Cognitive Insecurity
Will AI replace our memories with synthetic fabrications?
We all know AI is coming for our jobs, and our free time, and our search results, and our cover letters, and our laborious coding tasks, and certainly for our marketing (budgets and content). But because all that exists outside of my head, out in the world, I don’t worry too much about how it might alter my fundamental perception of reality.
Unless of course AI ruins my brain.
No, I’m not specifically talking about the MIT study that shows we’re already indumbnifying ourselves by outsourcing formerly brain-powered activities to machines. Though that’s not exactly great news either.
I’m talking about deep, core elements of our identities. Who we are, what shapes us into the beings we be. Our innermost thoughts and ideas and, crucially, our memories.
Remember as a kid how your grandfather used to take you boating on the river? No? Well that’s my memory, so it’s fine that you don’t share it. But you can have it too, if you want it.
Maybe your memories are of playing stickball in the street, or hanging out in the park with your friends, or just being alone watching TV in the dark. Whatever they are, good or bad, they’re your memories. And they contribute to who you are, to why you are the way you are. Maybe a lot, maybe a little—I’m not solving nature vs. nurture here—but I think it’s safe to say our experiences, and our understanding of those experiences, are a fundamental part of our being. Thankfully nothing can take that away from us.
Except AI.
We like to pretend otherwise, but human memory is demonstrably fallible. True crime TV has taught me that eyewitness accounts are notoriously incorrect, and that’s based in fact. The same event witnessed by two people is often interpreted quite differently.
One high profile example of this is what happened to former NBC Nightly News anchor Brian Williams. Williams was fired in 2015 for fabricating a story about being in a helicopter when it was hit by a rocket-propelled grenade. The implication was that he lied, but in fact it appears that his brain betrayed him, conflating details of an oft-repeated story and creating a false memory. He was simply incorrect. He repeated the real version of the story for more than a decade before it eventually escaped containment and grew into a full fledged vibrant experience he was sure, incredibly and incorrectly, that he had.
Ever see two old friends argue about a memory? (“That didn’t happen to you, that’s my story!”) One of them is wrong, unintentionally, because the other’s story planted itself so vividly in the brain as to become real. A real memory of a real experience, all because the story was oft told. Or, really, because memory is fragile.
Photographs play a tremendous role in supporting our fragile memories. Family albums are filled with reminders of the faces of old friends, the apartments we lived in, the places we went. This evidentiary nature of photography is one of its great powers. The greatest, perhaps.
Imagine it’s a rainy day and you’ve got time on your hands so you’re looking through old photos. You’re moving them from here to there, organizing and saving and backing up. Your online identity somehow catches wind of it, and the next thing you know your social apps are serving pertinent ads for apps that help organize/save/backup old photos.1
It being 2025, it isn’t long before the algorithm serves up an ad for an AI powered app. This one not only organizes/saves/backs up, it can also bring old photos to life through the magic of interpolation, animating those crusty old photos into exciting AI videos.
So on a whim you start plugging old photos into Dreamlux and Runway and the like, and out comes an animation of your long lost loved one moving and speaking and hugging you, right there, live and in person. You can see it with your own eyes, and every fiber of you feels it to be real. But it’s not — at least not in the same way that photographs and videos typically are. But it sure feels like it.
There is value in this, no doubt. I assume the ability to see someone from a still photograph come to life, to move and breathe and smile and interact with the you in the picture… That has the potential to be an invaluable gift.
Reddit founder Alexis Ohanian recently reported on his own experience with this. He dropped an image of him and his mother from 30+ years ago into Midjourney. It animated their embrace, and he saw a loved one come to life.
I have no doubt this is a profoundly moving experience. And I am not equipped to know whether or not it is “worth it.” But the possible side effects strike me as terrifying.
What about the uncanny valley? What about the fact that AI is interpolating data, filling in where no actual information exists? It means the machine is making up the shape of your mother’s face, or the sound of her voice.
You had some recollection, back in the recesses of your brain, what her actual voice sounded like. And the shape of her smile, or the way she would arch her eyebrow. But now new information has been added to your mental archive, modified by the machine. One pixel of misinformation and the whole lot is polluted. Now, to you, the shape of your mother’s face and the sound of her voice are slightly different, bent by what the machine made up, no longer purely what actually was.
I don’t think this is better. I don’t know the answer, but I think we would do well to tread lightly.
As someone who finds context crucial and the power in photographs and videos to stem from the fact that they are (typically) based in reality… New input — the kind that’s close but not quite right — is damaging the signal-to-noise ratio. It may be beautiful, profoundly meaningful noise that brings momentary joy. But at what cost?
If it means I lose the authenticity in my memories, it doesn’t seem worth it. No matter how fallible those memories may be, they’re all I’ve got.
It’s sort of like when you’re furniture shopping and you get ads for couches immediately after purchasing a couch. If there’s a time in my life I’m least likely to purchase a couch, it’s in the hours and days immediately following the purchase of a couch.
Animating a memory with A.I. is a lie. (like the hesitant look in the lady with the child being turned into a smile) It's creating a false memory. Great way to surrender ones life to the new Master (A.I.) and weaken ones own emotional core.
I totally get your concern about AI potentially distorting memories—especially when apps animate old photos of loved ones. A flawed interpolation could easily imprint an expression or posture that never existed, and yeah, once you see it, that version might stick in your mind more than the original. It’s like a false memory wrapped in nostalgia tech.
That said, I think we’re heading into a very different kind of terrain—where memories aren’t just reconstructed from fragments like photos or home videos, but where AI can synthesize something much more complete. Think of all the digital traces we leave behind: texts, emails, voice recordings, video clips, journals. Layered together, they can form the foundation of a fairly realistic facsimile—something that can evolve over time and even be tuned to reflect different periods or dynamics in your relationship.
It’s not about reanimating someone with eerie accuracy, but maybe about creating something useful. Something therapeutic. Imagine a bot that understands both your loved one and your relationship with them—your regrets, your questions, the gaps you never got to close. Suddenly you might get a second chance at a conversation you never had, shaped by everything that person ever wrote, recorded, or shared—and everything you bring to it.
Of course, before all that gets serious, you’d probably give your grandparents Mohawks or make your dad do a TikTok dance. That’s just part of the play phase before the tech matures into something deeper.
It reminds me of journalism, oddly enough—where the moment you try to be completely objective, you reveal your subjectivity through your choices. In that way, your personal AI, trained on you and your memories, won’t be neutral. But maybe that’s the point. It will know you, and it will adapt its tone, delivery, even its philosophy based on how you think and feel.
Now take it a step further. What happens when enough people start opting in—volunteering DNA, metadata, life histories—and AI begins suggesting optimal paths for their lives? Education, love, health, meaning. At some point, it won’t just be about remembering the past; it’ll be about refining the future.
We’ll have to decide how much we want to outsource, how much of ourselves we want reflected back to us—and how much room we leave for surprise, imperfection, and growth. But I think it’s coming, whether we’re ready or not.