Hungry Hungry Hippos - H3
Deep Papers
00:00
Using State Space Models to Model Long Sequences
The fast Fourier transform is the main reason, but like why are they so good at that? I think there's maybe two ways to look at it. So one just from a efficiency standpoint. You can kind of use a state space model to model audio waveforms. They have these nice demos where they can get a machine to kind of count from one to 10 with kind of no priors.
Transcript
Play full episode
