We Are Going to Remember People With AI. The Question Is How.

This weekend, The Wall Street Journal published a fascinating story with an intentionally provocative headline: “A Dead Father, Reincarnated With AI: Is This How We Will Remember Loved Ones Now?”

The story is about Brian Dema, whose father, J. Richard Dema, passed four years ago. Richard left behind an extraordinary amount of material, including roughly 17,500 pages of writings, along with recordings, videos, business philosophy and other pieces of his life. Brian and a team of collaborators used that material to create “RichieBot,” an AI experience designed to reflect his father’s personality, memories and voice.

As someone who spends a lot of time thinking about memory, storytelling and how technology can help us preserve the people we love, I found the story fascinating. But what stayed with me most wasn’t the technology. It was the question in the headline: Is this how we will remember the people we love?

Increasingly, I think the answer is yes. But I also think we need to be incredibly careful about what we mean when we say remember.

There is a big difference between using artificial intelligence to preserve more of someone and using it to pretend that person never left.  At Reflekta, it is something we think about constantly.

We Have Always Tried to Hold On

Human beings have always found ways to carry the people we love forward. We keep photographs in drawers and on our phones. We save letters. We replay old voicemails. We tell the same family stories over and over, sometimes to the point where everyone at the table knows exactly how they end.

We inherit recipes, expressions, traditions, gestures and pieces of advice. We name children after grandparents. We go back to places because someone we loved once stood there. None of those things erase loss, and they aren't supposed to. They are simply ways of saying that this person was here, that they mattered, and that some part of them is still being carried forward.

AI doesn't change that very human instinct. What it changes is how much of a person we may be able to preserve.

For most of history, an enormous amount of a human life simply disappeared when that person passed. Maybe we were left with photographs, some letters, a few home videos. If we were lucky, perhaps there was a recording of their voice. But so much of what actually made someone them was incredibly fragile.

Their laugh. Their phrasing. The story about how they met your grandmother. The advice they gave when you were scared. The way they remembered their childhood home. The little family story everyone assumed somebody else would remember.

Those things tend to disappear quietly, and usually we don't realize what we've lost until we want to ask one more question.

Increasingly, that doesn't have to happen.

Preservation Is Not Resurrection

This is where I think the language we use becomes important.

Words like “reincarnation,” “digital ghost” and “bringing someone back” make for powerful headlines, but I worry that they can blur the line between what this technology is and what it isn't.

An AI representation of a person is not that person. It doesn't contain their consciousness. It hasn't reversed loss, and it shouldn't pretend it has. It cannot create a new authentic memory with someone who is no longer here.

What it can do is help us interact with what that person actually left behind, their stories, their recordings, their memories, their words, their photographs and the parts of themselves they chose to preserve.

To me, that is a much more interesting idea anyway.

Imagine a family archive containing hundreds of photographs, recorded conversations, letters, stories and videos. Traditionally, somebody might look through those things every few years before they eventually wind up on a hard drive, in a box in the basement or tucked away in an attic.

AI has the potential to make that archive conversational.

A grandchild could ask what school was like when Grandpa was ten. They could ask how Grandma met Grandpa, whether Dad was nervous the day he became a father, or what Mom believed mattered most in life.

The important thing is that the machine shouldn't be inventing those answers. Its job should be to help uncover and organize answers that were already there.

The Technology Shouldn't Be the Main Character

Brian Dema has described RichieBot as part of a larger experiment around a question I think is incredibly important: as AI expands what humans can create, how do we preserve what makes us human?

That question resonates with me because it is easy, especially right now, to make the technology itself the story. We talk about models, avatars, voice cloning, agents and increasingly powerful AI systems. All of that is exciting, but I don't think any of it should be the main character in this particular story.

The person should be the main character.

The technology should ultimately fade into the background, and the human story should remain.

That philosophy guides a lot of what we are building at Reflekta. We believe people should have agency over their stories, their likeness, their voice and their memories. They should understand what is being preserved, how it will be used, who will have access to it, and how it can be changed or removed.

Consent can't simply be a box somebody checked years ago and forgot about. Families also need to understand where the boundaries are between something a person actually said and something an AI system may be inferring.

Most of all, I think the purpose should always be connection, not illusion.

That will matter even more as this technology becomes easier to use. What currently requires thousands of documents, technical expertise and a significant amount of work will eventually become available to almost anyone. When that happens, these won't just be technical questions. They will be questions millions of families are going to have to think about.

So I think we should start thinking about them now.

The Hard Questions Are Part of the Work

The Journal story raises understandable questions about grief, emotional attachment and the effect these kinds of experiences could have on different people. I don't think those questions are reasons to walk away from the technology. I think they are reasons to build it carefully.

We should be asking whether someone consented to being represented this way. We should ask what source material an AI is using, who controls the resulting representation, who owns the underlying material, whether a family can correct something that is wrong, whether they can delete it entirely, and whether someone can decide during their lifetime that they simply don't want anything like this created.

We should also be asking a harder, more emotional question: Are we helping someone remember a person more deeply, or are we encouraging them to believe that person somehow never really left?

Those questions don't have easy answers. They probably shouldn't.

They are not just engineering questions. They are human questions, which means the answers are going to require something more than better technology.

We Shouldn't Need 17,500 Pages

Richard Dema left behind roughly 17,500 pages of material. That is extraordinary. Most of us won't leave anything close to that behind. Our parents probably won't either, and most of our grandparents certainly didn't.

In some ways, that's precisely why I think this technology has so much potential.

You shouldn't need to leave behind thousands of pages for your grandchildren to know something meaningful about you. You shouldn't need to be famous, unusually prolific or meticulously organized for your stories to survive.

We can simply start asking the questions now.

Ask Dad about his first job. Ask Mom what she remembers about becoming a mother. Ask Grandma what her childhood home looked like. Ask Grandpa what he was afraid of when he was twenty years old. Ask about the family vacation everybody still laughs about, or the recipe nobody ever bothered to write down.

Ask the questions we always assume there will be time to ask.

That, to me, is where the extraordinary possibility of this technology lives. Not in resurrection, not in immortality, and certainly not in pretending technology can solve one of the oldest and hardest parts of being human.

It is something much simpler, and I think much more meaningful: remembering better.

For thousands of years, memory has traveled from one generation to another largely through storytelling. Someone spoke, somebody else listened, and if we were lucky, the story survived long enough to be told again.

AI gives us the chance to extend that chain in ways that would have been almost unimaginable even a few years ago. We should approach that possibility carefully, and with a great deal of humility, but I don't think we should be afraid of it.

The goal isn't to make the people we love live forever. It is to make sure that more of who they were, what they believed, what they experienced and what they wanted us to remember doesn't simply disappear.