2 ms·
> One obvious direction is predicting the next video frame. This is being done. There are incredible generative video models...
by qeternity 2y ago
> One obvious direction is predicting the next video frame.
This is being done. There are incredible generative video models...
- p1esk 2y agoI would not call them “incredible”. Video generation is still at gpt-1 level. It will need several orders of magnitude scaling to get to gpt-2 level. We’ve just started making very first baby steps in this direction.