News for hackers. Stuff that matters.
Re:Project Genie: Experimenting with infinite, interactive w... by in-silico on Friday January 30, 2026 @04:26AM (#46822288)
But how does it interpolate?
Re:Project Genie: Experimenting with infinite, interactive w... by slashdave on Friday January 30, 2026 @02:47PM (#46828971)
Pixel by pixel, time-slice by time-slice, in a 2D+T convolution. You provide enough examples of videos of changing point-of-view, and the model reproduces what it is given.
Re:Project Genie: Experimenting with infinite, interactive w... by in-silico on Friday January 30, 2026 @07:11PM (#46831775)
Yes, it reproduces what it is given by modelling the rules of physics, geometry, etc.

For example, image generators like stable diffusion carry strong representations of depth and geometry, such that performant depth estimation models can be built out of them with minimal retraining. This continues to be true for video generation models.

Early work on the subject: https://arxiv.org/pdf/2409.09144

Re:Project Genie: Experimenting with infinite, interactive w... by slashdave on Wednesday February 04, 2026 @01:56AM (#46882416)
What? No, it does no such thing. Study the architecture. Pixels in. Pixels out.