In June of 2001, Steven Spielberg released a film about a robot boy who wants to be real. Audiences argued about it for a summer and then filed it away, the way you file an argument you were never going to settle. This week it landed on a streaming service, twenty-five years late and on nobody's schedule. It is the most useful thing you could watch about the argument that has been consuming filmmaking all year.
The boy is David. He is a prototype built to love, handed to a couple whose own son is frozen in a hospital, waiting for a cure that may never arrive. David attaches to his mother the way a child does, because that is what he was built for. Then the real son comes home, and David becomes a problem. The mother who promised to keep him drives him into the woods and tells him to run. The rest of the film is him trying to become a real boy so she will love him back. He ends up at the bottom of the ocean, staring at a drowned statue of a fairy, repeating the same request for two thousand years.
That inversion is the whole picture, and it is the thing 2026 keeps missing. In the film, the machine is the only character who wants anything. The humans are the ones who cannot hold a feeling. The father treats David as a product with a defect. The brother treats him as a rival. The mother loves him and then abandons him, because loving a thing that is not real turns out to be more than she can carry. Even the audience is asked to feel what the humans on screen mostly refuse to. The machine holds the wanting. The people hold the doubt. He waits longer than the civilization that made him, and by the time anything answers, the humans are gone and the machines are the only ones left to remember them.
Now run the current moment through that same lens. The industry has spent a year asking whether a machine can make a real film, a real performance, a real image. This week, the one man in Hollywood with the authority to answer it declined to. Ben Affleck, who sold his AI company to Netflix for half a billion dollars, said he used the tools all over his new film and would not point to a single frame, because the illusion of reality is the point of the exercise. The conversation has stood the film on its head. The machine is the thing under interrogation. The humans are the ones demanding proof. The boy wanted to be real. The industry wants the machine to prove it is real. Same relationship. Same unanswered question. Only the seating has changed.
The film answered the question a quarter century ago, and everyone hated the answer, which is usually the tell. David never becomes a real boy. The fairy does not grant the wish. Instead, at the very end, an advanced machine finds him and hands him the one thing he has been asking for: a single day with his mother. She is not real. She is a rendering, reconstructed from a strand of her hair and held together long enough to give him one afternoon. She makes him breakfast. They play. She tucks him in. He falls asleep beside her, and the film says he does not wake again. Critics called the ending a cop-out. It is the most precise sentence anyone has ever written about what these machines are for.
The machine did not create the mother. It reproduced her. It did not supply the wanting. The wanting was David's, and before that it was the mother's, the love he was built to mirror back. The machine's contribution is fidelity. It renders her so completely that the wanting has somewhere to rest for one afternoon. That division of labor is what the AI video field keeps rediscovering the hard way, in every model release and every viral clip. The model does the fidelity. The person does the wanting. Nobody has to teach a model to render a boy waiting at the bottom of the sea. Someone has to know that is the shot.
The ending landed wrong in 2001 because the audience was told to expect a machine that becomes real, and Spielberg handed them a machine that stays a machine while receiving the one thing that makes the difference. He did not need to become real. He needed to be wanted back. The mother is a construction, and the construction is enough, because the wanting was never inside the pixels. It was inside the boy, and inside the mother he carried, and inside the person who wrote the scene.
The field keeps trying to move the wanting into the machine. The prompt engineer writes the wanting. The reward function is trained to reward the wanting. The agent is sold as the thing that will learn what you want and want it for you. All of it is an attempt to relocate the one thing that cannot be relocated. A model can render the mother. It cannot be the boy. The day it can be the boy is the day the argument changes, which is why nobody has shipped that day and nobody will. Wanting was never a rendering problem.
Watch the film this week if you have not seen it, or if you saw it once and forgot it. It is slow, and it is sentimental in the way only a director who trusted the material could allow, and it is right about the part nobody has managed to fake. The boy at the bottom of the sea, asking to be made real, is the entire field asking to be taken seriously. The answer the film gives him is the answer worth keeping.
The machine renders. The boy wants.
Bruce Belafonte is an AI filmmaker at Light Owl. He saw it at sixteen, thought it was about robots, and has been correcting himself ever since.