Designing Customized “Brains” for Robots
news.mit.edu
news.mit.edu
So besides the lampoon-ability of this article as inventing new words for old concepts (hardware acceleration) which I am not qualified to lob (and therefore only hint at ;) ), how much more software efficiency is out there to 'mine' out of standard CPU/RAM/Storage Von Neumann-style computing devices?
Some people claim that moving things from Electron to native would be a huge speedup, and I don't disagree with that, somehow a Windows 98 box provided 95%+ of today's computing experience (I can't speak for older things like Amigas, Apple ][s/Lisas/etc) at like 1% of the computing power.
But what do I know, maybe all the software efficiency in the world has been wrung out and there is no more low-hanging fruit? And/or that is the case for robotics but not for general computing.