Here's a human-made video of a Rube Goldberg machine:

https://www.youtube.com/watch?v=Ss-P4qLLUyk

This outdoor Rube Goldberg Machine goes around my entire yard, and swishes a basketball shot after 70 steps. This video was filmed in one take, meaning there...

The Swish Machine: 70 Step Basketball Trickshot (Rube Goldberg Machine)

 that the ability to generate a video like that (indistinguishable from reality even on close human inspection) from a prompt will require an AI that's within 1 year of AGI.

Since this is another leading indicator, I thought it would be good to have a Manifold market for it.

Yudkowsky's thinking is that it takes a high-fidelity physics model to make all the balls and dominoes and things interact realistically. If you understand the physical world at that level, AGI isn't far behind.

Maybe a minute? Don't quote me on that yet. Let's discuss in the comments what best hews to the spirit of the question, about a leading indicator for AGI.

Ask other clarifying questions in the comments as well -- I'll add them to this FAQ!

[Ignore any auto-generated clarifications below this line. I'll add clarifications to the FAQ above.]