In the two examples below we show the results of policy evaluation steps for a simple MDP. The values of the states after convergence are shown in the grid. In the left grid the original policy was "Alway Go Right" and in the right grid the original policy was "Always Go Forward". -10.00 100.00 -10.00 -10.00 1.09 -10.00 -10.00 100.00 -10.00 -10.00 70.20 -10.00 Go Right -10.00 -7.88> -10.00 -10.00 48.74 -10.00 -10.00 -8.69 -10.00 -10.00 33.30 -10.00 Go Forward a) (10p) If you perform policy extraction for each of the two cases what policies will you obtain? b) (10p) How do you explain the fact that a bad policy led to a desirable outcome? Why are the values of states different from -10 (left grid) and 100 (right grid)?