I've been a bit obsessed recently by Scott Alexander's market about whether an AI will be able to generate a high-quality film to a prompt by 2028. I find it surprising that it's been sitting almost exactly at 50% for almost 6 months. Right now there's some pretty heavy market-making going on, and I think we might have "Keynesian-beauty-contest'd" ourselves. (I'm as guilty of this as anybody.)

As well as the market makers, there are a few people who think the true probability is significantly higher and a few who think it is significantly lower. While it's possible both groups are wrong, and the true probability is 50%, it is also plausible that we are all missing something, and it should be significantly higher or lower.

If I'm right, then at some point, presumably the market will reach a new equilibrium?

To define "reaching a new equilibrium", I will require the market to settle either below 44% or above 58% for at least a month. I reserve the right to wait longer than a month if I think some whale is throwing mana at trying to make this market resolve positively. (Seems like a small risk since the main market is very liquid and this one probably won't be.)

Title says "by September", but specifically, it's a year today: 15th September 2024.

I may bet in this market. I promise not to abuse my (very minor) discretionary powers outlined above to try and turn a profit.

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

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

52% chance. IE "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.