videos) to reconstruct artificial versions of loved ones who have died. As Mancini suggests, “This digital twin can then respond to text messages, emails and participate in video calls, mirroring the deceased’s personality traits and communication style.”

The technology has been around for the past few years and is taking several forms. A Washington Post article last year highlighted an effort by Hossein Rahnama, a professor at Toronto Metropolitan University and research affiliate with MIT Media Lab, to preserve the memory of people who are about to die. As the Post notes:

Rahnama receives emails almost weekly from people who are terminally ill, asking if there is a way to conserve their legacy for their loved ones. He said he now has a beta group of 25 people testing his product. His goal is for consumers to one day be able to create their own eternal digital entities.

Commenting on these developments, the Post article suggests: “artificial intelligence and virtual reality, along with other technological advances, have taken us a huge step closer to bringing the dead back to life.”

In My Head

The current and future AI innovations make Sacks’s question about outsourcing memory increasingly urgent. We will soon have the opportunity to interact with AI-generated versions not only of lost loved ones but also of people we’ve never met. The latter could include relatives who died before we knew them and living people (famous or otherwise) whom we don’t know personally. We will create new memories interacting with artificially generated versions of people who aren’t really there.

On one level, our identity shaped by memory already includes asynchronous and imperfect interactions with loved ones and strangers alike. When we look at old pictures or read old letters, we augment and sometimes alter our memories. We arguably do the same when we read biographies of celebrities or other people we’ve never met in real life. Even more abstract ideas conveyed in books and articles can shape how we remember.

If Sacks is right, the more we embrace grief technology, the more we risk confusing identity. It’s one thing to have a box of pictures. But artificial intelligence that generates new and ongoing “conversations” will ultimately supplant memories of loved ones with fantasies that emerge from our longings to be with them. Over time, we may be unable to distinguish memories tied to a past we actually experienced from those corresponding to a past we wish we’d lived.

Stanley Hauerwas, whom I interviewed in last week’s newsletter, has written about the importance of remembering as a moral task. But I think it is only moral if Sacks is right to link memory to identity. If who we are is in part a function of how we remember, then the integrity of our identity depends on the integrity of our memory.

We could, of course, dismiss the importance of the integrity of our identity—we could decide that all that matters is who we are in this moment in time. But if we see our life as a narrative that includes past, present, and future, then remembering rightly—or as rightly as we can—may affect not only who we are but also who we are becoming.

In fact, part of the gift of memory may be our imperfect recall of it. We work to remember because the details of our past experiences and feelings elude us. Outsourcing memory to technology—or having technology create memories for us—ultimately deprives us of the practice of remembering. And at some point, we may no longer remember how to remember. Or, as Hauerwas said to me last week: we may no longer “know the determinative narratives that shape our lives.”

In the World

Writing this post called to mind Miroslav Volf’s powerful book, The End of Memory: Remembering Rightly in a Violent World. Volf examines the ethical dimensions of memory and the dangers of misremembering. Our imperfect recall risks using memory not only as a shield against past harms but also as a sword that conjures overly harsh versions of past wrongdoers. Drawing from some of his own experiences, Volf notes that “when we remember the past, it is not only the past; it breaks into the present and gains a new lease on life.”

Volf also focuses on the importance of remembering rightly. I suggested above that one risk of grief technology is that we will replace actual memories with artificial ones. In doing so, will will not remember rightly. And we may eventually find ourselves further from reality rather than closer to an understanding of who we really are.