[Weekly Update] How LLM will affect Movies

Look, I know I'm the kind of assholes who forgets about his girl's birthday. But I swear it's not because I don't care. You just underestimated how scatterbrained I can be!

Anyway, yeah, I forgot to write an update again. Sorry uwu.

CHANNEL UPDATE

Summer Horror:
Voting result for our summary horror video is out. Looks like most people want to hear more about Japanese urban legends. That means, next month, we'll talk about movies about the slit mouth woman, the coin locker babies, and the teketeke. Maybe I'll toss in some ghosts stories like Hanako of the toilet, too. Google at your own risk, yo.

Next Video:
Our next video is ready to go, and will be publish in a few days. I didn't put up a forecast, because it's a video about garbage movies. But if you really want to see the movies before I spoil them, they are A Legend (available on YouTube for rental), The Journey of Flower (available on YouTube illegally, not to be confused with the 2015 TV show), and Bureau 749 (Good luck finding this one).

Bonus Video:
For our monthly bonus video, we are working out way through the 80s classics. Since we are doing horror next month, I'm thinking of covering a horror film. We have previously talked about classics like The Thing, and Alien. So we can only go more fringe from here. The options are: Hellraiser, Child's Play, Dawn of the Dead. Let me know which one interests you the most!

MEDIA TALK

Like an old man learning to use a smart phone, I've been messing around with LLM, Large Language Models such as ChatGPT, Gemini, an other text based generative AI with logos that look like an anus.

I know there is both ethical and environmental concern regarding the technology. But being in the industry that's severely affected by it, I feel the need to at least understand it on a personal level before bracing for its impact.

So, I tossed in a few cliche scenes from other movies, as well as ask a few questions, to test the creative limit of LLM. Here is what I find, and what they'll mean for the film industry:

• One of the first things I learn, and the most unintentionally funny bit, is that LLM can accidentally generate something that violates its own ToS. In my case, I asked ChatGPT to generate a story summary about a young adult healing from his traumatic past induced by his abusive parent. I don't know what ChatGPT decided to show me, but apparently it's too graphic to be shown.

While there are other LLM that do not have this sort of censorship, the best ones all have some form of safeguard, and for good reasons. If I know anything about artists, is that we love to push boundaries. With such a clear boundary set in place, unreachable by LLM, it's perhaps reasonable to expect indie artist to want to cross that line with their work.

So, my prediction is: Expect a return of the 80s style gruesome body horror, with graphic kill scenes that you can't get from your typical generative AI. Alternatively, there may be more films focusing on more sensitive issues that corporations are too much of a coward to touch.

In the long run, this means further normalization of discussion on these sensitive topics. AI guideline will, in turn, relaxed, causing artists to push even further. What that means ultimately? I'm not sure.

• Another thing I noticed is the incredible leap of logic made by LLM. One of the harder part of script writing is to create a scenario so complicated that your audience don't know how to solve, and then you have to let your characters solve it. LLM are experts at solving your problems... Except the solution is hit or miss. Still, it offers solutions, and it offers them non-stop.

So I asked the LLM how to create a set piece set inside an elevator. The goal is to escape a serial killer hunting the characters. And the sequence has to utilize each character's strength. What ChatGPT gave me is... wild. The musician, with her expert hearing, realizes the elevator cable is about to snap. So she drew the diagram of the elevator on the wall with her lipsticks, and allow her tech savvy friend to hack the elevator, stopping it just short of falling, closing the door on the killer.

It reads like some complete non-sense from early Sierra point-and-click games, with a serious case of moon-logic. But this is also the exact kind of BS I saw in a lot of escape room movies, including that dumb sh*t movie literally called Escape Room.

So, expect some seriously creative and non-sensical problem solving in future movies. If you see moon-logic, this is probably why. I'm actually kinda excited about this one, since if it's good, it feels smart and superhuman. And if it's bad, it's still creatively stupid.

• Last example for today, I asked ChatGPT to give me an idea: Imagine a meteorite landed in Central Park, and an alien creature spawns from it. What should the meteorite turns out to be? I tell ChatGPT I want to avoid the cliché alien egg. So, ChatGPT tells me it can be an dimensional rift, an arrangement of doors, a lantern, or strangest of all: a memory.

It's pretty clear that, outside of the rift (which itself is a cliche), ChatGPT has no idea what the concept is about. Of course not, LLM is not sentient, it simply guesses the next word. These creative answers, while unique, aren't exactly visual. Imagine a lantern smashing into Central Park isn't exactly tonally scary, or sci-fi, or funny, or interesting. Not to mention, what would a physical manifestation of a memory looks like.

The problem persists even when I ask it to generate an image. Generative AI seems to have the most trouble when it comes to creating interesting visuals, as it has no idea what visual is considered unique and interesting. My guess is that with text, there is enough volume out there to gauge human interests. But in our visually driven society, any image that appears frequently is an image that's already been outdated.

In other words, future artists will have to race AI on creating unique and never before seen visuals. While many will fail in this quest, those who succeed will definitely create something striking.

So, will I be using LLM in the future? Probably not. Honestly, a regular chatbot might be better to bounce ideas off of, as it is just an unbiased wall that doesn't alter my style or thinking. As it stands, I feel that LLM is best at showing me what the current cliche is, so I can avoid them. In other words, the best part of ChatGPT is how uncreative it is.

Anyway, I'll see you with a new video tomorrow!