Robotic Nation Evidence

Animals have been useful to humans for many centuries, yet many animals seem to get by in the wild with limited visual processing power. If a human brain requires 10^14-10^17 operations per second to simulate, then the brain of a stout infantfish, the smallest known vertebrate, should only require 10^7-10^10 operations per second to simulate if it is as complex for the fish's size as a human's brain is for a human's size. Assuming standard phylogenetic relationships, it should be another 100 fold simpler than that. This fish manages to see, though I doubt anyone knows how well, but it presumably does it with the same basic neurological setup as other fish, some of which see reasonably well. It looks to me like if we knew how we could simulate a whole vertebrate, visual cortex and all, on a laptop, or in some small cheap piece of custom hardware, and could then program it for various functions. Hardware doesn't seem to be critical.
 

I work in automotive engineering. I think general purpose robotics will require more than just software and CPU. A general purpose robot will need to be a robust and durable device. Especially once you begin to rely on it. I think one of the challenges will be purely mechanical. A general purpose robot will have lots of moving parts, many designed at the limits of manufacturing and material limits. Once the software and CPUs are there, there will still be a lag as money from sales pays for the learning to make them robust.