How to Scale Large Scale Networks

The tricky part in practice is, as you said, right, like the modern networks that we use and deep learning are so big and so complicated. But I think there's a lot of interesting stuff to be down there still though. Right? So if if you can show that they are scalable, then then you can go back to the implementation level and think about how those computations may be implemented at the level.Yeah, I think that's fair, right? Yeah, we can do that. And I think, yeah, a lot of what we need to explore is what what is that like what kind of biases should we build in that are stable or that wouldn't restrict the scale