Twitter Q&A with Geoff Hinton
The Robot Brains Podcast
00:00
Is There a Learning Argument for Spiking?
When i think about the transformer architectures, they are also a kind of design to define agreements or correlations. I mean, there've been your scientist saying for years and years that they'd be crazy not to use ta spike time unless people like abelas ha talk about sinfire chains. It would be very satisfying to show that when you start learning particular consequential data, like auditory data, then they really do make use of the spike times sensible way. But if i think about spiking, let's say i try to play devil's advocate here and try to maybe argue against a strong believin in spiking. No, we do devolve wheels, or we did. Now, you
Transcript
Play full episode
