Concepts of an Experimental Study of a Prototype: Generative False Memories

Scene from the 1897 film The Haunted Castle by Georges Méliès

William Gibson wrote a poem called Agrippa (A Book of the Dead), the central gimmick to the work being that, depending on which version you read, it could only be read once before it “ate itself.” This was achieved through photosensitive paper for the physical copies and some clever programming for the version that came on a floppy disk. There’s very little information about this work besides some scattered YouTube videos, a few mentions in blogs, and very rarely copies of the digital version of the poem that have been decrypted to undo the corruption. The point of this ephemeral approach is that it left the reader with an impression of the poem rather than something they could refer back to as needed or re-read to bolster the memory. Rather than a concrete memory, you’re left with something more nebulous, even fragile. False memories are more common than people think, even leading to a widespread conspiracy theory called The Mandela Effect. It refers to any sort of common misconception, but specifically focuses on nostalgia, such as the spelling of the Berenstain Bears, the color of Pikachu’s tail, or (namely) the imprisonment of Nelson Mandela. The conspiracy side of it is that these are not false memories, but people experiencing different realities that overlap and absorb one another. If that sounds overly silly to you, then congratulations are in order because it means you have a sense of humility. Nobody likes the idea of forgetting things, especially as we get older, but it’s quite the response to explain away a fuzzy recall with an intertwined multiverse.

You just don’t remember that thing very well. That’s it. It’s nothing to be ashamed of.

I bring all this up because it feels like a number of people work under a vague impression of something rather than anything more concrete or easily referenced. This shouldn’t be any serious problem; inspiration often involves the donning of rose-tinted glasses, leading you to make a better version of what you were trying to emulate. However, this doesn’t always pan out. In the worst case scenario, it can lead to a kind of revisionism, remembering things as being better than they once were or glossing over something problematic because of the warm and fuzzies it makes you feel.

Somewhere in the middle of this spectrum, at present, is generative AI and an “experiment” that seems to be yielding the same sloppy results over and over again.

I’ve talked elsewhere about a pattern of behavior I see on LinkedIn, in which someone who loudly and proudly admits up front that they have no experience in filmmaking (at best, they’ve watched all the special features on a DVD once) tries their hand at filmmaking by way of an AI-Based Video Generator. The result is something that isn’t actually all that cinematic. It seems to be trying to hit all the right notes of what modern cinema looks like, but it’s all a bit… off. It’s not even off by a little bit. It’s off in that people are literally walking through walls, reacting to things happening behind them, and even changing their entire physiology within the same shot. That’s an extreme case, mind you; the other results are often just boring and derivative. The point remains is that you’ve got someone who has only the vaguest understanding of how a movie is put together trying their very best to emulate what they’ve been consuming over the years, and it becomes obvious they look at movies in a very different way than you, and that includes how they remember them.

To put this in perspective, I’m going to tell you a rule of filmmaking, and if you’ve never heard of this before, you’re never going to look at movies the same way ever again. It’s called the 180 Rule and it goes like this: When you’ve got two characters talking to each other, the camera can move (or cut) to anywhere else in the space so long as Character A is on the left and Character B is on the right. Essentially, you’re drawing an imaginary line between these two characters, and your point of view cannot cross it. If it does, it becomes disorienting and can even break immersion. After all, the camera is supposed to be your point of view, and that means there should be a consistency to it. When you’re watching two people talk, and your eyes move left and right, you’re not suddenly on the other side of them with their positions swapped. That would make the dialogue extremely awkward.

It’s true that some movies deliberately break this rule for narrative purposes, to intentionally make the audience feel an unease. It’s a “rule” in the loosest sense of the word. You can break it, but you have to have a reason for doing so, a reason that serves the story.

On the whole, most people probably don’t think of this. Most people probably don’t notice or aren’t bothered by little mistakes and errors in a movie. They don’t pay attention to the set of rules that the filmmakers are following in order to tell a cohesive and immersive story. It’s usually only the diehard movie buffs or people who work in the industry directly that understand the language of cinema and can therefore have a completely different moviegoing experience than the average Joe and Jane Consumer.

One of the best pieces of advice I ever got for anything was, when it comes to making movies, you not only have to watch a lot of movies, but you have to watch them intelligently. Why is this your favorite movie? Is it the performance of the lead actor? If so, what makes it so effective? Is it the story? If so, how is the story structured and do you care that it’s actually rather derivative of similar narratives? If you simply enjoy the movie, then you simply enjoy the movie, but that’s not going to make you a filmmaker. You watch a lot of movies? Cool, but that doesn’t make you a director any more than binge-watching The Pitt makes me a surgeon.

When these AI Enthusiasts set about making their “experiment in cinema” and ultimately exhibiting their rough drafts (which they only bring up when receiving criticism), they’re not bringing a full understanding of what it means to make something cinematic. They’re bringing what is at best an impression of something cinematic, a faint echo of something established that they’ve tucked away in their memory that’s been left to deteriorate by that oddly mundane process of inscribing electrical information onto meat. They believe they have an idea of how a movie is put together, but their reference is their own faulty memory. It looks cinematic to them because, in their minds, that’s what cinema looks like. In actuality, it’s just how they remember it, and they don’t remember it very well. It always makes me think of what Roger Ebert said about Battlefield Earth’s unorthodox cinematography, “The director, Roger Christian, has learned from better films that directors sometimes tilt their cameras, but he has not learned why.”

To close, I’d like to speak to the AI Enthusiasts directly for a moment, assuming they’ve gotten this far before their brainrot-addled noodles tuned out and went back to Instagram.

When people point out your continuity errors and all the other problems with what you’ve presented, firing back with “it’s just an experiment” isn’t really helping to move the needle on the development of this tool you’ve touted as the greatest invention since the wheel. I’ve seen the rest of your posts, and you’re not learning why the cameras are sometimes tilted. There’s no shortage of so-called experiments from a host of other enthusiasts who are convinced they’re now a one-person marketing agency capable of going toe-to-toe with some of the greats of the business all from the comfort of their borrowed Chromebook at a Starbucks they’ve never actually ordered anything from. I can’t tell their experiments from yours, and they’re not even calling their slop an experiment. They’re insisting it’s a final product. Meanwhile, here you are pretending you’re a filmmaker and producing a piece of brainrot indistinguishable from all the other disposable mental junk food, and I’m supposed to be at once impressed while acknowledging that you don’t actually know what you’re doing? You’re not really experimenting. That’s just an excuse you’ve thrown out there to avoid having to actually learn anything about the craft you’re claiming is about to be disrupted by an aggregate of mediocrity in the form of resource-hungry datacenters trained on other people’s hard work.

If you’re not going to take this seriously, if you’re not going to check your work, if you’re not going to learn from the errors that have been pointed out to you, maybe take your own advice and let the legitimately talented people from the film industry work with these tools and see what they can produce.

Originally published at https://www.linkedin.com on 10 MAY 26 SUN