
https://twitter.com/solarise_webdev/status/1598708384441798656?t=gSyX2gzixpMryYOsyyxt7A&s=19
Close date updated to 2027-12-31 6:59 pm
Dec 2, 12:26pm: By 2028, will AI be able to make a full animated movie with consistent plot, design and characters? → 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?


@Kormann Maybe, maybe not. Not sure if you watched the original Corridor Digital video, but they had to do A LOT of manual work to make it even visually coherent, forget about all of the writing and acting. And that was for a cool-for-the-novelty-but-not-actually-very-good couple of minutes about two dudes play rock paper scissors.

@ChanaMessinger Is there a minimum length or complexity? Could it be a 1 hour animation of a snail moving through some leaves?




@StrayClimb Hmm, I'll think about whether I disagree with 70% enough to trade, I probably don't. But I don't think legal restrictions could prevent this from occurring. It's much too easy to build with 2027's ai algorithms and compute substrates.



@IsaacKing like are you just confused why I don't bet every capabilities market up? or did I bet in a way that seems actively in conflict

@IsaacKing like, I think there will be tons of legal restrictions on ai. but I don't think any of them will stop ai from being able to do anything a human can.



Does "consistent plot" mean that you have to like it about as much as a regular Disney movie? Or just passable?
@EmrikGarden it's meant to get at the thing where GPT3 sometimes loses track of who the characters in the stories are

does a pipeline consisting of
prompt --> (LLM) --> movie script --> (video gen model) --> movie
count
@Lauro I think that resolves to yes if you're not doing more decision making it selecting in the intermediate steps and no otherwise

2028? easy peasy. question is if it'll be done by 2024. I'd bet no on that, despite thinking general models will be mildly superhuman in many more subdomains by then, and narrow models will be starkly superhuman in a great many things - this is an extremely tall order, very few humans could just do this! but it seems likely to be easier than whole-cell simulation; I also think deepmind will pull off whole cell simulation before this happens.

The exclusion of 'good' from the requirements is intentional, right?










By 2028, will AI be able to make a full animated movie with consistent plot, design and characters?, 8k, beautiful, illustration, trending on art station, picture of the day, epic composition