Serendipper wrote:What are we trying to discover here? If you desire to create a robot that can walk and you've no idea how, then the best strategy is to create an evolutionary algorithm to let the robot learn what you have no idea how to teach. In the design of the algorithm, there can be no plan because you cannot presume you know how a robot should execute commands in the proper order and timing to produce a stable gait. So the idea is to not make any assumptions and let "nature" take it's course.
You have to let what works be the test for what is right and not what you dictate to be right from the beginning. If you could dictate what is right from the beginning, then there would be no point in the evolution. If you knew how to program a robot to walk, then just program it; no need for algorithmic learning.
You see?
It's not that the universe is a chaotic and random mess, but a dark place where we must build our own eyes and fumble around learning as we go. If it could have been understood from the beginning, then there would be no point to any of this. Life is the discovery of the unknown; not rediscovering what is already known.
Sure, and the robot will learn because of regularities - gravity, objects no being able to occupy the same space, downhills causing this set of issues, uphillls these, and so on. The robot will also be perceiving its environment and not just getting random perceptions.
Put a robot in
'random existence fluctuating everywhere and with chaotic random manifestations of being'and it will not learn how to get anywhere or really, it might get somewhere, but it will not be able to learn how to do this. Like Skinner's superstitious pidgeons the robot might keep trying something, but the results would be random.
Perhaps we are talking past each other somehow.