In early 2028, will an AI be able to generate a full high-quality movie to a prompt?

4.6k

Ṁ21kṀ14m

2028

36%

chance

ALL

EG "make me a 120 minute Star Trek / Star Wars crossover". It should be more or less comparable to a big-budget studio film, although it doesn't have to pass a full Turing Test as long as it's pretty good. The AI doesn't have to be available to the public, as long as it's confirmed to exist.

Scott Alexander's 5 year predictions

Get

1,000

to start trading!

People are also trading

By 2029, will an AI be able to generate Video Games comparable to ~2023 'AA' Mid Market Games?

69% chance

In 2028, will an AI be able to generate a full high-quality movie to a prompt, with human aid?

61% chance

By 2028, will AI be able to make a full animated movie with consistent plot, design and characters with no input besides the original prompt?

71% chance

In 2028, will an AI be able to generate a full low-quality movie to a prompt?

80% chance

In early 2028, will an AI be able to generate a full high-quality movie script to a prompt?

66% chance

By what year will an AI be able to generate a full, high-quality movie to a prompt?

In early 2028, will an AI be able to generate a full high-quality movie from a script?

40% chance

Multi year: Will an AI be able to generate a full high quality movie to a prompt?

In 2038, will a movie be able to generate a high-quality AI to a prompt?

69% chance

at the end of 2026, will an AI be able to generate a full high-quality tv ep to a prompt?

17% chance

Sort by:

@ScottAlexander We should interpret the example prompt of "make me a 120 minute Star Trek / Star Wars crossover" to mean that the prompt couldn't include a script, directions, and so on, correct?

@Panfilo that is not gonna be the bottleneck, but yes

@Panfilo but the system prompt can contain anything, since that's part of "the AI"

@0xseraphim I don't think I agree with "anything," since a system prompt that contained a full prewritten Star Wars / Star Trek crossover script already approved for quality would, for example, clearly fly in the face of the spirit of the question.

Dan Greenheck (@dangreenheck) on X

Ever wonder what $1874.40 of Opus 5.5 tokens looks like? Wonder no more. I highly recommend watching the whole video on 1080p on a large screen with sound on. Pay attention to all the little details. - Birds flying and diving into the ocean. - Cloth, signs and lights swaying in the wind. - Cra…

Do you guys know something about Scott Alexander's taste that I don't?

@Panfilo I think this is a reaction to some of the emergent capabilities of Astra and Opus 5.5

@Panfilo they’re showing their taste

opened a Ṁ4,450 YES at 35% order

Such a movie would be written in code. Diffusion models would be the compilation step at best.

https://manifold.markets/ScottAlexander/in-2028-will-an-ai-be-able-to-gener#aeipdeiypa4

@VitorBosshard can we get an updated take on this?

market has gone up by 1 percentage point since then! 😂

It would be cool to see a breakdown of who thinks we are one breakthrough away from this in the next 16 months, vs. who thinks we are more than one breakthrough away (but still thinks it will happen).

@Panfilo I don't think a "breakthrough" is necessarily a requirement. We could get there just with incrementally smarter models over the next 15 months. That being said, it's highly plausible that there will be breakthroughs in that timeframe (including those discovered through RSI runs) that have implications for long-form video gen

opened a Ṁ5,000 YES at 30% order

@BionicD0LPH1N I would add to check at examples from September 22 2025. GPT-5 and Opus 4.1 were, lets just say, not great at animations. This was presented as an impressive animation by Opus 4.1. SOTA from last year movie scenes recreation:

@BionicD0LPH1N Another great example:

@BionicD0LPH1N My favorite one so far:

bought Ṁ1,000 NO

@BionicD0LPH1N it’s cute but none of them make any sense and they’re just a few minutes long.

Meng To (@MengTo) on X

Opus 5.5 is next level. I’m genuinely impressed by the three.js details. I generated a playable boat scene through Japanese landscapes with dynamic weather, day and night lighting, realistic textures and 3D characters. Live site: https://t.co/YMW4Zp52In The water reflections, physics, scenery an…

@MachiNi Seems you’re not used to following AI progress? Wait a few months.

I'd argue the one with the girl sending messages to the claude demon in the sky makes sense. At the end it surrounds the word "you" in her letter, sends it back to her and she hangs it up on her wall.

If we're on the timeline where the first movie that resolves this question is a sequence of 3D javascript scenes generated by claude and then smooshed together with higgsfield or something, that's going to be weird

@BionicD0LPH1N A few model generations away, surely:

@BionicD0LPH1N I am. That’s precisely why I’m so underwhelmed.

bought Ṁ100 NO

@BionicD0LPH1N for anyone who thinks this is a ‘masterpiece,’ it’s worth knowing that Scott Alexander has refined taste. He seems to have a high bar for what counts as good poetry. I’d be surprised if his taste in film was significantly inferior.

I think the poetry is likelier to resolve yes because LLMs are simply, and obviously, much better at writing.

/ScottAlexander/in-2028-will-an-ai-be-able-to-write

@MachiNi https://www.bbc.com/news/articles/ck7v4y45893go

Think we're already there with writing. People are just in denial about it.

@MachiNi It's not a masterpiece. What matters is the rate of progress and the state of things a year ago vs today. I have trouble seeing the gap between a year ago and today being greater than the gap between today and resolution of this market. We shall see.