Neural networks are notoriously hard to debug, one reason AdaBoost became extremely intriguing (human-understandable weight decisions). The low-power capability is extremely interesting from a surveillance-state perspective, as now you can have behavior detection in-camera.
I have to agree with LeCun that this isn’t groundbreaking (except power consumption IMHO). Sejnowski’s claim that “GPUs will never be on cellphones” seems a bit disingenuous. We’ll see how things fall out.
New IBM chip inspired by human brain
Aren’t they all?
This topic was automatically closed after 5 days. New replies are no longer allowed.