re: fruitfly navigation and avoidence?
The MERS go autonomous on carefully selected stretches.
Remember the fiasco last year in the DOD sponsored robotics contest navigating thru Death Valley? No vehicle made it even one mile. That included CM, MIT and many other vaunted AI labs. Robotic vision and nav systems don't work well except in artificially restricted college lab setings, as they require static well defined obsticals. Meanwhile, a fruit fly does 3-D real time navigation, obstacle avoidance and sometimes even flyswatter avoidance. No AI system comes anywhere close to this performance. We do not have viable theories to construct algorithms to navigate in this manner, even in non real time. In other words, it's not just # of computes.