Which two instructions in the "programming language" of the 2011 HW would be the most difficult for robots to follow?
Intsructions like 7 and 12 would be the most difficult to follow for a robot. This is because rotating a part(joint) while holding
a pencil would also require recalculation of force applied to the point of contact to balance the torque change (which humans do
unconsciously)
The robot following the learning paradigm as in Kalakrishnan is clearly gaining some expertise. Which aspects of the execution may
be called implicit or automatic, and which aspects may be more explicit? What could be the "chunks" in this structure?
The premise of "implicit learning" in robots can be debated. One may argue that at every point in time the robot is driven by the goal
of picking up the pencil and hence is paying full attention to it. But at the same time machine learning can be considered as implicit
since it results in the development of a abstract representation. However even this knowledge can be passed on from one machine to another
and thus the premise that implicit knowledge is not transferable (as it cannot be articulated), which holds in humans does not hold in machines.
But for the sake for argument let us consider machine learning to be implicit and other aspects of learning in robots to be explicit.
In this case the calculation of forces to be applied and point of application of forces can be considered as implicit whereas the detection of pencil,
finding the door knob, etc from the environment as explicit.
The chunks in the structure can be the image (or its representation) of the pencil, door knob, door. Even the knowledge of point of applicatio of
force and the corresponding force can be considered as a chunk.
Comment on whether human learning may also be following similar "reward" based processes? Consider the learning process for the fire-fighting
expert who knows how to fight complex fires.
Human learning, in my opinion, is by and large driven by reward/feedback. Feedback may be both in the form of positive reinforcements and/or
negative reinforcements. A baby stops going towards the fire after getting his hands burnt. But feeback need not only be explicit. Physiological
or psychological rewards can be just as effective in guiding the process of learning in an individual.
Considering the case of a fire-fighter. He learns to identify the points where the process of dousing fire will be most effective.
This learning is driven by the reward of putting out the fire, which in turn is reward in the act of dousing fires which is his occupation.
Even in the larger frame of reference it is driven by the reward of saving lives (both his and others).
In a way humans have acknowledged the role of rewards in learning. Otherwise teachers would grant extra marks for good handwriting and fathers would
not promise material gifts on doing well in class.