Q1. Which two instructions in the "programming language" of the 2011 HW would be the most difficult for robots to follow?
All the instructions in the "programming language" of the 2011 HW are difficult but I think, the two most difficult ones are -
Q2. 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?
Kalakrishnan writes in his report says that the control policy is represented as a discretized trajectory of desired end-effector positions, orientations, forces and torques. This means that end-effector positions, orientations, force and torques constitute the problem space variables and the robot is continuously manipulating these only by interacting with the environment to train itself and this fact is evident if we see the Learnt force/torque control policy graph, where the robot does many trials to come up with a force control policy for the pen-grasping task and this policy. Clearly, the policy is derived through interaction with environment (in this case pen).
Now,it is important to note that one features of implicit knowledge is that Implicit knowledge is acquired relatively directly from the environment. So, I think, in the experiment, this is an example where implicit learning involved in the process.
About explicit learning, I think that the features like initial positioning of the hand comprise the part on explicit learning. Each time a new trial starts, there is a need to refer to the initial set of commands to arrive at a starting posture also one should note that no interaction with environment is involved in the process.
Also,Chunks can be the dimensions in the low-dimensional embedding obtained which represent an inter-relation between the variables(end-effector positions, orientation, force and torque) which must hold for favorable execution.
Q3. 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.
Yes, human learning can also be following such "reward" based processes. It is because there can several types of reward like positive and negative reward or internal and external rewards. Some of these are very hard to be recognized but they might be motivating the individual to learn. Like, in the fire fighter example, the reward may be fast and effective removal of fire. Clearly, this is not providing the fire fighter any physical or material advantage but it might be motivating him to learn the more efficient ways to fight the fire.