back to index

MIND-BLOWING Sora demo | OpenAI's newest model #ai #openai


Whisper Transcript | Transcript Only Page

00:00:00.000 | What's amazing about this model, this model on its own has rendered physics.
00:00:04.300 | If you look at some of the other demos on this tool, you can
00:00:06.700 | actually see fluid moving.
00:00:08.160 | Those are like physics models that would typically be deterministically
00:00:11.300 | written, meaning that a programmer, an engineer writes the physics, and
00:00:14.480 | then the unreal engine renders the physics.
00:00:16.560 | In this case, the model figured out the physics and the model probably
00:00:19.960 | did it in such a reduced way that you could use a reduced amount of
00:00:23.180 | compute to generate something that would otherwise take a ton of compute time.
00:00:26.500 | And I think it really speaks to what's so incredible right now, which is
00:00:29.520 | that software is writing itself, and it's replacing a lot of human determinism.
00:00:33.440 | The way that humans think about constructing software is being
00:00:36.640 | completely redone with these learn models.
00:00:38.640 | This is incredible because Pixar spent years, maybe a decade
00:00:42.440 | trying to figure out hair.
00:00:43.920 | I remember talking to Ed Catmull and he told me hair was just so hard
00:00:46.960 | to get right with wind and lighting.
00:00:49.240 | They literally built the physics engine for years to make hair
00:00:52.720 | across the uncanny valley.
00:00:54.040 | What you're saying here, Friedberg, is somehow this language model,
00:00:57.760 | this image model just outputs it.