You can virtually bring back the dead with AI. Should you?

9 hours ago7 min read5 comments

The black-and-white photograph of my grandfather Max was a static relic of the 1940s, a frozen moment of a man in his army uniform. With a few clicks inside Google’s Veo 3 video generator, that relic was shattered.The image bloomed into color, and then, with an uncanny and deliberate grace, he stepped out of the frame. He walked toward me, his arms swinging in that familiar, lanky gait, a ghost summoned from silicon.This is the new frontier of artificial intelligence: the ability to virtually resurrect the dead. It is a power that feels both miraculous and profoundly unsettling, forcing us to confront a question we’ve only ever pondered in science fiction: just because we can, does it mean we should? The ethical dimensions of this technology were hilariously, and terrifyingly, illustrated in a recent Saturday Night Live sketch that served as a perfect cultural Petri dish for our anxieties.In it, a family uses an AI photo app to animate old pictures for their grandmother, only to watch the digital recreations of her loved ones descend into surreal horror—a father roasting the family dog, a mother’s torso detaching and floating as a nuclear bomb detonates. The comedy works because it’s rooted in a tangible truth about the current state of AI video generation; these models, for all their advances, are still in their technological infancy, prone to spectacular and often disturbing failures of interpretation.The core of the issue lies in the data these AIs are fed. While large language models like GPT-4 train on a near-infinite corpus of text, the available library of high-quality, well-labeled video is comparatively minuscule.Models like Veo and OpenAI’s Sora are primarily trained on public videos from platforms like YouTube and Instagram, which are inherently biased toward the dramatic, the viral, and the visually engaging. We post videos of wedding cake disasters and heated arguments, not hour-long clips of someone quietly grilling a hot dog or gently holding a baby.Consequently, AI models become masters of chaos but stumble over the mundane. They lack the foundational understanding of simple, everyday physics that a human takes for granted.This has spawned an entirely new market, with companies like Waffle Video now paying creators to generate bespoke training data—footage of people performing boring, routine tasks to teach AIs what ‘normal’ looks like. Until this data gap is closed, we will continue to see these ‘avocado chair’ moments, where the AI’s imagination spectacularly misfires.The psychological impact of these errors cannot be overstated. Animating a long-departed relative like my grandfather Max, a man I never met, feels like a fascinating historical experiment.But applying the same technology to a recently deceased parent, child, or spouse is a different matter entirely. The SNL sketch, in its absurdity, highlights a real danger: the potential for these digital recreations to corrupt our most cherished memories.A splitting kitten in a test video is a amusing glitch; a splitting grandmother is a traumatic image that, once generated, cannot be unseen. The technology is advancing at a breakneck pace.The animated portraits from My Heritage’s Deep Nostalgia, which felt revolutionary just a few years ago, now seem primitive. We are on an exponential curve, and it is plausible that within a few years, these models will be far more stable and reliable.But this rapid progress demands a parallel conversation about digital dignity and the ethics of posthumous representation. Do we have the right to reanimate someone without their consent? What are the psychological consequences for the living, potentially hindering the natural process of grief by offering a digital facsimile that can never truly replace a human relationship? As a society, we are navigating this uncharted territory without a map.The current generation of AI video tools, for all their wow factor, are not yet trustworthy custodians of our personal histories. The risk of a random, horrific visual detour is still too high.The prudent path, for now, is one of cautious patience. We must allow the technology to mature beyond its current propensity for nightmare fuel and develop robust ethical frameworks to guide its use. The power to bring back the dead is now at our fingertips, but with it comes the profound responsibility to decide when, and if, we should ever press play.

#AI video generation

#digital resurrection

#ethics

#technology limitations

#SNL sketch

#featured

Stay Informed. Act Smarter.

Get weekly highlights, major headlines, and expert insights — then put your knowledge to work in our live prediction markets.