Skip to playerSkip to main content
  • 13 hours ago
Transcript
00:00I know your background is you came from Amazon, Google, AI. Is that something that you really
00:05come into this role with the express purpose of doing, sort of implementing and embracing AI?
00:10Yeah, you know, when I was at Google, it really became pretty clear to me that video was one of
00:15the hardest types of AI. You simply have more electrons moving through the NVIDIA chips.
00:20You know, like a document has a few, you know, let's call it a few little electrons that have
00:26to move through when you have a document, even if it's a six page or 10 page or a hundred page
00:29or a thousand page document, it's not as much as when you have a 4K frame and you have
00:34thousands of 4K frames to be able to find either the image, the sound, the emotion, you know,
00:40the context, the logo. That's really hard to do. And multimodal AI is difficult and it's expensive
00:49because you just simply have more moving through those NVIDIA chips. And quite frankly, you've
00:53got more variability across all those modalities. Modalities are like text and image.
00:58And sound. And in some cases, even emotion in the video, being able to source that and
01:04understand it really hard problem in AI.
Be the first to comment
Add your comment

Recommended