I think this is finally the right direction for AI to move in. For a while it has seemed like AI researchers were too enamored with the idea that humans were inborn with certain ideas like language and motor capability, and that they should try to replicate that inbuiltness in a machine.
My take is that humans are born with little more than sensory perception, and a mechanism for recognizing patterns and cause and effect (this is known as memory).
If you observe the actions of young children closely, I think you'll find that they are consistent with the actions of someone trying to conduct experiments in the world beginning with very little information.
When my younger siblings were babies they would do the old routine of "holding my hands up around my face and waving them around." I think this is the process of forming the deduction "when I fire my motor neurons something appears in my sight." And eventually this association becomes so strong that the child recognizes the hand as something he directly controls, i.e. the self.
Very cool idea of using a low cost camera sensor to provide feedback on how well the robot is moving. I would guess (not having seen the code) that Matt is using the metric of how far the robot can move to tune parameters for controlling the legs (genetic algorithm or genetic programming).
Isn't it a bit incorrect (or at least irrelevant) to say that the robot is running Ubuntu?
Maybe things have changed drastically in the last few years, but the last time I checked, Ubuntu was a pretty installer and package manager for gnu/linux.
My take is that humans are born with little more than sensory perception, and a mechanism for recognizing patterns and cause and effect (this is known as memory).
If you observe the actions of young children closely, I think you'll find that they are consistent with the actions of someone trying to conduct experiments in the world beginning with very little information.
When my younger siblings were babies they would do the old routine of "holding my hands up around my face and waving them around." I think this is the process of forming the deduction "when I fire my motor neurons something appears in my sight." And eventually this association becomes so strong that the child recognizes the hand as something he directly controls, i.e. the self.