One Year Later: AI Learned to Make Movies. But Did It Learn to Make Us Feel?

A little more than a year ago, a film professor and a student in Chile made a strange little movie with artificial intelligence. It was not a glossy Silicon Valley demonstration, a Hollywood studio experiment, or a commercial designed to prove that the latest model could produce prettier pixels than the competition. It was called El Mono con la Lengua Afuera, or The Monkey With Its Tongue Out, and it followed an encounter between a filmmaker and a spider monkey as real images collided with synthetic ones and questions about surveillance, control, memory, and the relationship between humans and algorithms began to surface.

At the 2025 Sparknify Human vs. AI Film Festival in Silicon Valley, the film won the Best AI Generated Award. Its creators were Sebastián Arriagada, a professor at the University of Chile School of Cinema and Television, and Emiliano Urbina, then a fifth year student. The University of Chile later described how the project grew out of Arriagada's course Artificial Intelligence: Future Images, with Urbina joining him to rework an earlier short film into something closer to a collaboration among filmmaker, student, and machine.

What made the film interesting was not simply that artificial intelligence could create an image. Even in 2025, that was no longer particularly shocking. The more important question was what happened when the machine stopped being merely a tool for making images and became something closer to a creative collaborator.

At the time, that question still felt experimental.

One year later, it feels almost quaint.

Because in roughly the time it takes Hollywood to develop a screenplay, generative AI has gone through another technological mutation. It did not merely learn to make better pictures. It learned to move them, make them speak, add sound, maintain characters across scenes, follow camera instructions, and participate in editing and post production. What began as a novelty capable of producing strange five second clips is rapidly becoming a production environment.

The question is no longer whether artificial intelligence will become part of filmmaking.

It already has.

The more difficult question is what happens when it becomes normal.

When the Demo Became a Camera

For the first few years of generative AI, video was impressive largely because it existed at all. A dog could run through a field. A woman could walk down a street. A spaceship could fly over a city. Yet the illusion rarely survived close inspection. Hands mutated, faces subtly changed, objects disappeared, gravity behaved strangely, and a character who looked one way in one shot might appear to be a distant cousin in the next.

Then the models started improving with extraordinary speed.

In late 2025, OpenAI introduced Sora 2 with improvements in physical accuracy, realism, controllability, dialogue, and synchronized sound. Google was pushing Veo toward longer and more coherent sequences, richer audio, and stronger cinematic control. Runway was refining camera movement, scene composition, timing, and visual consistency. Across the industry, the race was no longer simply about making a convincing clip. The real challenge was whether AI could begin behaving like something a filmmaker could actually direct.

That distinction is critical. Cinema is not a collection of beautiful five second videos. A character must remain the same character. A room must remain the same room. Wardrobe has to persist. Objects need to stay where they were placed. A director needs to decide where the camera goes, what happens next, and how one shot connects to another.

Over the past year, that problem has begun to look increasingly solvable. Newer systems emphasize character identity, object consistency, camera control, and higher resolution output. Filmmakers can increasingly provide beginning and ending frames, reference images, visual styles, or movement instructions, then iterate on the result instead of simply prompting again and hoping for something useful. The first generation of AI video asked, What can the machine imagine?

The emerging generation asks a much more important question: Can the filmmaker direct it?

That is the moment when a technological curiosity begins to resemble a medium.

AI Moves Into the Editing Room

Another transformation has been quieter but perhaps even more consequential. Generative AI is no longer confined to creating images or shots from scratch. It is moving into the spaces between traditional filmmaking processes.

Adobe and other creative software companies have increasingly embedded generative capabilities into tools already used by professional filmmakers, editors, designers, and visual effects teams. Instead of regenerating an entire sequence because one detail is wrong, creators can increasingly change individual objects, extend environments, alter backgrounds, manipulate camera movement, modify audio, or create additional frames.

The workflow is beginning to shift from something like, generate something and hope it works, to a much more familiar creative cycle: generate it, direct it, change it, edit it, and finish it.

That sounds much more like filmmaking. And much less like prompting. This may ultimately matter more than the spectacular AI clips circulating online. Most professional filmmaking is not a single moment of invention. It is revision. Directors change performances. Editors restructure scenes. Cinematographers alter framing. Visual effects teams refine details. Sound designers build atmosphere. A film slowly emerges through thousands of decisions. Once AI becomes editable, controllable, and iterative, it stops being just an image generator and starts becoming part of that decision making process.

Hollywood Crosses the Line

For years, Hollywood's public relationship with generative AI was defined largely by fear. Writers worried about scripts. Actors worried about digital replicas. Visual effects artists worried about automation. Studios worried about copyright. Unions worried about jobs.

Those concerns have not disappeared. If anything, they have become more serious as the technology becomes more capable. But another shift has occurred at the same time. Major entertainment companies are beginning to treat artificial intelligence not simply as an external threat, but as infrastructure they may eventually need to understand, license, and use.

One of the clearest signs came when Disney announced a major partnership with OpenAI involving its characters, intellectual property, and future technology development, along with a billion dollar investment. The symbolic importance was difficult to miss. One of the most powerful owners of entertainment intellectual property in the world was no longer standing outside the generative AI revolution watching cautiously. It was stepping inside. The debate had moved from the studio gates to the boardroom.

Some of cinema's most established figures are also grappling publicly with the transition. Martin Scorsese has argued that cinema is still a relatively young medium and must remain open to evolution. James Cameron has described generative AI as a major new wave of cinema technology while warning that filmmakers must master and control it rather than allow it to replace artists.

Those two positions are not actually far apart. Together, they capture the strange place cinema now finds itself. The technology is becoming too powerful to ignore, but embracing it without asking what it changes would be equally reckless.

The Real Disruption May Be Economic

The most spectacular AI images attract attention, but the economics could reshape the industry far more profoundly. Generative tools can already compress parts of development, storyboarding, concept design, visualization, localization, post production, sound, visual effects, and marketing. The consequence is not necessarily a future in which someone types, make me a two hour science fiction movie, and receives a masterpiece ninety seconds later. That scenario makes for a dramatic headline, but it misses the transformation happening right now.

The more immediate change is that a five person team may increasingly be able to attempt something that once required fifty people. A filmmaker with a modest budget may be able to visualize a world that once required millions of dollars. An independent creator may gain access to capabilities previously available only to major studios. A director may test twenty visual ideas before constructing a set. An advertiser may create multiple versions of a campaign without reshooting it. A small production company may suddenly gain access to capabilities once spread across several specialized departments.

That changes who gets to make movies.

It may also change who gets paid to make them.

Filmmaking has always been constrained by capital. Cameras cost money. Sets cost money. Actors cost money. Lighting, travel, visual effects, editing, sound, distribution, and marketing all cost money. Artificial intelligence has the potential to lower some of those barriers dramatically.

For creators who have historically been excluded because they lacked capital, equipment, or institutional access, this could be liberating. A teenager in Taipei, a student in Santiago, or an independent director in Nairobi may be able to visualize ideas that would previously have required access to Hollywood infrastructure. But democratization comes with an uncomfortable second half.

When everyone gains access to extraordinary production tools, technical polish itself becomes less extraordinary. The bottleneck moves.

Yesterday, production capability was scarce. Tomorrow, the scarce resources may be taste, judgment, originality, perspective, story, and emotional intelligence. In other words, the more capable the machine becomes, the more important the human decisions surrounding it may become.

When Infinite Creativity Starts Looking the Same

This brings us back to The Monkey With Its Tongue Out. The University of Chile's account of the film now reads almost prophetically. Its creators were not merely fascinated by what artificial intelligence could produce. They were also concerned about what might happen if the technology became too dominant, too universal, or too culturally powerful. If millions of creators rely on the same models, trained on overlapping bodies of visual culture, could those systems begin pulling our imagination toward the same center?

That concern looks increasingly important.

Generative models can produce almost anything, at least in theory. But what happens when millions of people ask the same systems to imagine beauty? The same systems to imagine love? The same systems to imagine fear, power, success, desire, or the future?

Do we receive infinite creativity?

Or infinite variations of a statistical consensus?

Artificial intelligence could democratize filmmaking. It could also homogenize it. It could give millions of people a voice while quietly teaching those same millions what an image is supposed to look like.

The central battle in AI cinema may therefore have very little to do with resolution, frame rate, or prompt accuracy. It may be a battle over imagination itself.

The Machine Can Depict Sorrow. But Can It Understand It?

There is an even more fundamental question hiding beneath all of this. Cinema was never fundamentally about images.

A technically perfect image can be boring. A badly exposed image can be unforgettable. A multimillion dollar visual effects sequence can leave an audience cold, while a simple close up of a person's face can make an entire theater cry. The true currency of cinema is emotion. That is where the AI conversation becomes much more difficult.

Generative models can learn what sadness looks like. They can learn the musical patterns associated with grief. They can generate a trembling voice, rain against a window, a lonely figure standing under a streetlight, or the facial movements we associate with heartbreak. They can study millions of examples of how human beings have represented sorrow.

But the machine has never lost someone it loves. It has never been jealous. It has never wanted to be accepted. It has never experienced humiliation. It has never fallen in love, watched a parent grow old, or feared death.

And yet it may still be capable of creating an image that causes you to experience those emotions.

That possibility changes the question completely.

Perhaps a machine does not need to experience an emotion in order to create the conditions that produce emotion in us. A violin does not feel sorrow either. A camera does not understand love. A paintbrush has never grieved.

The difference, of course, is that historically there was always a human being standing behind those tools. Now we are beginning to experiment with something that can participate in the creative process itself. And that is why the next test of artificial intelligence should not be another benchmark.

It should be an audience.

September 26: Put the Machine in Front of an Audience

On September 26, 2026, the Sparknify Human vs. AI Film Festival returns to San Francisco with an experiment built around one deceptively simple idea. Do not tell the audience who made the film.

More than 3,000 films from around the world entered this year's festival, with 32 short films selected across Human Produced and AI Generated categories. Together, they explore eight fundamental emotions: Joy, Sorrow, Intensity, Serenity, Love, Hate, Desire, and Disgust.

The films will be screened at the Delancey Street Screening Room on San Francisco's Embarcadero. But during the screening, one crucial piece of information is intentionally withheld. The audience will not know whether each film was created by traditional human filmmakers or with generative artificial intelligence until after the viewing.

That changes the experience completely.

If you do not know how the movie was made, you cannot reward the human filmmaker simply because you want humans to win. You cannot praise the AI film simply because the technology is impressive. You cannot enter the theater with your conclusion already written.

You are left with only the film. And yourself. Did it make you laugh? Did it disturb you? Did you feel desire, disgust, serenity, anger, sadness, or love? Did something happen inside you?

For decades, the famous Turing Test asked whether a machine could communicate convincingly enough for a person to mistake it for another human. Cinema may now require a different kind of test.

The question is no longer simply whether AI can fool you into believing a human made the movie.

The question is whether AI can make you feel something that is real.

That question has become much harder to answer over the past twelve months. The technology has advanced at extraordinary speed. Images are better. Motion is better. Continuity is better. Sound is better. Control is better. Professional workflows are beginning to emerge. The distance between demonstration and filmmakin g is narrowing faster than almost anyone expected.

Now it is the audience's turn.

On September 26 in San Francisco, sit in the dark. Watch the screen without knowing what created what. Forget the production method for a moment and pay attention to your own reaction.

Because after a year in which artificial intelligence learned so much about how movies are made, there is still one question technology cannot answer for us.

Human vs. AI. Can you feel the difference?